CS432 Handouts PDF

Booklet of Network Modeling and Simulation
CONTENT
Course code CS432
Credits 3+0
Instructor
Dr. Ali Hammad Akbar
Lecturing style
Video lectures of short
duration (5-7 minutes)
Evaluation
Quizzes
Assignment
Simulation modules
Mid-term and final

The complex world of networks
Networks are many and diverse. Can we simplify it? Or at least the discussion of it?
Need for NeMS

Let us explore the need and motivation to perform Network Modeling and Simulation (NeMS) by
looking at the technology landscape. The landscape consists of the people, technology and their
relationships.
Technology Landscape
1. Communications systems: Evolving rapidly
2. User demands: High performance networks
3. Service providers: Rapidly expanding their network infrastructure
Network researchers face the protocol war by developing new communications techniques,
architectures and capabilities. Equipment vendors are releasing new devices with increasing
capability and complexity. Technology developers and OEMs are developing NG equipment.
Network designers and developers are working on how to satisfy the QoS demands of users amidst
emerging technologies and techniques viz a viz legacy counterparts? Network Engineer in
operations is thinking about what is the right approach to solving problems? Do I buy latest device
from company X that claims to solve all my problems? Do I replace underlying technology of my
system with the latest generation? Next Generation Network Architect wonders how do I know how
this new approach will interact with already existing protocols? How do I build confidence in the
utility of this approach without producing and deploying the technology? Is there one solution?
Actually not! There are various ways to answer and satisfy the goal seekers. These include
 Prototyping & empirical testing
 Trial field deployment
 Modeling and Simulation (M & S)
 Analysis
The order represents decreasing costs but increasing abstraction. It is upto the network engineer to
trade them off.
What is NeMS?
Network Modeling and Simulation is often considered a single term. In reality, it is not! Simulation
is the imitation of behaviour of real-world system or Computational re-enactment according to rules
described in model. Whereas modeling is a step that precedes simulations. Together they form an
iterative process approximating the real world systems. Model is the logical representation of a
complex entity, system, phenomena or a process. In communications, network model could be
analytical representation, mathematical form as a state Machine or closed or approximate form.
Computer simulation is the execution of computer software that reproduces behavior with a certain
degree of accuracy to provide visual insight. It is basically a template on which a computer program
runs. It has
 Inputs
 Outputs
 Behaviour
Formally, simulations are pieces of computer software that implement algorithms, take inputs and
give outputs.
The model definition could be

 Descriptive
 Analytical
 Mathematical
 Algorithmic
The computer models can be described into various types

 Stochastic vs Deterministic
 Continuous vs discrete
 Steady state vs Dynamic
 Local or Distributed
 Linear or nonlinear
 Open or closed

These models must be applied according to the perspective. It is important to Model only what you
understand. Likewise, understanding your model is equally necessary. Model what you need & no
more so that it is neither underdefined nor overdefined.
Simulation Building Process

Consider a one hop communication scenario between two wireless notebooks connected through a
WiFi AP. The simulation entities would include wireless computers and their packets (multiple
instances), WiFi AP (single instance), and a traffic generator (single instance) that creates wireless
computers and their packets. The states of the system would include WiFi AP (idle or busy). Each
computer generates a number of packets with each packet successful/failed. The events would
include wireless computer creation, packet generation, wireless AP activity. Queues would be
needed to schedule the events. These would contain frames waiting in output queue of wireless
computer and frames (packets) at WiFi AP input queue. To make the simulation exciting, dynamic
and insightful, randomization has to be performed. These would include random realizations of
packet lengths, no. of frames per wireless computer, no. of wireless computers, BER (and PER) and
packet drop ratio in WiFi AP input queue. It is further needed to distribute various entities and
events. The distributions could include Uniform and Gaussian etc for packet lengths and no. of
frames per wireless computer.
How Does The Simulation Run?
Following steps are executed during the running of the simulation.
 Inputs created/initialized
 Events of transmission, reception and noise occur
 Randomness causes queues to behave and err
 Packet successes/failures
 Simulation logs are compiled and presented as the output in desirable formats
Components of a simulator
 A self-contained program
 Event queue
 Simulation clock
 State variables
 Event routines
 Input routine
 Report generation routine
 Initialization routine
 Main program
Types of simulations
 Monte Carlo simulation
 Trace driven
 Discrete events
 Continuous events
When to simulate
 Analytical model not feasible (complex)
 Analytical model not possible (too simple)
 Simulate to verify analysis
 Otherwise simulations are unnecessary
When not to simulate
 Analytical model gives good enough representation
 Simulation takes months
 Simulation is expensive
 Simulation is non-scalable
General mistakes
 Inappropriate levels of details
 Improper selection of programming language
 Unverified models
 Improper initial conditions
 Short run times
 Poor random number generators
 Inadequate time estimate
 No achievable goals
 Incomplete mix of essential skills
 Inadequate level of user participation
 Inability to manage simulation project
Inappropriate levels of details

 Include what is relevant
 Too fine simulations computationally heavy
 Many interdependent parameters
 Difficult to assess their interplay
 Tip: Necessity & sufficiency
Improper programming language

 Scope & type of simulation determine best choice
 Object oriented vs. procedural
– Types/diversity of simulation parameters
 Interpreted vs. compiled
– Machine dependence
– Speed
–
Unverified models
 Programming is non trivial
 Semantic mistakes
 make simulations get
 Wrong results
 Misleading results
 Modular verification a must
Improper initial conditions

 Initial condition not steady state
 Often a late realization
 Surprisingly wrong results
 May never converge
Short run times

 Strong dependence on Initial conditions
 Don't achieve true
 steady state
Poor random number generators

 Lacking pseudo-random sequence leads to predictability
 Wrong choice of seed value could cause inadvertent correlation between processes
– Use celebrated RNGs
Inadequate time estimate

 Models overstate the simulations
– Implementations get delayed
 Software development life cycle must assess model complexity
No achievable goals
 Goals not defined
– Tangible output analysis
– Logs and trace files
 Goals are unreal
– Affects simulation complexity and implementation
Incomplete mix of essential skills

 Domain knowledge
 Statistics
 Programming
 Project management
 Past experience
Inadequate level of user participation

 From modeling to implementation
 UI design
 Output analysis
Inability to manage simulation project

 Simulations are not monolithic
 Need software engineering tools
– Multivariate design
– Code management
– Track changes
Simulation inaccuracies
 Over reliance on link budget methods for abstraction
 Overly simplistic modeling of radio layers
Over reliance on link budget methods for abstraction

 Link budget losses overly static
– Fair enough for steady state analysis
– Dynamic analysis not possible
 Results are misleading
Overly simplistic modeling of radio layers

 Lowest layer often ignored
– No bit level BER & delay
 Often the Achilles heel
 Wrong results in highly dynamic use cases
Development of Systems Simulation
A “Still I am not dead yet!” scenario
Available Not available

h = height (feet) Mass of object
t = time in motion (seconds) Air resistance
v = initial velocity (feet per second, + is up) Location of object
s = initial height (feet)
a = acceleration (feet per second per second)
/* Height of an object moving under gravity. */

/* Initial height s and velocity v constants. */
main()
{
float h, v = 100.0, s = 1000.0;
int t;
Development Process
 Problem formulation
 Data collection & analysis
 Simulation development
 Model validation, verification, & calibration
 “What-if” analysis
 Sensitivity estimation
Problem formulation
 Identify controllable and uncontrollable inputs
Data collection & analysis

 What to collect
 How much to collect
 Cost and accuracy trade off
Simulation development
 Codify, codify and codify!
Model validation, verification, & calibration

 Validation
 Is it the right system?
 Emulates real phenomenon

 Verification
 Are we building the system right?
 Implementation must correspond to the model
 Calibration
 Parameter estimation
 Tweaking/tuning to ensure that simulated data follows real data
“What-if” analysis
 Performance measures with different inputs
Sensitivity analysis
 Relative importance of different parameters with respect to output
 Even with respect to each other
Life cycle of Simulation Development
Recommended Text and References
NeMS contents cover

 Well known mathematical models, equations and forms
 Widely used simulation tools and code reusability
 Their inter-relationship
NeMS contents don't cover

 Mathematical derivations from scratch
 Programming dexterity
Uptill now
Basics of NeMS
 Mohsen Guizani et al, “Network Modeling and Simulation” John Wiley , 2010.
Basics of NeMS
 Jack Burbank et al, “An Introduction to Network Modeling & Simulation for the Practicing
Engineer” John Wiley , 2011.
Basics of NeMS
 John A. Sokolowski & Catherine M. Banks, “Modeling and Simulation Fundamentals” John
Wiley , 2010.
Next Roadmap
 TicToc tutorial
 OMNET++ Manual
 Website: https://omnetpp.org
 INET Framework for OMNeT++
 OMNET++ Wiki
 Mixim Sourceforge Page
Introduction to OMNET++
What is OMNET++
 Objective Modular Network Testbed in C++
– Simulation kernel
– Component-based simulation library
 A framework, not a simulator
 Designed to create & simulate any network
Simulation Kernel
Getting a free copy

 www.omnetpp.org
 Download the latest release (4.6 in our case)
“Omnetpp-4.6-src-windows.zip”
 Complete folder
– C++ compiler
– CMD line build environment
 Download source code
Compile & Install

 Compiling and installing on Windows self-contained
 Enter OMNeT++ folder that you unzipped
 Run the file called Mingwenv.cmd
Minimum GNU environment for Windows

Compilers provide access to functionality
Of Microsoft C runtime and some
Language-specific runtimes
Compile & Install

 When terminal appears, enter the commands
/.configure
Make Build process produces
Debug and release
Binaries
Debug is elaborate
But slow
Release is optimized
& fast
Debug mode
 Does not optimize the binary it produces
 Source code and generated instructions relationship is complex
 Allows accurate breakpoints setting
 Allows code step-through one line at a time
 Compiled with full symbolic debug information
Release mode
 Enables optimizations
 Generates instructions without any debug data
 Lots of code could be completely removed or rewritten
 Resulting executable may not match with written code
Running first time

 OMNeT++ comes with an Eclipse-based Simulation IDE
 Type omnetpp
Select the default workspace
 A workspace is a logical collection of projects
 A workspace called p2p may contain only peer to peer applications

Design of OMNET++
Model Structure
 Model consists of modules
 Modules communicate with message passing
 Modules are C++ files
– Implement simulation class library
– Run in simulation kernel
 Module types
– Simple (active modules)
– Compound
 Simple modules can be grouped into compound modules and so forth

 Modules communicate through gates (connections)
– Directly between modules or through intermediaries
No. of hierarchy levels not limited
 Gates
– Gates
– Input output interfaces of modules
– Allow message passing
– Linked via connection (TPROP, RDATA, BER) Input output interfacing
1. Define
Module
Types
2. Instantiate
3. Network implements system model them
 Channels
– Connection types with specific properties
– Reusable at several places
– Standard Host talking to another Standard Host via an Ethernet cable
 Message; tuple (time stamp, arbitrary data, ...)

 Network; A compound module with no external gates
Module Parameters
 Pass configuration data to simple modules
 Define model topology
 String, numeric, boolean
 Constants, random numbers
 Expressions as references
Internal architecture of OMNET++
OMNeT++ simulation programs possess a modular structure.

Model Component Library
 Consists of the code of compiled simple and compound modules
Simulation Kernel & SIM Class Library

 Modules are instantiated and concrete simulation model is built by simulation kernel
 SIM covers most of the common simulation tasks through classes
– Generate random number (distributions)
– Queues (F IFO, priority)
 Messages (hold arbitrary data structures)
 Routing (explore topology, generate graph data structure)
Envir, Cmdenv and Tkenv Libraries

 Simulation executes in an environment
 Defines and determines
– Where input data come from
– Where simulation results go to
– What happens to debugging output?
– Controls the simulation execution
– How model is visualized
What is NED Language?
 A network description language
 Creates network topologies in OMNeT++
 You create alternately create topology graphically as well
 Correspondingly NED source code is automatically generated
Typical Ingredients of NED description

 Network definitions
 Compound module definitions
 Simple module declarations
Network Definition
 Network definitions are compound modules
– Self-contained simulation models
Simple Module Declaration

 Describes the interface of modules
– Gates
– Parameters
Compound module definitions
 Declaration of external interface
– Gates
– Parameters
 Definition of
– Sub modules
– Their interconnections
Let us create a topology called My_Network using Graphical Editor
Network name
=
My_Network
Module name
=
My_Module
Compound module
name=
standardHost
Inside
standardHost
\
More About NED Language

 Inheritance
 Modules and channels can be subclassed
 Derived modules and channels may add
– New parameters
– Gates
 Similarly compound modules may add
– New parameters
– Connections
Example
Simple
Compound
GenericTCPClientApp BaseHost
Module
Derived
(Extended) FTPApp BaseHost + WebClientApp
Module
Interface instantiation
 Module and channel interfaces can be used as a placeholder
– where normally a module or channel type would be used
 Concrete module or channel type determined
– At network setup time by a parameter
Example:
MobileHost compound module
Design time IMobility
Run time
ConstantSpeedMobility RandomWayPointMobility
Packages
 Addresses name clashes between different models
 Simplifies specifying which NED files are needed by a specific simulation model
Example:
package book.simulations;
Package is a mechanism to organize various classes and files. The simulation project
inside of OMNeT++ is called "Book" and this NED file is found in the "simulations" folder of
the
Project.
Configuring OMNET++ Simulations
Separation of Model and Experiments

 Always good practice to try to separate different aspects of simulation
 Model topology
– NED file
– MSG file
 Model behavior
– C++ code
 Provides cleaner model
Configuring simulations
 How to capture the effect of different inputs?
– Run to run variables
 C++ and NED code do not have such variables
 INI files provide a mechanism to specify these parameters
– omnet.ini
INI File Syntax

 Basically an ASCII text file
 Consists of
– Key-value pairs
<key>=<value>
INI File Editor

 INI File Editor lets the user configure simulation models for execution
 Both form-based and source editing
INI File Editor

 Considers all NED declarations
– Simple modules
– Compound modules
– Channels, etc
 Fully relates this information to the INI file contents
 Editor knows which INI file keys match which module parameters
Example omnet.ini
My_Network
wildcarded as **
[General] No of Apps
network = book.simulations.My_Network
#We will make standardHost a TCP Session Application in
order for it to communicate#
**.standardHost.numTcpApps = 1
Example
Application Name
[General]
#We will make standardHost a TCP Session Application in
order for it to communicate#
**.standardHost.tcpApp[0].typename = "TCPSessionApp"
**.standardHost.tcpApp[0].connectAddress = "standardHost1"
**.standardHost.tcpApp[0].connectPort = 1000
#We will make standardHost1 a TCP Echo Application, this
means that it
will send #an echo packet once it receives a packet.
**.standardHost1.numTcpApps = 1
**.standardHost1.tcpApp[0].typename = "TCPEchoApp"
**.standardHost1.tcpApp[0].localPort = 1000
**.standardHost1.tcpApp[0].echoFactor = 3.0
#**.ppp[*].queueType = "DropTailQueue"
[General]
#We will make standardHost a TCP Session Application in order for
it to communicate# Who to
connect
**.standardHost.numTcpApps = 1 with whom
#We will make standardHost1 a TCP Echo Application, this means
that it
will send #an echo packet once it receives a packet.
Which port to
[General] connect to
#We will make standardHost a TCP Session Application in order for it
to communicate
#We will make standardHost1 a TCP Echo Application, this means
that it will send #an echo packet once it receives a packet.
#**.ppp[*].queue.frameCapacity = 10
#**.eth[*].queueType = "DropTailQueue"
Reply size =
Echo Packet* EF
[General]
#We will make standardHost a TCP Session Application in order for it to
communicate
#We will make standardHost1 a TCP Echo Application, this means that it will
send #an echo packet once it receives a packet.
Queuing behaviour
[General]
communicate
Buffer Size
[General]
communicate
Example
Building Simulation Programs
Using GUI Project Builder

 Initial build takes longer on indexing before building the project
 Dependency generation in the generated make files
– Classes, functions, methods, variables, macros
Using Mingwenv
 Once you have the source files ( *.ned, *.msg, *.cc, *.h) in a directory
– Change the working directory to there
 Type
$ opp_makemake
 This will create a file named Makefile
 Type
$ make
Your simulation program should build
A makefile is used to tell the compiler which source files you want to compile.
It'll also do things like name your executable and place it in a specific location.
Where to next!
Running Simulations
What is Simulation Run?

 Launch the built project make file
OMNET++ IDE Features

 Single runs
 Batch runs
 Run numbers
 Graphical mode (Tkenv)
 Command mode(Cmdenv)
 Simulation configuration
 Recording event logs
 Debug support
Quick Run
 In Project Explorer, select a project
 Clicking Run button on the toolbar
 Runs vary
– Folder
• Runs if single ini file present
– ini file
• Use this as the main ini file
– NED file
• Scan for available ini file
Launch Configuration
Run omnet.ini
From /queuenet
Run queuenet Launch

numbe Configuration
r
R=0
One
omnet.
One or more
ini files
Directories where
the NED files
are read from
Animation and Tracing

OMNeT++ is capable of
 Animating
– Flow of messages on network charts
 Reflecting
– State changes of the nodes in the display
 Animation is automatic
 No programming need for simulating engineer
 Suitable network simulations
– Rarely need fully customizable animation capabilities
Simulation Tracing
 Simple modules may write textual debug (trace) information like printf()
 OMNET++ provides Module output window
– Special window to display output stream
 Eases following the module execution
Simulation Object Inspection
 An object inspector is a GUI window associated with a simulation object
– Displays contents and properties
 Three types
– Network Display
– Log Viewer
– Object Inspector
Tkenv
Tkenv is a graphical runtime interface for simulations
 It provides
– Network visualization
– Message flow animation
– Log of message flow
– Display of textual module logs
 Inspectors
 Visualization of statistics
– Histograms, etc. during simulation execution
 Event log recording for later analysis
Tkenv in action
Object inspector Timeline

Future Events Set (FES) on log scale
Network
display Log
viewer
Organizing and Performing Experiments
Need for organizing experiments

Stuart Kurkow, “MANET Simulation
Studies:
The Incredibles,” ACM’s Mobile Computing and Communications
Review, 9(4):
50-61, 2005
Repeatable
 Fellow researcher should be able to repeat
Unbiased
 Results must not be specific to scenario used in experiment
Rigorous
 Scenarios & conditions for experiments must be truly representative
Statistically sound
 Experiments results must not violate mathematical principles
Relationship between terminologies

How to organize experiments
Model
 The executable
 (C++ files & external libraries + NED files
 Invariant for the purpose of experimentation
 INI file not part of model
Study
 One or more experiments to investigate a phenomenon
 Usually many experiments
 One or more models
Experiment
 Exploration of a parameter space on a model
 Only and only one model
Measurement
 A set of simulation runs on the same model with same parameters
 Characterized by INI file
 But with different seeds
 May involve replication for averaging out
Replication
 One repetition of a measurement
 Replication can be characterized by the seed values it uses
Run
 One instance of running the simulation
 Characterized by exact time date
 computer (host name)
Example
Handover
optimization
No of hosts
= 10
Load
= 3.8
Effect of
Mobile
No. of hosts
IPv6
Traffic load
nodes
Sequence Charts
Event Log Tables

 An event log file contains
― Tabulated log of messages sent during simulation
 Between modules
 Self-messages (timers)
― Event details that prompts such sending or reception
 User can control
― Amount of data recorded from messages
― Start/stop time
― Which modules to include in the log
Event Log File Creation (1 of 2)

 Type
$ record-eventlog = true
 Output placed in
/results directory
 Filename
${configname}-${runnumber}.elog
Using INI file event log configuration
Record event
Sequence Chart
 Displays event log files in a graphical form
• Helps focus on causes & consequences of events/messages
• Helps users understand
― Complex simulation models
― Verify implementation for desired behavior
Compound module
Understanding the legend: Axis
Initialization event
Compound module
Axis
Event number
Axis with attached

Vector data
Module full path

as axis label
Self-message Self-message
Message processing Message send
Processing event
Event
Parts of Sequence Charts
What is Timeline?
 Simulation time mapped onto the horizontal axis
 Various ways
– Intervals between interesting events often of different magnitudes
 Example
– MAC (ms)
– Higher layers (ms)
Types of Timeline
 Linear: simulation time proportional to distance measured in pixels
 Event number: event number proportional to the distance measured in pixels
 Step: distance between subsequent events is same
 Nonlinear: distance between subsequent events is nonlinear function of simulation time
between them
Interpreting Sequence Charts

 Zero Simulation Time Regions
 Gutter
 Events
 Messages
 Displaying Module State on Axes
Zero Simulation Time Regions
Gutter
Events Processing
Messages
Displaying Module State on Axes

TicToc Tutorial
TicToc with 2-nodes
• Two nodes, Tic and Toc
• One node initializes by sending a message to the other
• Every time a node receives the message
― Sends it back
• Continue indefinitely
― Till user stops
Creating an empty project
• Open the OMNeT++ IDE
• Navigate to File | New | OMNeT++ Project
• Enter a Name for the project
• Next
Select the Tictoc example file in the Examples folder

You have created Tictoc example project
Opening NED file

• In newly createdproject, navigate to the simulations folder in the Project Explorer
• Open Tictoc.ned
Understanding toctoc1.ned
Opening Simple Module

• Open project explorer
• Open src folder of this project
• Open Txc.ned
Understanding Txc.ned
Opening Simple Module

• Open project explorer
• Open src folder of this project
• Open Txc.cc
Understanding Txc.ned
Understanding omnet.ini
Compiling & Running on Tkenv

Extending TicToc
Refine graphics &

 Tictoc2.ned
Add debugging output
 Txc2.cc
Tictoc2.ned
Tkenv output
Extending TicToc
Add State Variables

• Add a counter as a class member to the module
• Delete the message after 10 exchanges
• Txc3.cc
Txc3.cc
Output:
Adding parameters
Adding parameters
• Add input parameters to the simulations

– Count = 10 now into a parameter that the user can define
• tictoc4.ned
• Txc4.cc
• Omnet.ini
Boolean parameter (decides if module should send out first message in its initialization code)
• tictoc4.ned
• Txc4.cc
• Omnet.ini
tictoc4.ned
Using Inheritance
Using Inheritance
 What is different between tic and toc?
― Parameter values
― Display string
 Inheritance allows to create a simple module
― Then derive modules from ittictoc5.ned
tictoc5.ned
Modeling processing delay
Modeling processing delay

 So far, no processing delay in tictoc
 We need timer in
 Tictoc module to send itself “Event” message
 tictoc6.ned
 txc6.c
Strategy
tx6.cc
Output:
Random numbers and parameters
Random numbers and parameters

 Introduce random numbers in simulation
– Randomly lose packet
– Change delay from 1s to a random value
 txc7.cc
 tictoc7.ned Or omnetpp.ini
txc7.cc
Timeout, cancelling timers
Timeout, cancelling timers
 Getting closer to real world working protocols
 Stop-and-wait protocol
 txc8.cc
 tictoc8.ned
 omnetpp.ini
Strategy
txc8.cc
Output
Retransmitting same message
Retransmitting same message)

 So far we used “tictocMsg”
 It was created afresh everytime
– At tic
– At toc
 In reality, original packet needs to be retransmitted
 Solution: Keep a copy with tic
 txc9.cc
 tictoc9.ned
 omnetpp.ini
Strategy
 Create two new functions
 Conditionally call them in tic and toc
Txc9.cc
generateNewMessage()
sendCopyOf(cMessage *msg)
More than 2 nodes
More than 2 nodes

 Create several tic modules
 Connect them into a network
 One of the nodes generates a message
 Others toss it around in random directions
 Until it arrives at a predetermined destination
 tictoc10.ned
 omnetpp.ini
 txc10.cc
Tictoc10.ned
Output
Channels & inner type definitions
Channels & inner type definitions

 With growing topology
– We can improve connection section
 tictoc11.ned
 omnetpp.ini
 txc11.cc
 Connections with same delay parameter can be typified as channel
 Such channel can then be replicated between gates
Tictoc11.ned
Using two-way connections

Using two-way connections
 So far, each node pair is connected with two connections
 Two-way connection can reduce coding size
 tictoc12.ned
 txc12.cc
 omnetpp.ini
 We define two-way (inout) gates Instead of in and out gates
Tictoc12.ned
Defining our message class

Defining our message class
 Instead of hardcoding tic[3], we need flexibility
 Draw out a random destination
 Add Destination address
 tictoc13.ned
 txc13.cc
 tictoc13.msg
 omnetpp.ini
Strategy : Avoid boilerplate code writing
Strategy :Avoid boilerplate code writing
\
Txc13.cc
\
Output
Displaying no. of packets sent/received

 No. of messages at each node
 tictoc14.ned
 txc14.cc
 tictoc14.msg
 omnetpp.ini
Txc14.cc
Object Inspector in Tkenv
Adding statistics collection
Adding statistics collection

 When packet traverses multiple hops, it becomes important to collect network statistics
– Average hop count
– Max, min etc
 tictoc15.ned
 txc15.cc
 tictoc15.msg
 omnetpp.ini
Strategy
Txc15.cc
Visualizing output scalars & vectorsVisualizing output scalars & vectors
 OMNET++ allows to visualize outputs of scalar and vector files

– Filtering
– Processing
– Displaying
Analyzing Results
What is Simulation Analysis?

 Analyzing simulation results is lengthy and time consuming process
 Result are recorded as scalar values, vector values and histograms
 User can apply statistical methods
– Extract the relevant information
– Draw conclusions
Analysis File (.anf)

 A file that automates the steps to analyze the results
– Loading result files
– Filter them
– Transform data
Creating Analysis File
Using the Analysis Editor

\
Datasets
 Describe a set of input data, the processing applied to them and the charts
 Displayed as a tree of processing steps and charts
 Nodes are used for
― Adding and discarding data
― Applying processing to vectors and scalars
― Selecting the operands of the operations
― Content of charts, and for creating charts
Editing Datasets
What is Compute Vectors?
 Both Compute Vectors and Apply to Vectors nodes compute new vectors from other vectors
What is Compute Scalars?

 The Compute Scalars dataset node adds new scalars to the dataset whose values are
computed from other statistics in the dataset
Finally we are done!
Computation Examples 1
Bit rate
 Assume several source modules in the network that generate CBR traffic
 Parameterized with packet length (in bytes) and send interval (seconds)
 Both parameters saved as scalars by each module (pkLen, sendInterval)
 To use the bit rate for further computations or charts
– Add a Compute Scalar node with the following content to create an additional bit
rate scalar for each source module
Value: pkLen*8/sendInterval
Name: bitrate
Throughput
 Assume several sink modules record rcvdByteCount scalars, and simulation duration is
saved globally as the duration scalar of the top-level module.
 We are interested in the throughput at each sink module
 We need to refer to the duration scalar by its qualified name (prefix it with the full name of
its module)
 rcvdByteCount can be left unqualified
Value:8*rcvdByteCount/Network.duration
Name: throughput
Total Received Bytes

 We are interested in the total number of bytes received in the network
 We can use the sum() function
 We store the result as a new scalar of the toplevel module, Network.
Value: sum(rcvdByteCount)
Name: totalRcvdBytes
Module: Network
Bytes Received by Hosts
 If several modules record scalars named rcvdByteCount
 We are only interested in the ones recorded from network hosts
 you can qualify the scalar name with a pattern
Value: sum(**.host*.**.rcvdByteCount)
Name: totalHostRcvdBytes
Module: Network
Average of Peak Delay

 If several modules record vectors named end-to-end delay
 We are interested in average of the peak end-to-end delays experienced by each module
 We can use the max() function on the vectors to get the peak
 Then we need mean() to obtain their averages
Value: mean(max('end-to-end delay'))
Name: avgPeakDelay
Module: Network
Computation Examples 2
Packet loss per client-server pair

 3 clients (cli0, cli1, cli2) and 3 servers (srv0, srv1,
srv2) in the network
 Each client sends datagrams to the corresponding server
 Packet loss per client-server pair computed from the number of sent and received packets.
 We use the i variable to match the corresponding clients and servers.
Value: Net.cli${i={0..2}}.pkSent -
Net.srv{i}.pkRcvd
Name: pkLoss
Module: Net.srv${i}
Total No. of Transport Packets

 When input scalars are recorded by different modules
– We need the host variable to match TCP and UDP modules under the same host
 Compute the total number of transport packets (the sum of the TCP and UDP packet
counts) for each host
Value: ${host=**}.udp.pkCount +
${host}.tcp.pkCount
Name: transportPkCount
Module: ${host}
Modules with largest RTT

 A network has various modules recording ping round-trip delays (RTT)
 We want to count the modules with large RTT values (where the average RTT is more than
twice the global average in the network)
 We need to do it in steps
Step 1:
Value: mean('rtt:vector')
Name: average
Step 2:
Value: average / mean(**.average)
Name: relativeAverage
Step 3:
Value: count(relativeAverage)
Grouping: value > 2.0 ? "Above" : "Normal"
Name: num${group}
Module: Net
Simulation Models and INET

What is Simulation Model?
As we know that
OMNET++ is not a simulation itself
 It is a framework that allows other simulation frameworks
– To be created
– To be simulated
 Simulation frameworks are simulation libraries
– Implement protocols
–
Types of Simulation Model
 Domain-specific functionality is provided by model frameworks
– WSNs
– Ad-hoc networks
– Internet protocols,
– Performance modeling
– Photonic networks, etc.,
 Developed as independent projects
 Reusability of models in OMNeT++ is due to its modular architecture
 Simulation models are easily integrated into OMNET++
Some Well-known Types

 INET Framework
 OverSim
 Veins
 INETMANET
 MIXIM
 Castalia
INET
 The INET Framework can be considered the standard protocol model library of OMNeT++
 Contains models for the Internet stack
― TCP, UDP, IPv4, IPv6, OSPF, BGP, etc
 Wired and wireless link layers
 Ethernet, PPP, 802.11, etc)
 Support for mobility
 QoS support
 DiffServ, RSVP
 Several application models
 Maintained by OMNeT++ team officially
OverSim
 Overlay and peer-to-peer network simulation framework
 Contains several models for
 Structured
― Chord
― Kademlia
― Pastry
 Unstructured
― GIA
Veins
 Inter-Vehicular Communication (IVC) simulation framework
 It is a road traffic microsimulation model
INETMANET
 Fork of INET framework
 Simulation frAAamework for mobile ad-hoc networks
 Written and maintained by Alfonso Ariza.
MIXIM
 Modeling framework created for
― Mobile wireless
― Fixed wireless
― WSNs
― BANs and VANs
― Ad-hoc networks
 Radiowave propagation
 Interference estimation
 Power consumption
 Wireless MAC protocols
CASTALIA
 Simulation framework for networks of low-power embedded devices
 Offers models for
― Temporal path loss
― Fine-grain interference
― RSSI calculation
― Physical process model
― Node clock drift
― MAC protocols
Design Tour of INET 1
In this module
We shall take a guided
Tour of INET to
 Understand how ARP works in Ethernet environments
 Walk through features of INET
 Peek into various
– Packets
– Queues
– Internal tables
Why ARP scenario?

 While ARP is not the most important protocol, it is very interesting
 It relates to
– Ethernet
– IP
– And other higher layer protocols
Scenario
 Client computer opens TCP session with server
 Rest of operations (including ARP) follow
– ARP has to learn the MAC address for the default router
Usage Diagram for ARP
On simulation start
Ethernet autoconfiguration precedes ARP
Entities at work
 Various compound modules interact with each other
 TCP host on Ethernet
 Router
 TCP server
 How end-to-end transmission takes place?
TCP Client
Router
TCP Server
End-to-end transmission
Ethernet Compound Module

 In order to further understand how INET works, let us explore Ethernet (Compound
Module)
 Consists of
― Arp
― Encap
― And Mac
Ethernet Compound Module

arpTest.client.eth[0].arp
Inside ARP PaAcket
ARP Packet Class (Generated by .msg file)

Packet Queue (Contains IP Packet)
ARP Cache Build-up
Introduction to top-down approach to modelling and simulation
Top-down approach to NeMS

 Networks are complex to design
 One-time design of simulation is cumbersome
 Top-down: Phased roll-out of model-simulate cycle
― Iterative
Rolling-out of model at every layer to Design a Network
Design goodness (QoE) is user-centric aspects
Strategy
Rules for Mathematical Reading
What is mathematical modeling?
A Representation of an object, a system, or an idea in some form other than that of the entity itself.
(Shannon)
Quantification
 The act of counting and measuring that maps human sense, observations and experiences
into members of some set of numbers
 Facts represented as quantitative facts are the basis of science
Formalism
 Mathematics creates models that have certain relationships
 Statements of mathematics can be considered to be statements about the consequences of
certain string manipulation rules
Best practices to read mathematical expressions

A) Understanding math is like understanding a foreign language
B) Learn the formulas you already understand
C) Always learn what the formula will give you and the conditions
D) Keep a chart of the formulas you need to know
E) Math is often written in different ways, but with the same meaning
What is an equation?
 A statement that the values of two mathematical expressions are equal
 indicated by “=” sign
 What is a formula then!
Constituents of an equation?
 Expressions consist of one or more of these arguments
– Numerical constants
– Symbolic names
– Mathematical operators
– Functions
– Conditional expressions
Easy math writing

 2-3-4 rule
 Consider splitting every
– Sentence of more than 2 lines
– Sentence with more than 3 verbs
– Paragraph with more than 4 "long" sentences
 Use mnemonics
– s for speed
– v for velocity
– t for time
Easy math writing
 Organize into segments
– An entity intended to be read comfortably from beginning to end!
 Segments are standalone
– Definite start
– Definite end
 Segments should be represented linearly
QoE—Usability
Everything starts with “You”
What is usability?
 Usability (Ub) is defined as the ease of use with which network users can access the network
and services
 Ergonomic and technological facilitation
– Networks should make users’ jobs easier
 Some design decisions have a negative affect on usability
– Strict security
 Some choices are user friendly
– WiFi
– DHCP
Understanding usability
Sanjay Kumar Gupta, “Usability Models Based on Network Artifacts for Rural Development” Int. J.
Computer Technology & Applications,Vol 4 (3),508-513
 Ub: Usability as ease of use
 Ue: Use effort
 Ub 1/Ue
Usability expressions
Connotations
 Usability (Ub) is expressed as a function of network devices
 The top-down approach implies that the assessment of overall usability has to be based on
the performance of
– Hubs/switches
– Routers/gateways
QoE—Scalability
Ability to grow
What is scalability?
 Scalability refers to the ability to grow (or add)
 Factors to be added
― Number of applications
― Number of sites
― Addressing at sites
― No. of users
― No. of servers
Effects of growth
 Efficiency decreases with increasing factors
– But increases with increasing “other” factors
 Execution time increases with increasing factors
– But decreases with increasing “other” factors
Understanding efficiency & speed-up

 Execution time tends to vary with problem size
― Must be normalized when comparing network performance at different traffic
volumes
ERelative = T1 ¸ (No. of hosts ´ TNo of hosts)
SRelative = No. of hosts ´ E1
Understanding execution time

TimeExecution = TCompute + TComm + TIdle
Tmsg = ts + twL
Connotations
 Scalability is expressed as a function of factors in the network
 This criteria affects the design choices made for the network model
QoE—Planning for Expansion
Need to expand is ever increasing
Why plan?
 Expansion is unavoidable
 Unplanned expansion causes performance degradation
– Execution time
– Efficiency
 Planning is necessary
– Preemption is key
– Late planning is no planning
Considerations for Planning

 Nodes and locations
– End hosts
– Switches
– Routers
 Equipment scalability
– No of ports
 Naming system
– Extensible tuple
(Node ID, Network ID)
 Application-specific protocol choices
– Email
– File transfer, sharing & access
– DB access & updating
– Web browsing
– Network game
– Remote terminal
– Videoconferencing
– Video on demand (VoD)
QoE—Expanding Access to Data

Scalability without continued access to data is futile
Access to data
 Social networking has emerged
 Extranets need topology definitions & dedicated bandwidth allocation
– Classic 80-20 rule ´
 Increased access
– Data available to more departments
– Increased utilization of network services
Metcalfe's Law
 Community value of a network grows as the square of the number of its users
 Often cited as an explanation for the rapid growth of the Internet
Expression for Metcalfe's Law
n(n − 1)/2 or O(n2) connections between “n” nodes
Manifestation of Metcalfe's Law
 Can be seen in network applications
Connotations
 Network model is more scalable than the number of nodes and servers in the topology
 The total traffic load generated depends upon the user activity
QoE—Constraints on Scalability
The whole cannot be greater than the sum of its parts
(Apologies to Aristotle)
Recall parts of model!

 Nodes
― Computing
― Memory
 Protocols
― Operation
― Message formats
 Devices
― Ports
― Specifications
 And more!
Identify their upper bounds

 Nodes
― Nmax
 Protocols
― Operation: Omax
― Message formats: Mmax
 Devices
― Ports: Pmax
― Specifications: Smax
Maximum scaled up network

 Given by MinMax decision rule
Min(Nmax,Omax,Mmax,Smax)
 The strength of the chain is determined by the weakest link
Specific example
 Constrained addressing
― IPv4
― Top-level exhaustion occurred on 31 Jan 2011
― 24 Sep 2015 for North America
 Unconstrained addressing (for now!)
― IPv6
 With everything as IoT, 2128 is the constraint
QoE—Availability
The degree to which a system, subsystem or equipment is in a specified operable and committable
state
Everything may fail, not if but when!

 Networks Nodes Links
 Typical failure of components represented by the famous
Network Availability
 Percent uptime per year, month, week, day, or hour to total time in that period
– For example:
• 24/7 operation
 Network is up for 165 hours in the 168-hour week
– Availability is 98.21%
Application perspective
 Applications may require different levels
– Real time
• Video/Audio
– Commerce
• Non-repudiable transactions
– Non-real time
• Email
Availability vs reliability
 Reliability is the ability of a system to complete its function
– accuracy
– error rates
– Stability
 Even if a system is available does not mean its reliable
Availability vs capacity
 A system that runs out of capacity becomes unavailable
– ATM connection admission control
– Regulates no. of cells into network
 If capacity & QoS for connection unavailable
– cells dropped
Availability vs redundancy
 Redundancy is not a goal
 It is provided to achieve a level of availability
– Only a means!
Availability vs resiliency
 How much stress can be taken by network?
– Availability difficult to maintain
– No. of failures that make a system unavailable
 How soon can a network rebound?
– Availability difficult to achieve
QoE—Disaster Recovery
Amat Victoria Curam (Latin)
Victory Loves Preparation
Benjamin B. M. Shao, “Allocating Redundancy to Critical Information Technology Functions for

Disaster Recovery,” Proc. 10Th Americas Conference on Information Systems, Aug. 2004
The question
How to allocate redundancy to IT functions such that the overall survivability of these IT functions
against disasters is maximized and the cost remains under budget.
Redundancy
 Redundancy in preparation for disasters provides disaster preparation
– Proactive prevention
– Reactive recovery
– Backup facilities
Redundancy Allocation Scenario
Redundancy Allocation Model

 IT function can be implemented by a number of IT assets
– Computing hardware
– Communication links
– IT personnel, and
 other infrastructure
 m fails against d only when all of its selected solutions fail at same time
– As long as one of the selected solutions survives, m would still be operational
QoE—Specifying Requirements
Measurable is achievable
Availability in %age per annum

 Uptime of 99.70%
– 30 mins downtime
 Uptime of 99.95%
– 5 mins downtime
 Map onto totally deviant requirements
Availability in calendar year

 Downtime on weekdays
– vs weekends
 Project deadlines
Availability in spurts
 Staggered vs onetime
 99.70% uptime
― 30 minutes per year
― 10.70 sec per hour
 Acceptable for some users not to others
 Allowed for few applications
QoE—Five Nines Availability
The devil is in the details of availability
5 9s as best-case availability
 Some enterprises may want 99.999%
– 5 minutes downtime per year
 Sometime or all the time?
– A million $ worth question for managers
 Repair time inclusive or exclusive
– In service upgrades (hot-swaps) possible?
 Hardware manufacturers provide 5 9s
 However sum is not equal to parts
– Carrier and power outages
– faulty software in routers & switches
 Unexpected and sudden increase in bandwidth or server usage
 Configuration problems, human errors (90% of all!)
 Security breaches, and software glitches
Shifting Impact of 9s on time
99.999% Availability might require triple redundancy
One being active, one in hot standby ready to be used immediately, one in standby or maintenance
QoE—Cost of downtime
40 percent of companies that
shut down for three days failed within 36 months (Contingency Planning and Management
magazine)
Source: Top Business Continuity Priorities for 2004.©EnvoyWorldWide - February, 2004
Source: New England Disaster Recovery Information X-Change (NEDRIX)

Step-wise approach to measure downtime cost
1. Identify Business Continuity Components
2. Define What You Protect
3. Prioritize Business Functions
4. Classify Outage Types,
5. Calculate cost
Identify Business Continuity Components

 People
 Property
 Systems
 Data
Define What You're Protecting

 Define core competencies
― product, service, process, or methodology
Prioritize Business Functions
 Business functions necessary to sustain that core competency
― And associated IT infrastructure
 80% of available resources restore 20 % systems, applications, and data
Outage Types, Frequencies, & Duration

 Branch Outage
 Regional outage
 Data center outage
 National outage
Calculate cost
Frequency x Duration x Hourly Cost = Lost Profits
Example
 If there were 90 branch outages in an average year
― Each lasting an average of one-and-a-half hours
― Costing $300/hour 90 outages x 1.5 hours x $300/hour = $ 40,500
 Cost of branch outages for a year =$40,500
QoE—MTBF AND MTTR
Averaging out the availability
Availability as MTBF
 Mean time bw failure (MTBF) & mean time to repair (MTTR)
 Component vs service
– Mean time bw service outage (MTBSO)
– Mean time to recover from service outage (MTTSO)
 Typical MTTF value is once per 4000 hrs or 166.7 days
 Typical acceptable MTTR value is one hour
Availability = MTBF/(MTBF + MTTR)
4,000/4,001 = 99.98% availability
 MTBF with MTTR help to assess frequency and length of service outage
– Mean value must be supported with variance
 The difference between MTTF and MTBF is the assumption of the former that the system
shall be repaired while in the later the system is replace
QoE—Network Performance
Composite metric that is end-to-end
Definition
 An overall working
 Many different ways to measure the performance of a network
― Each network is different in nature and design
 Modeled
 Simulated
 Measured
QoE—Optimum Network Utilization
Optimum is “As good as it gets”
Definition of optimum
 Selection of a best element (with regard to some criteria) from some set of available
alternatives
Optimum network utilization

 How much % of bandwidth capacity in a specific time period?
 Time varying phenomenon
– Instantaneous, averaged, weighted)
 Both goal & constraint
 Typical value is 70$%
– Exceeding this results in performance degradation
 WAN links utilization is more crucial than LAN
– Pay per packet
 Compression, caching and concatenation used to reduce WAN utilization
 LANs are over-budgeted
– Fast Ethernet)
 Full-duplex vs. half duplex switches
 User activity levels
 LANs suffer from exceeding utilization in switch-to-switch
QoE—Throughput
Throughput = Goodput + Badput
Definition of throughput
 Quantity of error free data transmitted/ sec
– Erroneous transmissions futile
 Ideally, should be the same as capacity
Deviation indicates the limitations of media type, device and network
QoE—Throughput of devices
Simulation of devices and specifications is vendor specific
Types of device throughputs

 Inter-networking devices give throughput as in
– TCP/IP: Packets per second
– ATM: Cells per second
 Sizes vary from 53, 64 to 1518 Bytes
Example—CISCO devices
 Traffic generators-device-traffic checkers in tandem measure throughput
– Smaller packets give better pps
 Cisco claims of 400 million pps for the Cisco Catalyst 6500 switch
CISCO claims throughput; which in actual is the capacity
QoE—Application Layer Throughput
Application layer uses lower layers unfairly
Definition
 Application layer throughput = goodput + badput
 Goodput vs badput
 Badput contributed by retxns, header etc
– Fraction of packets that collided/lost
 Fc = C/N
 Fc = L/N
Factors affecting goodput

 End-to-end error rates
 Protocol functions (handshaking, windows, & acks)
 Protocol parameters (frame size, retx timers)
 pps rate of networking devices
 Lost packets at networking devices
 Workstation & server performance factors:
– Disk-access speed
– Disk-caching size
– Device driver performance
 Computer bus performance (capacity/arbitration)
 Processor (CPU) performance
 Memory performance (access time for real and virtual memory)
 Operating system inefficiencies
 Application inefficiencies or bugs
An, Cheolhong, and Truong Q. Nguyen. "Error Resilient Video Coding using Cross-Layer
Optimization Approach." IEEE Transactions on Multimedia 10 (2008): 1406-1418.
Connotations
 Application layer throughput provides insight into “useful' transmissions
– It relates resource allocation down to physical layer throughput
QoE—Accuracy
Being accurate is not being precise
Definition
 Data sent and received should be the same
 Also referred as the number of error-free frames transmitted relative to the total number
of frames transmitted
Factors affecting accuracy

 Packet reordering at routers
 Power surges
– Lightning impulse of 1 s on 10 Mbps link
 Impedance mismatch problems
 Poor physical connections
 Failing devices
 Noise caused by electrical machinery
 WAN links give BER and SNR (10-5~10-11)
 LANs specify erroneous frames per 106 Bytes
 On shared Ethernet, collisions main cause of accuracy degradation
 First 64 Bytes collision (legal or runt frames)
 Typical acceptable value is .1% frames
 Late collisions are illegal
• Nahum, Erich M. "Validating an architectural simulator." Department of Computer Science,

University of Massachusetts at Amherst. 1996.
Accuracy = [(Real value – Error) / Real value] * 100

The frequency of events plays a key role in the overall accuracy
– Ei is the event i in the system
– freq(Ei) is the frequency of event i
– real(Ei) is the desirable (real) cost of event i
– sim(Ei) is the simulated (obtained) cost of event i
QoE—Efficiency
Boiling water analogy
Definition
 Application layer throughput = goodput + badput
 Goodput vs badput
 Badput contributed by retxns, header etc
– Fraction of packets that collided/lost
• Fc = C/N
• Fc = L/N
Factors affecting efficiency

 Access protocols
– high number of users showing activity
– Ethernet not efficient at high collision rates
 Frame size
– Using large frame is useful for single user on WAN links
 Serialization delay on WAN links results in unfair treatment
– for real-time shorter frames enquired in router
Kleinrock, Leonard. "Creating a mathematical theory of computer networks." Operations Research

50.1 (2002): 125-131.
If you scale capacity more slowly than throughput while holding the average response time
constant, then the channel efficiency (channel utilization) will increase
Average Efficiency
Latora, Vito, and Massimo Marchiori. "Efficient behavior of small-world networks." Physical
review letters 87.19 (2001): 198701.
 E(G) is the average efficiency of a network G

 n denotes the total nodes in a network
 d(i,j) denotes the shortest path between a node i and a neighboring node j
QoE—Delay and Jitter

Applications might forgive delay but not jitter
Delay
 Voice and video applications (especially interactive) demand minimum delay
 Other applications such as Telnet remote echo need timed performance
Sources of packet delay
Delay variation (jitter)

 The amount of time average delay varies
 Voice, video, and audio are intolerant of delay variation
Source of jitter
QoE
It is the small factors that matter the most
Causes of Delay
 Propagation
– Media type
– Length
 Transmission (serialization)
– 1024 Bytes on T1
 Switching delay
– upto 5-20 microsec for 64 Bytes frame
 Router delay
– Look-up, router architecture, configuration
– Software features that optimize the forwarding of packets
 NAT, IPSEC, QoS, ACL Causes of Delay (3 of 3)
 Queuing delay
– Dependent upon utilization
 Formula
Queue depth = Utilization/(1 – Utilization)
Queue Depth vs. Utilization
Implications of queuing delay

QoE—Delay variation
All animals are equal, but some animals are more equal than others (George Orwell)
Delay variation
 Amount of time average delay varies
 Voice, video, and audio are intolerant of delay variation
 Tradeoffs needed for efficiency for high-volume applications versus lowConcept of jitter
buffer to smoothen out the jitter
 Variations on the input side are smaller than the buffer
 Acceptable variation is 1-2% of the delay
Jitter types
 Jitter is quantified in two ways
 Delay jitter
― bounds maximum difference in total delay of different packets
― Assumes source is perfectly periodic
 Used for Interactive communication
– voice and video teleconferencing
 Helps to translate to maximum buffer size needed at the destination Second measure is rate
jitter
 Bounds difference in packet delivery rates at various times
 Measures difference between minimal and maximal inter-arrival times (reciprocal of rate)
 Useful measure for many real time applications
 Video broadcast over the net
 Slight deviation of rate translates to only a small deterioration in the perceived quality
Jitter Analysis Points
Kay, Rony. "Pragmatic network latency engineering fundamental facts and analysis." cPacket
Networks, White Paper (2009): 1-31.
Measurement of jitter
QoE—Response Time
Response time is relative phenomenon
Definition
 The amount of time between a request for some network service and a response to the
request
Measurement Points Locations

Tim R Norton. "End-To-End Response Time: Where to Measure?" Computer Measurement Group
Conference Proceedings, 1999.
Measurement of Response Time
[1] Reinder J., Bril., System Architecture and Networking. TU/e Informatica
[2] Sjodin, Mikael, and Hans Hansson. "Improved response-time analysis calculations." Real-Time
Systems Symposium, 1998. Proceedings., The 19th IEEE. IEEE, 1998.

Ceiling function represents maximum number of pre-emptions by higher priority processes
QoE—Security
Threat = Capability + Intention
Definition
 Protection of information systems from threat
– Hardware
– Software
– Information on them
 Avoidance from
– Disruption
– Misdirection of the services they provide
Implementation
 Includes controlling physical access to the hardware
 Protecting against harm via
– Network access
– Data
– Code injection
Trusted Computing Base

 Rainbow Series(orange book)
 Set of all hardware, firmware, and/or software components
 Critical to its security
 Bugs occurring inside jeopardize security of entire system
Bell-Lapadula Model
 Users as Subjects
 Predicates
– Devices and data as Objects
 Process algebra provides the action (verb) of subject over predicates
Bell-Lapadula Model
 Users as Subjects
 Predicates
– Devices and data as Objects
 Process algebra provides the action (verb) of subject over predicates
QoE—Reconnaissance Attacks
Prevention is better than cure
Definition
 Reconnaissance is a type of computer attack
 Intruder engages with the targeted system
– Gathers information about vulnerabilities
Types
 Active reconnaissance
 Port scanning
 Passive reconnaissance
 Sniffing
 War driving
 War dialing
Targeted Threat Index
Hardy, Seth, et al. "Targeted threat index: Characterizing and quantifying politically-motivated
targeted malware." Proceedings of the 23rd USENIX Security Symposium. 2014.
Targeted Threat Index

 Vulnerability of system
 Depends upon
– Target feature set
– Attacker methods
– Attacker aggressiveness
TTI = Method * Implementation
QoE—Security Requirements
Definition
 Enlist all the activities, actions, hardware/software
 Confidentiality
 Integrity
 Authorization
 Authenticity
 Availability
 Encryption
Assessing Security Levels

Burchett, Ian. "Quantifying Computer Network Security." (2011).
Common Vulnerability Scoring System
 Provides a repeatable quantitative score for computer security vulnerabilities
Vulnerability Compositing Method per Client
QoE—Manageability
Definition
 The level of human effort required to keep that system operating at a satisfactory level
– Deployment
– Configuration
– Upgrading
– Tuning
– Backup
– Failure recovery
Assessing Manageability
Candea, George. "Toward Quantifying System Manageability." UseNix HotDep. 2008.
Manageability Metric
The notion of efficiency of management operations, which is approximated by the time Timei the
system takes to complete Taski
Approximate complexity of a management task by the number of discrete, atomic steps (Stepsi)
required to complete Taski
Commentary
 Manageability is reduced proportionally to how long the management tasks take

 And to how many atomic steps are involved in each such task
 The fewer steps there are, the lower the exposed complexity of the system
 The faster the management tasks can be completed, the lower the likelihood of trouble
 Less management a system requires (i.e., the longer TotalTimeeval for the same Ntotal), the
easier it is to manage
 Equivalently, the less the system needs to be managed, the better
QoE—DoS Attack
Definition
 An attempt to make a machine or network resource unavailable to its intended users,
 Temporarily
 Indefinitely
Implementation
 Transmit a large number of packets
– TCP Syn attack
– Ping attack
 Server crashing attack
– Large computational load
A Simple Attack Analysis

He, Changhua. Analysis of security protocols for wireless networks. PhD Diss. Stanford University,
2005.

 Attack type: TCP SYN flooding DoS attacks
– n packets are used for attack
 Counter: Random drop queue 'Q'

– Q = queue depth
Attack success probability

 P = 1 − (1 − 1/Q)n
Attack failure probability
 1-P

Making Network Design Tradeoffs
Definition
 Make balance between desirable & incompatible features
 A compromise
 Often conflicting technical goals
 Make tradeoff a necessity
– Availability vs affordability
– Usability vs security
A Simple Communication tradeoff

Compressing of an image
 Reduces transmission time/costs
 At the expense of CPU time
 Tradeoff between computation and communication
Tradeoff at Network Level
 Throughput is at conflict with fairness

 Tradeoff can be implemented through weighted scheduling
A child with Rs. 100 in a convenience store!

Handle it as a knapsack problem!
A child with Rs. 100 in a convenience store!
Problem Set 1
Effect of Topology Factors

1. What is the total data rate of the network?
2. What is the application that is generating the maximum load per user in Administration
department?
3. What is the application that is generating the minimum load per user in Math and Science
department?
Effect of Routing Protocols

1. If RIP sends a routing packet every 30 seconds and each packet contains 25 routes (Each route is
20B), what is the bit rate?
Problem Set 2
Effects of Deployment/Protocol Behaviors

1. Where is the data center?
2. What is the data rate available for users of Eugene?
3. What is the maximum Internet speed available to the users?
4. Label the router that needs to implement firewall.
5. If a user in Medford sends out a broadcast 255.255.255.255, what is the impact?
Queuing Behaviors
1. A CISCO switch has 20 users (clients and servers), each offering packets at a rate of 200 packets
per second. If the average length of the packets is 64 Bytes, and the transmission rate of the switch
is 10 Mbps measure the load of all the users and the LAN utilization. Then measure the queue
depth
Understanding Network Design

1. Label the bastion host in the network.
2. Label the fastest end-to-end interoffice segment.
3. Label the slowest end-to-end interoffice segment.
4. How many total LAN segments are there?
5. Label at least one network where duplex auto-negotiation might help.
6. Label at least one segment where BERT can be used to measure BER.
Simulate FTP Scenario
A Real World Scenario
Factors affecting goodput

 End-to-end error rates
 Protocol functions (handshaking, windows, & acks)
 Protocol parameters (frame size, retx timers)
 pps rate of networking devices
 Lost packets at networking devices Workstation & server performance factors:
 Disk-access speed
 Disk-caching size
 Device driver performance
 Computer bus performance (capacity/arbitration)
 Processor (CPU) performance
 Memory performance (access time for real and virtual memory)
 Operating system inefficiencies
 Application inefficiencies or bugs
Implementation in INET
Source: https://omnetpp.org/doc/inet/api-current/neddoc/index.html
examples/inet/bulktransfer/BulkTransfer.ned
Usage diagram
Source: src/applications/tcpapp/TCPBasicClientApp.ned
numRequestsPerSession = exponential(3)
requestLength = truncnormal(20,5)
replyLength = exponential(1000000)
What to model?
1. Total time it takes to complete file transfer
2. Total goodput vs badput
3. Network utilization
4. Delay variation
5. Usability
6. Scalability
7. Availability
Parameters
What to model?
Simulating DoS Attack
Igor Kotenko & Alexander Ulanov , “Simulation of Internet DDoS Attacks and Defense ,” ISC
2006, LNCS 4176, pp. 327–342, 2006.
Kaur, Rupinderjit, Amrit Lal Sangal, and Kush Kumar. "Modeling and simulation of DDoS attack
using Omnet++." Signal Processing and Integrated Networks (SPIN), 2014 International
Conference on. IEEE, 2014.
What to model?
Configuring Ping of Death attack

cSimpleModule::initialize();
packetSize = par("packetSize");
sendIntervalPar = &par("sendInterval");
hopLimit = par("hopLimit");
count = par("count");
startTime = par("startTime");
stopTime = par("stopTime");
Summarizing top-down approach

Our Strategy
Application layer Roll-out for M&S
Simulate RTP with Packet Loss

Family of RTP
RTP
• Real-time Transport Protocol (RTP) is a network protocol
• Delivers audio/video over IP networks
• Streaming media
• Telephony
• Video teleconference
• Television service
• Push-to-talk over web
Delay/Jitter Analysis Points
Inet for Simulating RTP

(examples/rtp/unicast1/unicast1.ned)
src/nodes/rtp/RTPHost.ned
Usage Diagram and Statistics
Client Server Architectures

An architecture for data exchange
Definition
 One known server
 Always-on
 Permanent IP address
 Clients communicate with server
 Intermittently connected
Performance
 Distribution time for the client-server architecture denoted by Dcs

 Size of the file to be distributed (in bits) by F
 Number of peers that want to obtain a copy of the file is N
 dmin denotes the download rate of the peer with the lowest download rate
 Server upload rate is us
Web Server Modeling
Message Flow
Operation
 Handles multiple HTTP requests
 Accepts and parses the HTTP request
 Gets the requested file from the server's file system
 Creates and sends an HTTP response message consisting of the requested file
Characterizing web server

 Buffer size per client
 Number of clients
 File size that it handles
 Processing time
 Time out interval
HTTP Modeling
Time line operation
Variants
 HTTP is based on sequenced messages
 Underlying TCP handshaking determines the overall performance
– Persistent
– Non-persistent
– Pipelined
– Caching
Non-Persistent Connections
TCP handshaking required for every object
Modeling Non-persistence
 It requires 2 RTTs per object
 Total time for N objects
N*2RTT + N*Transmit time
 Consequent effect on simulated time is exacerbated in a multi-hop real world network
Persistent Connections
TCP handshaking required once
Modeling Persistence
 It requires 1 RTTs per object
 Total time for N objects
(N+1)*RTT + N*Transmit time
 Consequent effect on simulated time is noticed in a multi-hop real world network
Cache Response Time

Caching operation
 User sets browser: Web accesses via cache
 Browser sends all HTTP requests to cache
― Object in cache: cache returns object
― Else cache requests and returns object from origin server
Clients requesting objects through cache
Advantages of caching
 Reduces response time for client request
 Reduce traffic on an institution’s access link
Simulating Scenarios with and without cache

Factors affecting caching
 Average request rate from institution’s browsers to servers
 Round trip delay from institutional router to server
 Correlation between requests
 Average object size
Example
 Average object size = 100,000 bits
 Avg. request rate from institution’s browsers to origin servers = 15/sec
 Delay from institutional router to any origin server and back to router = 2 sec Utilization on
LAN = 15%
 Utilization on access link = 100%
 Total delay = Internet delay + access delay + LAN delay
= 2 sec + minutes + milliseconds
 If hit rate is .4
 40% satisfied locally
 60% requests satisfied by server
 Utilization of access link reduced to 60% (say 10 ms)
 Avg delay = Internet + access + LAN
= .6 * (2.01) s + ms < 1.4 secs
FTP Efficiency
FTP operation
 Client contacts FTP server at port 21
 Client obtains authorization
 Browses remote directory
 Server receives file transfer command
 Server opens TCP data connection to client
 After transfer connection closed
Control Signaling of FTP

Computational Efficiency of FTP
(COURTESY: ALEBRA TECHNOLOGIES INC.)
TCPU = Total CPU seconds recorded during the period of file transfer
ICPU = Measured CPU seconds when machine is idle for the equivalent period
MIPS = Machine performance rating in Millions of Instructions per second
TRATE = Transfer rate in megabytes per second
SMTP Scalability
Entities of SMTP Architecture
Recall scalability
 Ability to grow
 Scaling may include
― Number of user sites
― Inter-site topology
― No. of user agents
― User mailbox size
― No. of mail servers
― Outgoing queue size
Efficiency & speed-up for SMTP
Mail delivery time tends to vary with scaling factors
• Must be normalized when comparing SMTP performance at different traffic volumes
― On single server
― Servers confederation
ERelative = T1 ¸ (No. of hosts ´ TNo of hosts)
SRelative = No. of hosts ´ E1
DNS Load Distribution & Loss
Typifying DNS operation
Casalicchio, E., Caselli, M., Coletta, A., & Fovino, I. N. Aggregation of DNS health indicators:
issues, expectations and results
Health metrics
Incoming Bandwidth Consumption (IBC)
 Ratio between total amount of incoming data during a session over the duration of the
session
 Range: [0, IBC max]
 measured in Mbit/s
Health metrics
Incoming Traffic Variation (ITV)
 For each session i,
(IBCi − IBCi−1)/lengthi
 IBCi is incoming bandwidth consumption in ith session
 lengthi is duration of that session
Traffic Tolerance (TT)

 Measures the Round Trip Time (RTT) of a IP packet flowing between end-user node and
ISP’s recursive resolver in seconds
DNS Requests per Seconds (DNSR)

 It gives the total number of DNS queries in the session
Rate of Repeated Queries (RRQ)

 In a single session a name is resolved only once due to caching
 The metric returns no. of repeated DNS queries in a session for same name if the query is
lost
– Or not cached
Peer to Peer Scalability

Operation
 No always-on server
 Arbitrary end systems directly communicate
 peers are intermittently connected
 Change IP addresses
File Distribution Problem
Performance
 Distribution time for the P2P architecture denoted by DP2P

 Size of the file to be distributed (in bits) by F
 Number of peers that want to obtain a copy of the file is N
 dmin denotes the download rate of the peer with the lowest download rate
 Upload capacity of the system as a whole = the upload rate of the server plus the upload
rates of each of the individual peers, that is, utotal = us + u1 + ... + uN
 Server upload rate is us
Torrents Efficiency
Basic Torrent Operation
Factors affecting efficiency
 Heterogeneous upload capacity
 Diversities of neighbor selecting mechanisms
 Geographical distribution of peers
 Downloading rates of LocalBT clients
 Peer selection policy
Performance
Reliability of Circular DHT

Operation of Circular DHT
REDUNDANCY HANDLES FAILURES

Cost of Reliability
l leaf-set keepalive messages every T seconds

2-messages for probe and response Routing table probes every Trt
Summation computes expected number of routing table entries (128/b rows and 2b columns)
 Last expression is a binomial distribution
Problem Set 1
Network Latencies
Consider an institutional network connected to the Internet. Suppose that the average object size is
850,000 bits and that the average request rate from the institution’s browsers to the origin servers is
16 requests per second. Also suppose that the amount of time it takes from when the router on the
Internet side of the access link forwards an HTTP request until it receives the response is three
seconds on average.
Model the total average response time as the sum of the average access delay (that is, the delay
from Internet router to institution router) and the average Internet delay. For the average access
delay, use Δ/(1 – Δb), where Δ is the average time required to send an object over the access link
and b is the arrival rate of objects to the access link.
Now suppose a cache is installed in the institutional LAN. Suppose the miss rate is 0.4. Find the
total response time.
HTTP Performance
Suppose that an HTML file on a web server references eight (8) very small objects. Neglecting
transmission times, how much time it takes when non-persistent HTTP connection is used and the
browser is configured for five (5) parallel connections?
A. 18RTT B. 6RTT
C. 3RTT D. None of these
Problem Set 2
P2P Protocols
Suppose that peer 3 learns that peer 5 has left. How does peer 3 update its successor state
information?
A. It asks peer 4 B. It asks peer 8
C. It asks peer 2 D. None
User Activity Monitoring

For a 1 Mbps link, if each user generating 200 kbps is active for 20% of the time, what is the
probability that out of a total of 100 users, more than 5 users be active?
Simulate HTTP Persistence
HTTP Evolution
 RFC 793 does not support persistence
– HTTP1.0
 Additional mechanism needed
– Use keep-alive
 HTTP 1.1 is persistent by default
HTTP Support in OMNET++

Module Interface ITCPApp
 Template for TCP applications (Inheritance)

 It shows what gates a TCP app needs
 to be able to be used in StandardHost etc
HTTP Browser in OMNET++

src/applications/httptools/HttpBrowser.ned
Default support is HTTP 1.1
simple HttpBrowser like ITCPApp
{ parameters:
int httpProtocol = default(11); }
Supported Modes
 Random request mode

 Browser uses statistical distributions generate requests to random web servers
 Scripted mode
 Browsing behavior determined by a list of predefined web sites to visit at specific times
Simulate DNS Query Response

Basic Operation
DNS Support in OMNET++

 Extensions provide classes and functions to simulate DNS and MDNS traffic
 Implement RFC 1035
Supported DNS Operations

 Name servers with recursive resolving capabilities
 Authoritative servers with DNS zone configuration using master files
 Caching servers without zones
 Only recursively resolving
 DNS Cache base that can be extended
 Caches based on different policies possible
 DNS client that can query a DNS server
Simulate TCP Threading

Threaded Server
INET Support for TCP
 RFC 793 - Transmission Control Protocol

 RFC 896 - Congestion Control in IP/TCP Internetworks
 RFC 1122 - Requirements for Internet Hosts -- Communication Layers
 RFC 1323 - TCP Extensions for High Performance
 RFC 2018 - TCP Selective Acknowledgment Options
 RFC 2581 - TCP Congestion Control
 RFC 2883 - An Extension to the Selective Acknowledgement (SACK) Option for TCP
Features
• RFC 793 TCP states and state transitions
• Connection setup and teardown as in RFC 793
• Segment processing
• Receive buffer to cache above-sequence data
• Data not yet forwarded
Simulate HTTP Handshaking
HTTP Requests
HTTP/1.0: HTTP/1.1:
 GET  GET, POST, HEAD
 POST  PUT
 HEAD  uploads file in entity body to path
 asks server to leave requested object out specified in URL field
of response  DELETE
 deletes file specified in the URL field
HTTP Response
200 OK
request succeeded, requested object later in this msg
301 Moved Permanently
requested object moved, new location specified later in this msg (Location:)
400 Bad Request
request msg not understood by server
404 Not Found
requested document not found on this server
505 HTTP Version Not Supported
HttpBrowser Class Reference

(Inheritance Diagram)
Intro & Transport Services

Introduction
 Transport layer is the big brother
 Manages end to end delivery of data
 Modeling of transport layer is pivotal to the overall performance
Transport Services
 Multiplexing and demultiplexing
 Reliable, in-order delivery (TCP)
 Congestion control
 Flow control
 Connection setup
 Unreliable, unordered delivery: UDP
 “best-effort” IP
 Services not available
 Delay guarantees
 Bandwidth guarantees
Modeling Approach
Multiplexing & Demultiplexing

Basics
Capability of Port #
Cost of Multiplexing
Multiplexing Communication Link

Typical Multiplexing Techniques
Capacity Overhead
Capacity overhead ρ is defined as %age increase in the resource requirement of a practical

multiplexing scheme when compared to the optimal
Amount of resources allocated to application i at time t using the optimal and practical allocation
scheme respectively
Gain in Statistical (Packet) Multiplexing
 Each user: 100 kb/s when “active”

 active 10% of time
 Strict Multiplexing: 10 users
 Statistical Multiplexing: with 35 users, probability > 10 active at same time is less than
.0004
Checksum
Introduction
 Transport layer incorporates error detection
 Checksum is “checking the sum” both at the sender and receiver
 Performed at the header or the entire body
Operation in Brief & Performance
Overhead and Operational cost

 Divide the M-bit data into N-bit chunks
– Total chunks M/N
 Checksum is also N-bit
 Total sums M/N + 1
Undetected Errors
 Reordering of 2 byte words, i.e. 01 02 03 04 changes to 03 04 01 02
 Inserting zero-valued bytes i.e. 01 02 03 04 changes to 01 02 00 00 03 04
 Deleting zero-valued bytes i.e. 01 02 00 00 03 04 changes to 01 02 03 04
 Replacing a string of sixteen 0's with 1's or 1' with 0's
 Multiple errors which sum to zero, i.e. 01 02 03 04 changes to 01 03 03 03
Go Back N
Introduction
 Retransmission strategy (ARQ)
 No need to buffer at receiver
 Wheat and rice analogy!
– Go back N is wheat
– Fresher is better
Performance Amidst Packet Loss
Efficiency without Errors
 Choose N large enough to allow continuous transmission while waiting for an ACK for the
first packet of the window
If N > S/DTP E = min{1, N*DTP/S}
Selective Repeat
Introduction
 Retransmission strategy (ARQ)
 No need to retransmit all after loss
 Buffer requirements at receiver
 Wheat and rice analogy!
– Selective repeat is wheat
– Older is better
Performance Amidst Packet Loss
Efficiency without Errors
 Same as Go Back N
 If N > S/DTP E = min{1, N*DTP/S}
Efficiency with Errors

 Only packets containing errors will be retransmitted
E= 1 - P
Implications of buffer size
 Buffer limit at sender
― Number of un-ACKed packets at sender =< W
 Buffer limit at receiver
― Number of un-ACKed packets at sender cannot differ by more than W
RTT Estimation and Timeout
Fixed Window
First case
WS/R > RTT + S/R
 ACK for first segment in window returns before window’s worth of data sent
Delay Performance with Fixed Window
Fixed Window
Second case
WS/R < RTT + S/R
 Wait for ACK after sending window’s worth of data sent
 K is the number of windows that cover the object
Delay Performance with Fixed Window

TCP Timeout Value
 Longer than RTT
 As RTT varies
― Too short: premature timeout,
― Unnecessary retransmissions
― Too long: slow reaction to segment loss
Estimating RTT
 SampleRTT Measured time from segment transmission until ACK receipt
 Ignores retransmissions
 EstimatedRTT is “smoother”
 Averages several recent measurements, not just current SampleRTT
Relationship Between TimeOut and Estimated RTT
EstimatedRTT = (1- a)*EstimatedRTT + a*SampleRTT

TimeoutInterval = EstimatedRTT + 4*DevRTT
Reliable Data Transfer

 TCP offers data reliability
 In case data is lost or corrupted
― Error Control
― Flow control
― Congestion control
 Reliability at the cost of throughput
 With slow start and FRFR, throughput is given by
1.22  MSS
RTT L
Reliability Services
Handling Loss
Early TimeOut
Delayed Ack
Flow Control
Introduction
 Receiver throttles the sender by advertising a window
– Not larger than the amount of data that it can buffer
 TCP on the receive side must keep
LastByteRcvd − LastByteRead ≤ MaxRcvBuffer
Implication
 If local process reads data just as fast as it arrives
 Causes LastByteRead to be incremented at the same rate as LastByteRcvd
 Advertised window stays open
(AdvertisedWindow = MaxRcvBuffer)
 If receiving process falls behind, advertised window grows smaller with every segment that
arrives, until it eventually goes to 0
Advertised Window
Sender Window
LastByteSent − LastByteAcked ≤ AdvertisedWindow
Effective Window
EffectiveWindow = AdvertisedWindow − (LastByteSent − LastByteAcked)
Relationship between Max_Send and Max_Receive Buffer
LastByteWritten − LastByteAcked ≤ MaxSendBuffer
TCP Connection Management

Cost and Feasibility Model
Cohen, Edith, Haim Kaplan, and Jeffrey Oldham. "Managing TCP connections under persistent
HTTP." Computer Networks 31.11 (1999): 1709-1723.
Kurose and Ross. “Computer Networking Top-Down Approach Featuring the Internet”.
Holding Time
 Upon receiving an HTTP request r , the server decides on a holding-time interval T(r)
 The server then leaves the connection open for at most T(r) seconds from the moment it
received r
 If a new request r' arrives within the next T(r) seconds, then a new holding-time interval
T(r') is in effect
 Otherwise the connection is terminated after T(r) seconds
TCP State Transition
TCP Client Lifecycle

TCP Server Lifecycle
Connection Management Policy (1 of 3)

 Policy A is an algorithm that determines an interval T(r) for every request r
 Consider a request sequence s
 The profit (number of hits), PA of a policy A on s is the number of requests that did not
require opening a new connection
 The number of misses, MA of A on s is number of requests that require opening a new
connection
 The open-cost, HA of a policy A is total time connections are open
What to model?
 Trade-offs between open-cost and number of misses
Principles of Congestion Control

References
RFC 2914: Congestion Control Principles
Introduction
 Too many sources sending too much data too fast for network to handle
 Manifestations
 Lost packets
― buffer overflow at routers
 Long delays
― Queuing in router buffers
Infinite Buffer Scenario
Effects of Congestion
 Large delays when congested

 Maximum achievable throughput
Finite Buffer Scenario
Effects of Congestion
R/ R/ R/
R/
o o o R/
ut ut ut
R/ R/
in i i
a. No loss
b. Perfect loss
a. b. c
c. Imperfect loss
Combat Strategies (1 of 2)
End-end congestion control
 No explicit feedback from network
 Congestion inferred from end-system observed loss, delay
 Approach taken by TCP
 Network-assisted congestion control
 Routers provide feedback to end systems
 Single bit indicating congestion (SNA, DEC bit, TCP/IP ECN, ATM)
 Explicit rate sender should send at
ATM ABR Congestion Control
References
Introduction
 Available Bit Rate (ABR), a service used in ATM networks
 Source and destination don't need to be synchronized
 ABR does not guarantee against delay or data loss
 Allow network to allocate available bandwidth fairly over present ABR sources
Operation
 Elastic service
 If sender’s path is under loaded
 Use available bandwidth
 If sender’s path congested
 Sender throttled to minimum guaranteed rate
Combat Congestion
 Two-byte ER (explicit rate) field in RM cell
 Congested switch may lower ER value in cell
 Sender’s send rate thus minimum supportable rate on path
 EFCI bit in data cells is set to 1 in congested switch
 If data cell preceding RM cell has EFCI set, sender sets CI bit in returned RM cell
TCP Congestion Control

References
Introduction
 End-end control (no network assistance)
 Sender limits transmission
LastByteSent-LastByteAcked<= CongWin
 CongWin is dynamic, function of perceived network congestion
Operation
 Loss event = timeout or 3 duplicate acks
 TCP sender reduces rate (CongWin) after loss event
 Three mechanisms
― AIMD
― Slow start
― Conservative after timeout events
congestion
window
24 Kbytes
16 Kbytes
8 Kbytes
time
Leaky Bucket and Token Bucket
Leaky Bucket
 Buffering of the traffic to help manage and control the flow of traffic onto and through the
network
 “Leaky” means buffer that is constantly flowing
Operation
 Traffic enters into the buffers and is tagged, based on the amount of packets allowed by the
carrier
 If the user exceeds the amount of packets flow per increment then the buffer is filled and
begins to empty out the bottom side at a constant rate
Leaky Bucket Algorithm

Token Bucket
 Many traffic sources can be defined by token bucket scheme
 Provides concise description of load imposed by flow
 Easy to determine resource requirements
Operation
 Provides input parameters to policing function
 IP packet may be processed if sufficient octet tokens to match the IP data number of tokens
 If insufficient tokens available, the packet is relegated to best-effort service
 To transmit a packet through router, one token must be removed
 If token bucket is empty, packet is queued waiting for next token
 If there is backlog of packets & an empty bucket, packets emitted smoothly
Quality of Service
Background
 Broadband IP packet networks are multiservice, all-purpose communications platforms
 Spurred QoS efforts
 Simplest strategy to the one-size-fits-all best-effort service in today’s Internet: divide traffic
into classes
 Provide different levels of service to these different classes of traffic
Introduction
 QoS a non-issue for circuit-switched networks
 Layer 2 and 3 QoS approaches
 ATM and Frame Relay provide L2 QoS
 Provide circuit-like emulation
 Traffic agreements
 Traffic control
 Connection admission control
 Congestion notification
 Fragmentation
QoS at Network Layer
 IP QoS is concerned with end-to-end internetwork
 With every hop L3 QoS parameters mapping to L2 QoS
 Type of Service (TOS) field provides initial IP network class of service mechanism
 Three precedence bits classify eight categories of services
 Lower precedence dropped for higher precedence in congestion
 Network equipment vendors rarely provide precedence bits usage
QoS Models
 Two QoS models for IP packet networks
 IntServ
― Simulate “virtual circuit” of ATM or frame relay on L3
― Sets up an end-to-end route with fixed QoS parameters
 DiffServ
― Defining several common classes of service
― Each with associated queue priorities and drop precedence on a per-hop basis
Hard vs Soft QoS

 Hard guarantee applications will receive its requested quality of service (QoS) with certainty
 Soft guarantee application will receive its requested quality of service with high probability
Fair Queues
First In First Out
Motivation for FQ
 During periods of congestion, FIFO queuing benefits UDP flows over TCP flows
 A bursty flow can consume the entire buffer space of a FIFO queue
 PQ totally favours TCP over UDP
Introduction
 FQ is foundation for a scheduling disciplines designed to ensure that each flow has fair
access to network resources
 Prevents a bursty flow from consuming undue bandwidth share
 Also called per-flow or flow-based queuing
Operation
 Packets are first classified into flows by the system
 Assigned to a queue that is specifically dedicated to that flow
 Queues are then serviced one packet at a time in round-robin order
 Empty queues are skipped
Fair Queuing with Classifier
Benefits
 Primary benefit of FQ is extremely bursty or misbehaving flow does not degrade QoS
delivered to other flows
 Each flow is isolated into its own queue
 If a flow attempts to consume more than its share of BW, its queue is affected
Performance
 Allocation of single resource amongst N users
 Total resource µTotal
 Each user i requests ρi
 Each user i receives µi Conditions:
 No user receives more than its request
 No other user satisfying condition 1 has a higher minimum allocation
 Above condition remains recursively true as we remove the minimal user & reduce total
resource
 µTotal« µTotal –µi
 Conditions:
 µi= Min(µFair –ρi)
 Above condition remains recursively true as we remove the minimal user & reduce total
resource
μ Total = ∑ μ i
Priority Queues
Motivation
 Designed to provide a relatively simple method of supporting differentiated service classes
 To provide respective services to
– Interactive traffic
– Voice
– Video
– And best effort
Operation
 Packets classified and placed into different priority queues
 Packets scheduled from the head of a queue only if all queues of higher priority are empty
 Within each of the priority queues, packets are scheduled in FIFO order
Priority Queuing with Classifier
Trj : Resident time of an item j in queue k

Tsi : Service time of an item i i.e., processing time by the system
Variants
 Strict priority queuing
– packets in a high-priority queue are always scheduled before packets in lower-
priority queues
 Rate-controlled priority queuing
– High-priority queue scheduled before lower-priority queues
– Only if the amount of traffic in the high-priority queue stays below a user-configured
threshold
Static Window Modeling
Assumption
 Assume one link between client and server of rate R
 S: MSS (bits)
 O: object size (bits)
 No retransmissions (no loss, no corruption)
 Fixed congestion window, W segments
A simple one-link network connecting a client and a server
Operation (1 of 2)
 Server not permitted to have more than W unacknowledged outstanding segments
 Server receives request from client
 Server sends W segments back-to-back to the client
 . Server then sends one segment into the network for each acknowledgement it receives
 Server continues to send one segment for each acknowledgement until all of the segments of
the object have been sent
First Case
 Server receives ACK for first segment of first window before completing transmission of
first window
 WS/R > RTT + S/R
Delay = 2RTT + O/R
Static Window Modeling—2
A simple one-link network connecting a client and a server
Second Case
 Server transmits first window's worth of segments before the server receives ACK for first
segment in the window
WS/R < RTT + S/R
 It is a scenario where the propagation delay dominates transmission time
Delay = 2RTT + O/R+ (K-1)[S/R + RTT – WS/R]
Dynamic Window Modeling

TCP Congestion Dynamics
Assumptions
 Server starts with congestion window of one segment
 When it receives an ACK for segment, it increases its congestion window to two segments
 Sends two segments to the client
 Congestion window doubles every RTT
Example
 O/S = 15 segments
 K = 4 windows
 Q=2
 P = min{K-1,Q} = 2
Server idles P=2 times
End-to-End Windows
Limitations
 Cannot guarantee a minimum rate for a session
 Not suited for
– Voice and video
 Window size tradeoff requirements
– Limit no. of packets in subnet
– Full-speed transmission and max throughput
Delay-Throughput Trade-off
Node-by-Node Windows
Unfairness Problem in End-to-end
Long sessions with larger windows take precedence in intermediate devices
Virtual Circuit Windowing

 A separate window for every VC & pair of adjacent nodes along path of VC Main idea
 Receiver avoids accumulation of large no. of packets into its memory
– Slows down permit returns to sender
Backpressure Effect in VCs

Round Robin + Node-by-node
Little's Theorem
Big Questions
 What is the avg no. of customers in the system?
– The "typical" no. of packets either waiting in queue or undergoing service
 What is the avg delay per customer?
 The "typical" time a packet spends waiting in queue plus the service time
Definition
N= l´T
 N = No. of customers
 l = Arrival rate
 T = Time spent by customers (packets) in the system
Interpretation
 Little's Theorem expresses crowded systems
 Large N associated with long customer delays (T) & vice versa
 Not influenced by arrival process distribution, service distribution, service order, etc.
Probabilistic Little's Theorem

Time average
 The time average of a function is found by evaluating a measure space with the average
taken over a time, ΔT
 Pn(t) = Probability of n customers in the system at time t
Statistical (Ensemble) average

 Defined as the number that measures the central tendency of a given set of numbers
 A number of different averages
 Mean, median, mode and range
Probabilistic interpretation
 Little's Theorem admits also a probabilistic interpretation for stationary process
– Time avg replaceable with statistical avg
Application
 Little's Theorem becomes applicable to deterministic and probabilistic systems
– a situation does not exist where the theorem does not hold
– Often termed as law
Little's Theorem; Applications

End-to-end flow control
 Recall that end-to-end windows fail to provide adequate control of packet delay
 Little's theorem helps understand the relation
– Window size
– Delay
– Throughput
Average delay per packet

 n flow controlled sessions in the network with fixed window sizesW1,...Wn
 b = whether piggybacking supported or not
 l = throughput (total accepted input rate of sessions)
Throughput and Delay vs Active Flows

When network is heavily loaded,avg delay per packet increases approximately linearly with the
number of active sessions—the total throughput stays approximately constant
Arrivals as Poisson
M/M/1 system
 The M/M/1 queuing system consists of a single queuing station with a single server
– Communication context: a single transmission line
 Probability distribution of the service time is exponential with mean 1/m sec
Arrivals
 Customers (packets) arrive according to a Poisson process
 A(t) is a counting process that represents the total number of arrivals that have occurred
from to time t
Poisson Process
 A Poisson process is generally considered to be a good model for the aggregate traffic of a
large number of
– Similar and
– Independent users
 Merges n independent & identically distributed arrival processes
 Each process has arrival rate l/n
 So the aggregate process has arrival rate l
 No. of arrivals occurring in disjoint time intervals are independent
 No. of arrivals in any interval of length t is Poisson distributed with parameter lt
Poisson Distribution
Service Statistics
What is service?
 The set of activities performed at the receiving device
 Router
– MAC processing
– Lookup
– Forwarding decision
 Switch
– Header processing
– Port allocation table
Service distribution
 Sn is the service time of the nth customer
 Customer (packet) service times have an exponential distribution with parameter m
 m is also called service rate
 Represents the rate (in customers served per unit time) at which the server operates when
busy
 Service times are mutually independent
 Also independent of all inter-arrival times
 Density function
 Service distribution
Commentary
 In the context of a packet transmission, independence of inter-arrival and service times
implies,
– Length of an arriving packet does not affect the arrival time of the next packet
Exponential Distribution
Memorylessness
 Additional time needed to complete a customer's service in progress is independent of when
the service started
 Time up to the next arrival is independent of when the previous arrival occurred
Arrival Occupancy Distribution
System under change
 Users (packets) come and leave the system
– System under continuous change of occupancy
 It is possible that the times of customer arrivals are in some sense nontypical
Non-Typical Arrival
Typical Arrival
Occupancy distribution
 For M/M/1 systems
 pn = an for n =0,1,...
 Arriving customer finds the system in a "typical" state
 Future arrivals are independent of the current number in the system
Simulating TCP Receive Buffer

Operation
 RFC 1122 identifies host implementation requirements
― Includes receive buffer to cache sequenced data not yet forwarded
INET Support
 It stores bytes and not segments
 Few implementations store segments on the retransmission queue, and others store only the
data bytes
Receiver Window
LastByteSent – LastByteAcked
Receive Buffer Support

 inet::tcp::TCPReceiveQueue::getAmountOfBufferedBytes ()
 Returns the number of bytes currently buffered in queue
 inet::tcp::TCPReceiveQueue::getAmountOfFreeBytes (uint32 maxRcvBuffer)
 Returns the number of bytes currently free (=available) in queue
inet::tcp::TCPReceiveQueue Class
Departure Occupancy Distribution
System under change

 Users (packets) come and leave the system
– System under continuous change of occupancy
 It is possible that the times of customer departures are in some sense nontypical
Non-Typical Departure
Typical Departure
Occupancy distribution
 For M/M/1 systems
 dn = an for n =0,1,...
 For each time the number in the system increases from n to n+1 due to an arrival, there will
be corresponding decrease from n+1 to n due to departure
TCP BER Performance

Operation
 RFC 2581 identifies identifies the operation in the wake of TimeOut
TCPReno::recalculateSlowStartThreshold() [protected, virtual]

{
// set ssthresh to flight size/2, but at least 2 MSS
// (the formula below practically amounts to ssthresh=cwnd/2 most of the time)
uint flight_size = std::min(state->snd_cwnd, state->snd_wnd);
state->ssthresh = std::max(flight_size/2, 2*state->snd_mss);
if (ssthreshVector) ssthreshVector->record(state->ssthresh);
}
tcp_old::TCPReno Class Reference
Problem Set 2
TCP TimeOut
Suppose that the five measured SampleRTT values are 106, 120, 140, 90 & 115 ms. Compute the
EstimatedRTT after each of these SampleRTT values is obtained, using a value of α = 0.125 &
assuming that the value of EstimatedRTT was 100 ms just before the first of these five samples
were obtained. Compute also the DevRTT after each sample is obtained, assuming a value of β =
0.25 and assuming the value of DevRTT was 5 ms just before the first of these five samples was
obtained. Last, compute the TCP TimeoutInterval after each of these samples is obtained.
TCP Flow and Congestion Control

Host A is sending an enormous file to Host B over a TCP connection. Over this connection there is
never any packet loss and the timers never expire. Denote the transmission rate of the link
connecting Host A to the Internet by R bps. Suppose that the process in Host A is capable of sending
data into its TCP socket at a rate S bps, where S = 10 · R. Further suppose that the TCP receive
buffer is large enough to hold the entire file, and the send buffer can hold only one percent of the
file. What would prevent the process in Host A from continuously passing data to its TCP socket at
rate S bps? TCP flow control? TCP congestion control? Or something else? Elaborate.
Virtual Circuit Networks

Basics
 Source-to-destination paths behave much like telephone circuit
 Performance guaranteed
 Network actions along source-to-dest path needed
Operation
 Call setup, teardown for each call before data can flow
 Each packet carries VC identifier
 Every router on source-dest path maintains “state” for each passing connection
 Resources (bandwidth, buffers) allocated to VC
Packets Along the Same Path
Two Links Network Example
Stability Issues in VCs
 Arrival rate on link 1 using the shortest path

 Only one path is used for routing at anyone time if the shortest path update period is much
larger than the time required to empty the queue of waiting packets at the time of an update
Datagram Networks
Basics
 Two packets of the same user pair can travel along different routes
 A routing decision is required for each individual packet
Packets Along Different Paths
Complexity
 Each iteration of link state routing protocols
 n(n+1)/2 comparisons: O(n2)
 More efficient implementations possible: O(nlogn)
Oscillations
 Given these costs, finding new routes resulting in new costs
Input Processing
Basics
 Two key router functions:
 Run routing algorithms/protocol (RIP, OSPF, BGP)
 Forwarding datagrams from incoming to outgoing link
Router Functionality
Router Input
Distributed Switching
 Given datagram dest., lookup output port using forwarding table in input port memory
 Complete input port processing at ‘line speed’

Input port queuing
 Fabric slower than input ports combined
 Queuing may occur at input queues
Input Port Queuing

Output Processing
Operations
 Buffering required when datagrams arrive from fabric faster than the transmission rate
― If Rswitch is N times faster than Rline
 Scheduling discipline chooses among queued datagrams for transmission
Router Output Interface
Output Port Buffering
How much to Buffer?

 RFC 3439: average buffering equal to “typical” RTT (say 250 msec) times link capacity C
 C = 10 Gpbs link
 2.5 Gbit buffer
 With N flows, buffering equal to
Head of Line Blocking

Input Port Overflow
 Fabric slower than input ports combined queuing may occur at input queues
 Queuing delay and loss due to input buffer overflow!
Head of Line
 Queued datagram at front of queue prevents others in queue from moving forward
Scenario
Random Early Detection

Drop Tail
 Conventional tail drop algorithm
 A router buffers as many packets as it can
 Simply drops the ones it cannot buffer
 If buffers constantly full, network is congested
 Tail drop distributes buffer space unfairly among traffic flows
Active Queue Management

 When buffer becomes full or gets close to becoming full
 AQM is intelligent drop network congestion of network packets inside a buffer of NIC
 Often with the larger goal of reducing
Drop Tail
 Conventional tail drop algorithm
 A router buffers as many packets as it can
 Simply drops the ones it cannot buffer
 If buffers constantly full, network is congested
 Tail drop distributes buffer space unfairly among traffic flows
RED Operation
 Monitor avg queue size & drop packets based on probabilities
 If buffer empty, all incoming packets accepted
 As queue grows, P for dropping incoming packet grows
 When buffer full, P = 1 all incoming packets dropped
Operation:
RED with In & Out (RIO)

Background
 Similar to RED, but with two separate probability curves
 Has two classes, “In” and “Out” (of profile)
 “Out” class has lower minimum threshold
 Packets are dropped from this class first
 As avg queue length increases, “In” packets are dropped
 Since best-effort is included in the “Out” class, assured traffic can starve best-effort
Operation
For each packet arrival
if it is an In packet
calculate the average In queue size avg_in ;
calculate the average queue size avg_total ;
If it is an In packet.
if min_in < avg_in < max_in
calculate probability P in
with probability P in , drop this packet;
else if max_in < avg_in
drop this packet.
If it is an Out packet
if min_out < avg_total < max_out
calculate probability Pout;
with probability Pout drop this packet;
else if max_out < avg_total
drop this packet
Operation
Routing Algorithms
Interplay
 Routing algorithm determines end-end-path through network
 Forwarding table determines local forwarding
– at this router
– for IP destination address in arriving packet’s header
Graph abstraction
 Graph: G = (N,E)
 N = set of routers = { u, v, w, x, y, z }
 E = set of links ={ (u,v), (u,x), (v,x), (v,w), (x,w), (x,y), (w,y), (w,z), (y,z) }
Cost
 Cost could always be 1
 Or inversely related to bandwidth
 Or inversely related to congestion
 Cost of path
 (x1, x2, x3,…, xp) = c(x1,x2) + c(x2,x3) + … + c(xp-1,xp)
Algorithms
Key question: What is the least-cost path between u and z?
Routing algorithm: Algorithm that finds that least cost path
Complexity of Link State

Global Routing
 All routers have complete topology, link cost information
 Every node constructs a map of the connectivity to the network in the form of a graph
Shows which nodes are connected to which other nodes
Link State
 Each node independently calculates best path from it to every possible destination in the
network
 The collection of best paths will then form the node's routing tables
 Iterative: After k iterations, know path to k destination
Complexity
 For n nodes
 Each iteration: need to check all nodes, w, not in route discovered set N
 Full-mesh: n(n+1)/2
 Omega Notation: O(n2)
Complexity of Distance Vector

Distributed Routing
 Router knows physically-connected neighbors + link costs to neighbors
 Iterative process of computation
 Exchange of info with neighbors
Key Idea
 From time-to-time, each node sends its own distance vector estimate to neighbors
 when x receives new DV estimate from neighbor, it updates its own DV using B-F equation
Count to Infinity Problem
Link Cost Changes
 Node detects local link cost change
 Updates routing info
 Recalculates distance vector
 If DV changes, notify neighbours
Good news
 At time t0, y detects the link-cost change, updates its DV,

 & informs neighbors
 At time t1, z receives the update from y and updates its table
 It computes new least cost to x & sends neighbors its DV
 At time t2, y receives z’s update, updates
 y’s least costs do not change, y does not send message to z
Bad News!
 Good news travels fast

 Bad news travels slow
 Takes 44 iterations before Z eventually computes its path via Y to be larger than 50
Bad News Causes Loops
At time t0 Y detects the link cost change (the cost has changed from 4 to 60). Y computes its new
minimum cost path to X to have a
cost of 6 via node Z. Of course, we can see that this new cost via Z is wrong
But the only information node Y has is that its direct cost to X is 60 and that Z has last told Y that Z
could get to X with a cost of 5. So in order to get to X, Y would now route through Z, fully
expecting that Z will be able to get to X with a cost of 5
So in order to get to X, Y would now route through Z, fully expecting that Z will be able to get to X
with a cost of 5. As of t1 we have a routing loop—in order to get to X, Y routes through Z, and Z
routes through Y.
A routing loop is like a black hole—a packet arriving at Y or Z as of t1 will bounce back and forth
between these two nodes forever or until the routing tables are changed
Poisoned Reverse
Need
 Bad news travels very slow, especially if the cost change is large
 Ping-pong effect due to looping is undesirable
 Nodes are blindly following what is told to them
 Solution: Tell a small lie!
– Poison the link
Operation
 If Z routes through Y to get to X

 Z tells Y its (Z’s) distance to X is infinite (so Y won’t route to X via Z) table
 This lie prevents the loop
Performance
 Poisoned reverse does not work if more than 3-neighbors are involved in looping
 Other techniques such as packet or broadcast ID are incorporated
Hierarchical Routing; Complexity

Need
 All routers identical with a flat network is not true in practice
 Routers vary
– Connectivity
– Bandwidth
– Resources & Cost
Each network admin wants autonomy
Solution: Make a hierarchical relationship between them.
Methodology
 Collect routers into regions, “autonomous systems” (AS)
 Each AS within an ISP
 ISP may consist of one or more ASes
 In same AS run same routing protocol
 “intra-AS” routing proutersrotocol
– routers in different
AS run differentintra-AS routing protocol
 Gateway router:
– At “edge”
– Has link to router in another AS
 Forwarding table configured by both intra- and inter-AS routing algorithm
 intra-AS sets entries for internal dests
 inter-AS & intra-AS sets entries for external dests
Elastic Aggregates & TE
A Generalized Scenario
Traffic Aggregate
 Suppose that a request arrives for downloading a file of size V bytes
 V bytes must be transferred from s to t
 Number of download requests arriving over T interval is N(T)
V1 , V2 , . . . , VN(T)
Average Requests
 Over the interval T, if EV is the average file size
 Average requests for an aggregate amount
Offered Load
 Dividing both sides by T, we get
– Avg rate at which V(T) grows with time
– Avg rate at which download requests arrive
ρ = λ EV
 ρ = Offered load expressed in bytes/sec
 λ = Average arrival rate of download requests
Optimal Routing
Feasible Routing
 The sum of all flows on a link should stay below the link capacity
x(1) + x(2) + · · · + x(K) ≤ C
 Spare capacity
z = C − (x(1) + x(2) + · · · + x(K))
Optimization Problem
 Given a network and a set of demands, there may be many feasible routes
 To choose one route from a set, define an objective function
 Choose the route that optimizes the objective function
 Optimal routing is the one that maximizes the smallest spare capacity
 Reasonable, because any link in the network has a spare capacity of at least z
 Increases chance that a future demand between any pair of nodes finds sufficient free
capacity.
Limitations of Min Hop Routing
Scenario
Shortest path = Min hop routing

 If weight along each edge (link) set to 1
 Total bandwidth on a route is d × H
 Hmin takes min resources
 Least resource consumption
 H = no. of hops on chosen route
 Demand requires bandwidth d
Disadvantages
 Consider that x uses a and b to reach y
 It results in non-utilization of direct hops between them
 Other source-destination pairs would never use these resources
 Network is partitioned
Formulation of Routing Problem

Shortest path is most congested
 One or more links in a network get congested
– Form sub-paths on shortest path
 Unused bandwidth is available on other links
Routing as a User-Network-Traffic-QoS Phenomenon

Defining Routing Problems
 Shortest-widest path
 Widest-shortest path
 Least-loaded routing
 Maximally loaded routing
 Profile-based routing
.
Minimum Interference Routing
Route interference
 Any chosen route from a router a to another router b can possibly reduce the capacity
available for demands between other node pairs
– Often a phenomenon in ISP backbone sharing
Max Flow
 Maxflow (s, t) is a scalar
 Indicates the maximum amount of traffic that can be sent from s to t
 Exploits all possible paths through the network
– An upper bound on the total bits/sec that can be sent from s to t
Minimum interference
 Idealy zero interference
 If maxflow (s, t) remains unchanged
 Path used for the (a, b) demand does not share any link with the set of paths available for (s,
t)
 Non-zero minimum interference
 Paths share minimum hops
Problem Formulation
 After the (a, b) demand has been routed, the smallest maxflow value among all other (s, t)
pairs is maximized
Example
Consider four flows, w.r.t (a, b).
 (30, 15, 6) corresponds to path P1 for (a, b)
 (12, 19, 8) corresponds to path P2 for (a, b)
 (3, 12, 16) corresponds to path P3
 Route P 2 is the minimum interference route for the (a, b) demand
QoS Routing
Single Stream
 A single stream session comes with
 A given bandwidth requirement
 A specified end-to-end delay requirement
 Arrives at the network
 QoS routing is to find a “good” route for the session
Network Operator
 Wider and holistic objectives
– Minimization of total bandwidth consumed
– Maximization of the smallest spare capacity on the links of the network
Tradeoff
 End to end
 Hop by hop
 Two QoS models for IP packet networks
 IntServ
– Simulate the “virtual circuit” of ATM or frame relay on layer-3
 Sets up an end-to-end route with fixed QoS parameters
 DiffServ
 Defining several common classes of service with associated queue priorities and drop
precedence on a per-hop basis hops
Nonadditive Metrics
Definition
 Nonadditive link metrics cannot be summed over the links of a path to obtain the path metric
 Must be aggregated through another way
 Example: Bandwidth
 Requires d units of BW
 The least available link bandwidth along the path should be d
Application
 Wider and holistic objectives
– Minimization of total bandwidth consumed
– Maximization of the smallest spare capacity on the links of the network
Implications
 What if no path exists?
– S-D get isolated
 What if more than one path exists?
 BW measurement freq & accuracy is a tradeoff
 BW measurement is not exact
Solution
 Choose path with highest Prob of having d units
Additive Metrics; RMB

Definition
 Additive link metrics are summed over the links of a path to obtain the path metric
 Example: end-to-end delay
 If eah link offers t units of delay
 The total links N delay is Nt
Rate-based Mux
 A multiplexer takes input from various streams of traffic and puts them out on a single line
– Used fixed sized frames
 Rate matching of heterogeneous sources is required
 Example: WFQ
Weighted Fair Queuing

 WFQ supports fair distribution of BW for variable-length packets
– Weighted bit-by-bit round-robin scheduling
 Fair allocation of bandwidth
 Each queue receives its configured share of output port bandwidth
Finding Feasible Routes
Network Model
 G(N , L) is the network
 N is the set of nodes
 L is the set of links
 ξ1 = sum of prop delay & maximum TXN time on link l
 Cl be the available capacity on link
 K = source–destination pairs in the network
 Consider a path P through the network between a source router and a destination router
 Capacity on path P = minl∈P Cl
 H(P) = No. of hops (i.e., links) on path P
 Consider a path P through the network between a source router and a destination router
 Capacity on path P = minl∈P Cl
 H(P) = No. of hops (i.e., links) on path P
Problem
 Find, on connection arrival, a route connecting the SD pair
– Rate to be allocated on that route
– Connection’s delay and rate requirements are satisfied
– Capacity constraints are not violated
Upper bound on delay

 Required end-to-end delay
 If all the paths are computed
– Multi-commodity problem
– NP hard
Upper Bound on Performance

Route and Rate Allocation (RRA)
 λ connections distrib over I classes given to G(N , L) network
 ρikλ = number of class i connections for SD pair k
 If not all connections can be admitted due to capacity, select a subset for admission
Problem Formulation
 What is the maximum value (revenue) of the minimum weighted carried traffic (Wmin) that
any RRA algorithm can extract from the network?
Offline Routing (Integer Linear Program)

 sik = carried traffic of class i for SD pair k
 nij = No. of class i connections carried on path j
Non-Rate-Based Multiplexers
Additive Metric
 Non-rate muxes are unlike rate-based
– Rate requirement is relieved
 Other requirements emerge
– Bit error rate
– Packet Loss Probabilities
– Preferential links or paths
Multi-constrained Feasibility Problem

 m additive constraints are given
 Objective: find a path that satisfies all m constraints
Multi-constrained Feasibility Problem

 m additive constraints are given
 Objective: find a path that satisfies all m constraints
 In case several paths satisfying all m constraints are available
 No criterion specified for choosing one from this set of paths
– Not defined as an objective function in the optimization problem
Heuristic Interpretation As Constrained Region
λ(P) = α1λ1(P) + α2λ2(P)

m path metric values λ 1(P), ..., λi(P),...,λm(P), map to a single real value that represents the effective
path length.
Efficient Longest Prefix Match
Operation at Router
 Perform a logical AND of netmask and 32-bit destination IP address in the packet
 If result matches network prefix in the forwarding table entry,
 Next hop is the corresponding entry in table.
 Route lookup a search problem
Longest Prefix Match

 Multiple matches of forwarding table entries to a destination IP address are handled through
LPF
 If there are multiple matches to an IP address
– One matching longest network prefix is returned by the lookup function
Binary Trie
 Forwarding table organized as binary trie
– Essentially a binary tree
 Each vertex at level k corresponds k bits prefix
 Each vertex has 2 children
– k bit prefix expanded to (k + 1) bit prefix
 Route lookup essentially involves tracing 32-bit destination address in the trie to find the
vertex
 The entry in the forwarding table that matches the longest prefix
Sample Forwarding Table & Representation

Trie Traversal for 1000000
Level-Compressed Tries
Traversal Time
 Binary Tree is a graph
 Complexity of depth-first traversals is O(n+m)
 Complexity then becomes O(n + n-1), which is O(n)
Level Compress
 Rather than define a level for each bit of the address
– Define a level for groups of contiguous bits
 A simple case of level compression is to have a level for every K bits
 For N bits in address, then the number of levels is N/K
 Instead of two-way branch from each vertex of the trie 2 K-way branch
 Another view of level compression is to say that a subtree of height k is compressed into one
level
Level Compression or Prefix Expansion

Trie (Retrieval) Traversal for 1000000
Flooding; ARPANET Algorithm

Usage of Flooding
 An algorithm whereby a node broadcasts a topological update message to all nodes
 Sending the message to its neighbors
– Which in tum send the message to their neighbors, and so on
Indefinite Flood
 Transmission of messages never terminates
– Rule: node that receives a message relays it to all of its neighbors except from which
it received
Level Compress
 Instead of two-way branch from each vertex of the trie
– 2 K-way branch
 Another view of level compression is to say that a subtree of height k is compressed into one
level
Indefinite Flood Problem
 A failure of link (1-2) is communicated

to node 3 which triggers an indefinite
circulation of the failure message along
the loop (3,4,5) in both directions
ARPANET Solution
 Store enough information in update messages and network nodes
 To ensure that each message is transmitted by each node only a finite number of times
– Preferably only once
 ARPANET used Sequence Numbers
Operation
 When a node j receives a message that originated at some node i
 Check if its seq no. > seq no. the message last received from i
 Yes: message stored in memory
 Transmit to all its neighbors except sender
 No: discard
Flooding w/o Periodic Updates

Redundancy of Periodicity
 Periodic updates needed because if some updates are sent but not incorporated
 Node crashes
 Transmission errors
– Routing tables become inconsistent
 However under normal circumstances
– Not needed
Need-based Updates
 Zero seq no allowed only when node is recovering from a crash
– Situation where all of the node's incident links are down
– And it is in the process of bringing links up
 Separate seq no. for each origin node
The Problem
 Link (2,3) goes down, then link (1,2) goes down, and then link (2,3) comes up while node 2
resets its sequence number to zero
 Nodes 2 and 3 exchange their (conflicting) view of the status of the directed links (1 ,2) and
(2, 1)
 Both nodes discard each other's update message since it carries a sequence number zero
which is equal to the one stored in their respective memories.
The Solution
 Depending on the lexicographic rule used

– Either the (correct) view of node 2 regarding link (2, 1) will prevail right away
– Or else node 2 will issue a new update message with sequence number 1 and its view
will again prevail
Broadcast without Seq. Nos.
Redundancy of Periodicity
 Periodic updates needed because if some updates are sent but not incorporated
 Node crashes
– Routing tables become inconsistent
 However under normal circumstances
– Not needed
Need-based Updates
 Zero seq no allowed only when node is recovering from a crash
– Situation where all of the node's incident links are down
– And it is in the process of bringing links up
 Separate seq no. for each origin node
The Problem
 Link (2,3) goes down, then link (1,2) goes down, and then link (2,3) comes up while node 2
resets its sequence number to zero
 Nodes 2 and 3 exchange their (conflicting) view of the status of the directed links (1 ,2) and
(2, 1)
 Both nodes discard each other's update message since it carries a sequence number zero
which is equal to the one stored in their respective memories.
The Solution
 Depending on the lexicographic rule used

– Either the (correct) view of node 2 regarding link (2, 1) will prevail right away
– Or else node 2 will issue a new update message with sequence number 1 and its view
will again prevail
Problem Set 1
Topology for SPF Algorithm
Routing Algorithm Complexity

With the indicated link costs, use Dijkstra’s shortest-path algorithm to compute the shortest path
from x to all network nodes.
Inter-AS Connectivity
Operation of Inter-AS Routing Protocols

Consider the network. Suppose AS3 and AS2 are running OSPF for their intra-AS routing protocol.
Suppose AS1 and AS4 are running RIP for their intra-AS routing protocol. Suppose eBGP and
iBGP are used for the inter-AS routing protocol. Initially suppose there is no physical link between
AS2 and AS4.
Operation of Inter-AS Routing Protocols
a. Router 3c learns about prefix x from which routing protocol: OSPF, RIP, eBGP, or iBGP?
b. Router 3a learns about x from which routing protocol?
c. Router 1c learns about x from which routing protocol?
d. Router 1d learns about x from which routing protocol?
Problem Set 1
Switching Fabric Performance in Routers
If the maximum queuing delay is (n–1)D for a switching fabric n times faster than the input line
rates. Suppose that all packets are of the same length, n packets arrive at the same time to the n
input ports, and all n packets want to be forwarded to different output ports. What is the maximum
delay for a packet for the (a) memory, (b) bus, and (c) crossbar switching fabrics?
Subnetting
Consider a subnet with prefix 128.119.40.128/26. Give an example of one IP address that can be
assigned to this network.
Subnet Prefixes
Suppose an ISP owns the block of addresses of the form 128.119.40.64/26. Suppose it wants to
create four subnets from this block, with each block having the same number of IP addresses. What
are the prefixes (of form a.b.c.d/x) for the four subnets?
Fragmentation & Reassembly

Consider sending a 2400-byte datagram into a link that has an MTU of 700 bytes. Suppose the
original datagram is stamped with the identification number 422. How many fragments are
generated? What are the values in the various fields in the IP datagram(s) generated related to
fragmentation?
Simulate QoS Routing

Operation of QoS Routing
Each class of traffic needs a minimum bandwidth path. To avoid oscillations by, QoS routing
modifies the routing algorithms.
Key idea: Established routes continues to use the previous links till new paths (or links) are
discovered
Assumptions
Decision
Support in INET
 Basic DiffServ support
 Current queue modules
– DropTailQueue,
– DropTailQoSQueue
– REDQueue
 Classifier class: BasicDSCPClassifier
 classifyByDSCP() creates new packet classifiers
Simulate Routing Updates

EIGRP
 Cisco’s EIGRP is a hybrid routing protocol between distance vector and link state routing
protocols
 EIGRP offers routing based on composite metric
 Cisco released EIGRP specs as IETF’s RFC draft in 2013
Basic Operation
 EIGRP employs Diffusing Update Algorithm (DUAL)
 Propagates topology change minimizing path compute time
 Sends event-driven partial bound updates
– Weighted (Bandwidth + Delay)
ANSA
Automated Network Simulation and
Analysis
@Brno University of Technology
Czech Republic
EIGRP Simulation Module Structure

Simulate HSRP
Background
 Allows PC to keep communicating on an internetwork even if its default gateway becomes
unavailable
 Works by creating a virtual (phantom) router
– Virtual router has its own IP and MAC addresses
Hot Standby Router Protocol (HSRP)
Basic Operation
 Each PC is configured to use the virtual router as its default gateway
 When a PC broadcasts an ARP frame to find its default gateway, the active HSRP router
responds with virtual router’s MAC address
 Active router sends out HELLO periodically
 If the active router goes offline, a standby router takes over
 HSRP also works for proxy ARP
Automated Main Distribution Frame Housing Routers

Simulate Flooding
Message Complexity
 Flooding is a simple routing algorithm in which every incoming packet is sent through every
outgoing link except the one it arrived on
 Complexity
– M = Ω/(N-1)

 No. of messages at each node
 tictoc14.ned
 txc14.cc
 tictoc14.msg
Simulate TCP with BER

Reference Topology
TCP with BER
 A cross layer paradox
 TCP is for congestion control
 Packet loss due to PHY layer
– BER
 TCP wrongly interprets
– Goes into starvation
Line of Sight effect

 An object within the line-of-sight between two nodes s and b yields a weaker received signal
than that of a non obstructed pair s and a at the same distance
Support in Mixim
 Decider module
– Classifies incoming messages into receivable messages or noise
– Calculates the bit errors for the message
– Info. about current state of channel
PHY Layer Class Graph
Simulate Priority Queues

Recall!
 Designed to provide a relatively simple method of supporting differentiated service classes
 Cqueue provides FIFO by default
 Need to be modified for priority queuing
Operation
 Packets classified and placed into different priority queues
 Packets scheduled from the head of a queue only if all queues of higher priority are empty
 Within each of the priority queues, packets are scheduled in FIFO order

Support in INET
Member Functions
 simple PriorityQueue extends Queue
{ @class(PriorityQueue);
}
 void setSchedulingPriority(short p);
DLL Services
Need
 The datalink layer is to the link what the transport layer is to the path
 Upper layer necessitates its behaviour
― Reliability
― Flow control
― Error control
 Corresponding services must exist
Services Models
 Services offered
― Reliable (PPP)
― Unreliable (Ethernet)
 Point to point
 Multiaccess
Services
 Framing
 Link access
 Error control
 Contention control
EDEC Techniques
Block Diagram
Strategy
Capabilities of EDEC
Constraints
 All EDEC methods only work below a certain error rate
 If we allow any no. of errors in data bits and in check bits, then no EDEC method can
guarantee to work
– Any valid pattern can transform into any other valid pattern
Parity Checks
Operation
 Single bit parity detect single bit errors
― Even
― Odd
Limitations
 Probability of undetected errors in a frame protected by single-bit parity
― can approach 50 percent
 Burst errors cause such nondetections
Checksumming at DLL
Overhead of Parity schemes
 Single bit parity schemes provide little protection
 To provide enough resilience, redundancy increases linearly
 Solution:
― Treat data as k-bit integers
― Generate k-bit overhead
Operation
 RFC 1071 addresses Internet checksum algorithm
 1s complement of all sums of k-bit integers forms the Internet checksum
― 16-bit for TCP/UDP
 Carried in the segment header
Variants
 TCP and UDP: checksum computed over all fields
― Header + data
 IP: IP header
 XTP: one checksum is computed over the header and another checksum computed over
entire packet.
DLL vs Transport
 Transport layer is typically implemented in software
 Error detection has to be simple and fast
― Checksumming
 DLL implemented in NIC
― CRC is more robust
Horizontal & Vertical Parity

2D Generalization
 d bits in D are divided into i rows and j columns
 Parity value computed for each row and for column
 i + j + 1 parity bits comprise DLL frame’s error-detection bits
– 17 bits for 64 bits
• ~27%
Two-Dimensional Parity
Error Correction
 A single error is detectable
– And correctable
 Even an error in the parity bits themselves is also detectable and correctable
– Forward error correction (FEC)
Limitations
 Two-dimensional parity can also detect (but not correct!) any combination of two errors in a
packet
Cyclic Redundancy Check
Principle
 Checksum becomes weak
― Limited illegal rep
 CRC more powerful error-detection code
― Views data bits, D, as a binary number
― Choose r+1 bit G
― Goal: choose r CRC bits, R, so <D,R> exactly divisible by G (modulo 2)
 Receiver knows G,
 Divides <D,R> by G
 All zeros
 No error
 If non-zero remainder: error detected!
Modulo 2
 Modulo-2 arithmetic
 Addition & subtraction are identical
 Both equivalent to bitwise exclusive-or (XOR) of operands
 1011 XOR 0101 = 1110
 1001 XOR 1101 = 0100
Operation
 D.2r XOR R = nG
― Left shift by r then append R
― Multiple of Generator
Mathematical manipulation
 D.2r = nG XOR R
 If we divide D.2r by G, want remainder R to satisfy
D = 101110, d = 6, G = 1001, r =3
9 bits transmitted in this case are 101110 011

Throughput of MAC
Ideal MAC
 Broadcast channel of rate R bps
 When one node wants to transmit, it can send at rate R
― M nodes transmit
― Each sends at average rate R/M
 Fully decentralized
 No special node to coordinate TXNs
 No synchronization of clocks
 No slots
 Simple
Access Methods
Analysis
 Effective throughout depends upon various factors
– No. of active users
– No of resources
– Channel access methods
– Traffic volumes
 Probabilistic in nature
Channel Partitioning
Basic Idea
 Divide channel into smaller “pieces”
― Time slots
― Frequency
― Code
― Space
 Exclusive use
TDM
 Time divion multiplexing
 Access to channel in "rounds"
 Each station gets fixed length slot
 Length = packets trans time) in each round
– Unused slots go idle
TDM Example
Example: 6-station LAN, 1,3,4 have pkt, slots 2,5,6 idle
 Fraction of time slots being used
– Depends upon the frame size
FDM
 Frequency divion multiplexing
― Channel spectrum divided into frequency bands
 Each station assigned fixed frequency band
― Unused transmission time in frequency bands go idle
FDM Example
Example: 6-station LAN, 1,3,4 have pkt, frequency bands 2,5,6 idle
 Fraction of frequency bands being used
– Depends upon the transmission times of each user
Random Access Protocols
Basic Idea
 When node has packet to send
– transmit at full channel data rate R
– No a priori coordination among nodes
 Collisions are legal
– Two or more transmitting nodes cause collision
Packet attempt instants in space and time
Managing Collisions
 How to detect collisions?
― Voltage change
 How to recover from collisions?
― Wait &
― Retransmit
ALOHA
Basic Idea
 Just say as you like!
 Whenever and wherever
― Simplest
― No synchronization
Packet transmissions are independent

Packet reception success dependent upon others not transmitting
Probability of Success
P(success by given node) = P(node transmits) * P(no other node transmits in [t0-1,t0] * P(no other
node transmits in [t0-1,t0]
p . (1-p)N-1 . (1-p)N-1
= p . (1-p)2(N-1)
[choosing optimum p and n very large]
= 1/(2e) = .18
Slotted ALOHA
Basic Idea
 Minimize collisions
– Through synchronization
– Through frame size delimiting
Assumption
 All frames same size
 Time divided into equal size slots
― Time to transmit 1 frame
 Nodes start to transmit only slot beginning
 Nodes are synchronized
― If 2 or more nodes transmit in slot, all nodes detect collision
Operation
 when node obtains fresh frame, transmits in next slot
― if no collision: node can send new frame in next slot
― if collision: node retransmits frame in each subsequent slot with prob. p until success
Performance of Network with 3-Nodes
 30% success
 How many collisions?
 How many empty slots?
Pros
 Single active node can continuously transmit at full rate of channel
 Highly decentralized: only slots in nodes need to be in sync (master clock)
 Simple to implement
Cons
 Collisions, wasting slots idle slots
 Nodes may be not able to detect collision in time
 Clock synchronization needed
Probability of Success
 N nodes with many frames to send, each transmits in slot with probability p
 Prob that given node has success in a slot = p(1-p)N-1
 Prob that any node has a success = Np(1-p)N-1
 Max efficiency: find p* that maximizes
Np(1-p)N-1
 for many nodes, take limit of Np*(1-p*)N-1
― N goes to infinity
 Max efficiency = 1/e = .37
CSMA/CD
Basic Idea
Carrier Sensing
 Listen before transmit
 If channel sensed idle
– Transmit entire frame
 If sensed busy
– Defer transmission
Collisions Detection
– Within short time
 Colliding transmissions aborted
 Reduces channel wastage
CSMA/CD States
 Contention
 Transmission
 Idle
Binary (Exp) backoff
 After mth collision, NIC chooses K at random from {0,1,2, …, 2m-1}
 NIC waits K·512 bit times, returns to Step 2
― If idle, start trans
― If busy wait until idle, then transmits
 Longer backoff interval with more collisions
CSMA/CD Efficiency
Factors Affecting (1 of 2)
 Tprop = max prop delay between 2 nodes in LAN
 ttrans = time to transmit max-size frames
 Full load
– Worst
 Partial load
– Increases till a range
 No load
– Poor performance
 Efficiency goes to 1
 As tprop goes to 0
 As ttrans goes to infinity
 Efficiency goes to 0 vice versa
1
Efficiency =
1+ 5t prop /ttrans
Performance with Variants

Min Frame Size Computation
Min Frame Size
 Ethernet recommends 64 Bytes
– Inclusive of headers
 If the data portion of a frame < 46 bytes
 Pad field is used to fill out the frame to min. size
 Wireless networks necessitate lesser size
– ReTx cost to be minimized
 Optical networks require longer frames
Reasons
 Data field of 0 bytes is sometimes useful
– When a transceiver detects a collision, it truncates the current frame
 Stray bits and pieces of frames appear on the cable all the time
 To distinguish valid frames from garbage
 Collision detection can take as long as 2 τ
Reasons
 Prevent a station from completing the transmission of a short frame before the first bit has
even reached the far end of the cable
– where it may collide with another frame
Ethernet Calculation
 10-Mbps LAN
 Max length = 2500 m (four repeaters: 802.3 specs)
 RTT = 50 μsec in the worst case
― Therefore, the minimum frame must take at least this long to transmit
 At 10 Mbps, a bit takes 100 nsec
― 500 bits is the smallest frame that is guaranteed to work
 To add some margin of safety, round up to 512 bits or 64 bytes
Max Frame Size Computation

Background
 Ethernet recommends 1500 bytes as Max
 This limit was chosen arbitrarily for DIX standard
 Transceiver needs enough RAM to hold an entire frame
– More expensive transceivers
Factors
 Overhead
 Pipelining
Overhead of Variable Length Packet

 Each frame contains V as overhead bits
 Kmax:Max length of packet
 Message of length M
– Broken down into M/Kmax packets
Overhead of Variable Length Packet (2 of 2)

 As Kmax decreases, the number of frames increases
– Thus the total overhead in the message
 Processing load in a multiple hop network increases
Pipelining
 Split the message into smaller packets
– While the later packets arrive on the input queue of the node
– Former packets are leaving or may have already left the output queue
Pipelining Scenario
 Decreasing delay by shortening packets to take advantage of pipelining
Pipelining Scenario
 The total packet delay over two empty links equals twice the packet transmission time on a
link plus the overall propagation delay
Pipelining Scenario
 When each packet is split in two, a pipe lining effect occurs
 The total delay for the two half packets equals 1.5 times the original packet transmission
time on a link plus the overall propagation delay
Tradeoff between Overhead & Pipelining

 As the overhead V increases, Kmax should be increased
 As the path length j increases, Kmax should be reduced
Transmission Errors
 Large frames have a somewhat higher error probability than small frames
 Probability of error on reasonable-sized
 frames is on the order of 10-4 or less
– This effect is typically less important than the other effects
Fixed Frame Size Computation

Why Fixed Frame?
 Expectability of performance
– Latency
– Throughput
– Cell loss
 Resource pre-emption
Considerations
 How much should be the fixed size?
– Processing at the nodes
 Header to payload efficiency
– Padding requirement
 Applications (Voice/video)
– Achieve a small delay for stream-type traffic
 Assume an arrival rate of R and a packet length K
– First bit in a packet is then held up for a time K/R
– Waiting for the packet to be assembled
Fixed Frame (cell) networks

 ATM recommends 53 bytes (424 bits) as Max
– 48 bytes payload
– 5 bytes header
 Emulates circuit-like behaviour
– Good for interactive
– Bad for file transfer
Multi-Protocol Label Switching

Intro
 Multi Protocol Label Switching (MPLS)
– Fast packet switching & routing
– Provides designation, routing, & switching of traffic flows through MPLS
domain
 All packets labelled before being forwarded
 Network layer header not processed
 Although idea was to facilitate fast packet switching
 Main goal: support traffic engineering and QoS
Basic Idea
 Route once and switch many times
 Set of packets that have the same traffic characteristics are forwarded in the same manner
– Along the route that starts from an ingress node and ends at an egress node of an
MPLS network
MPLS Network Components
MPLS Enhanced Forwarding
Important Parameters
 Link utilization
 Voice jitter
 End to end delay
 Traffic Received when FRR vs link failures
Load Balancing in Data Centre

Intro
 Google, Microsoft, Facebook, and Amazon have built massive data centers
 Each houses tens to hundreds of thousands of hosts
 Concurrently support many distinct cloud applications
 Search, email, social networking, and e-commerce
 Top of Rack (TOR) switch interconnects the hosts in the rack
 With each other
 With other switches in the data center
 Form a data centre network
Basic Idea
 Each application is associated with a publicly visible IP address
 Clients send their requests and receive responses
 Inside, external requests first directed to load balancer
 Distributes and balances requests to hosts
– Also called L4 switch (with NAT)
Problems in Hierarchical Topology
Limited Host to Host Connectivity (1 of 2)

 40 simultaneous flows between 40 pairs of hosts in different racks
– 10 hosts in rack 1 sends a flow to a corresponding host in rack 5
– 10 flows between pairs of hosts in racks 2 and 6, 3 and 7, and 4 and 8
 40 flows crossing the 10 Gbps A-to-B link (and B-to-C link) each only receive 10 Gbps / 40
= 250 Mbps
Solution: Fully Connected Topology

Correctness of Stop and Wait
Stop and Wait Operation
 Client sends request
– Waits for response
 Server sends response
– Waits for ack
 Step-locked communication
 Most web and other servers based upon it
– Pipelining is deviation
Problems with Unnumbered Packets
Problems with Unnumbered ACKs
Problems Management using Seq. Nos.

Efficiency of Go Back N
Stop and Wait Operation
 Client sends request
– Waits for response
 Sends next request
 Each request travels all along the way to server
 Response travels backwards
Efficiency of Stop & Wait Operation
 8000 bit packet

 If RTT=30 msec, 1KB pkt every 30 msec 33kB/sec throughput over 1 Gbps link
Utilization of Go-Back N under No Loss

U 3L / R .0024
sender = = = 0.00081
RTT + L / R 30.008
Limitations of Go Back N
 Retransmissions, or delays waiting for time outs, occur in go back N due to following
 Errors in the forward direction
 Errors in the feedback direction
 Longer frames in the feedback than in the forward directions
Effect of Long Frames in Reverse Direction

 Ack for packet 1 does not arrive at the sending side by the time packet 6 finishes
transmission, thereby causing a retransmission of packet 0
 Probability that a frame is not acked by the time the window is exhausted is given by
Character-based Framing
Character Codes
 Character codes such as ASCII provide binary representations
– Keyboard characters and terminal control characters
– Also for various communication control characters
SYN Idle
 A string of SYN characters provides idle fill between frames when a sending DLC has no
data to send
– But a synchronous modem requires bits
STX and ETX

 STX (start of text) and ETX (end of text) are two other communication control characters
– Used to indicate the beginning and end of a frame
Simplified Frame Structure

Problem
 The header or the CRC might (through chance) contain a communication control character
– Since these always appear in known positions after STX or ETX, (no problem for the
receiver)
 The payload might contain ETX character
– Interpreted as ending the frame
Transparent Mode
 The transparent mode uses a special control character called DLE (data link escape)
― Inserted before the STX character to indicate the start of a frame in transparent mode
― Also inserted before intentional uses of communication control characters within
such frame
Bit-oriented Framing
Bit-oriented Protocols
 Bit-oriented synchronous protocol pass variable-length frames
– Image/voice data
– Web data
 Dedicated or switched Simplex, half and full duplex
Flags
 8-bit sequence (01111110) that delimits a frame's
– Start and End
 Procedure
– When DLL detects seq of 5 1s in a row in user data
– Inserts a 0 immediately after the 5th 1 in transmitted stream
 DLL at receiver removes inserted 0s by looking for seq of 5 1s followed by stuffed 0s
 Problem
 Confusion between possible appearances of the flag as a bit string within frame and actual
flag indicating end of the frame
Bit-Stuffing Example
 The frame after stuffing never contains more than five consecutive 1's
― Hence flag at the end of the frame is uniquely recognizable
Framing with Errors
Problems with framing
 Several peculiar problems arise
 When errors corrupt the framing information on the communication link
– Flagging
– CRC
– Length field
Flags
 If an error occurs in flag at end of a frame
– The receiver will not detect the end of frame
– Does not check the cyclic redundancy check (CRC)
 When next flag detected, receiver assumes CRC to be in position preceding flag
 This perceived CRC might be the actual CRC for the following frame
 But the receiver interprets two frames as one
 Receiver fails to detect the errors with a probability 2-L
 L is the length of the CRC
False Flag Example

Bits before the perceived flag are interpreted by the receiver as a CRC
– Accepting a false frame
–
 Called the data sensitivity problem of DLC
– Even though the CRC is capable of detecting any combination of three or fewer
errors
– A single error that creates or destroys a flag plus a special combination of data bits to
satisfy the perceived preceding CRC, causes an undetectable error
Length Fields
Purpose of Length Field
 Basic problem in framing is to inform the receiving DLC where each idle fill string ends
– Where each frame starts
– Where each frame ends
 Include length field in the frame header
Overhead of Length Field

 If the length is represented by ordinary binary numbers
 No. of bits in the length field has to be at least
 L = log2[Kmax+1]
– Kmax is the maximum frame size
Problems with Length Fields

 An error in this length field causes receiver to look for the CRC in the wrong place
– An incorrect frame is accepted with probability 2-L
– L is the length of the Length field
 Receiver does not know where to look for subsequent frames
Partial Solution-1
 DECNET uses a fixed-length header for each frame
– Places length of frame in header
– Header has its own CRC
 Limitation: transmitter must still resync after such an error
 Receiver will not know when next frame starts
Partial Solution-2
 A similar approach is to put the length field of one frame into the trailer of preceding frame
– Avoids inefficiency of the DECNET approach
– Requires special synchronizing seq after each detected error
Topology and Connectivity

Topology
 Physical connectivity
– Star
– Hub
– Mesh
– Bus
– Tree
 Connectivity is implied
 Wireless networks have constrained
– Topology
– Connectivity
Ad hoc Networks
 No infrastructure
 Nodes themselves
– Transmit
– Receive
– Relay (forward)
 An operational area in which nodes randomly placed
 Locations follow a spatial distribution
 Must communicate with neighbors
– Certain power
Spatial Reuse vs Connectivity

 The transmission range in the network is large
― At a time at most one transmission occurs
 With smaller transmission ranges, many transmissions can occur simultaneously
― Spatial reuse
― Multihop
Feasibility Region
 x1: location of the first node
 x2: location of second node
 Nodes distributed uniformly in [0, z]
x1 ≤ x2
 Two-node network
connected if x2 − x1 ≤ r
 Transmission range of
every node: r(n) ,
where n is the number
of nodes in network
Link Scheduling & Capacity

Hidden Terminal Problem
 Wireless nodes are blind
 Carrier sensing is hard
 Collision detection is harder
Link Scheduling
 MACA
 MACAW
Network Capacity
 Sum of all active connections
― Simultaneous
― Non interfering
 Varies with time
 Protocol design determines the effectiveness
Scheduling Constraints
Underlying Assumptions
 Multihop wireless network
 Topology has already been discovered
 Directed graph G(N , E)
― N is the set of nodes
― E is the set of directed edges
 An edge (i, j) ∈ E
 Transmission from i , addressed to j
 Decoded by j, provided that the SIR at j is adequately high
Constraints
 The edges can be grouped into subsets
― Edges in a subset can be activated in the same slot
― Receiver in each edge can decode the transmission from the tail (TX) node of the
edge
 Slotted time
 When such a set, S is activated one packet can be sent across each edge in S
Independent Sets
 S1 = {(1, 2), (5, 6), (3, 4)}
 S2 ={(2, 3), (1, 5)}
 S3 = {(2, 3), (4, 5), (1, 6)}
Centralized Scheduling
Scheduling Problem
 Schedule specifies a seq of independent sets to be activated
 Static link activation schedule
 Allocates MS slots to independent set S
 BW allocation follows
Maximum Schedulable Region

 Set of all such flow rates λ by L
– Flow on each link be less than the average link capacity under the schedule
Bluetooth Example
 Piconet is a centralized TDM system
 Master controls the clock
 Determining which device gets to communicate in which time slot
Marginal Buffering at Every Hop

Definition
 Multiplexer has no buffer to store data arriving in a slot but cannot be served in that slot
 Performance depends only on marginal distribution of arrival process
 Doesn't depend on correlations b/w arrivals in slots
Simple Analogy
 The basic idea of “bufferless” multiplexing/routing is
– Always forward a packet to an output port regardless of success
Multiplexer Network Scenario
 Traffic flow from location 1 to locations

2 and 3
 And from location 2 to location 3
– Old and new traffic
causes superposition
Comments
 Packet switching is unachievable with zero buffering
 At least the header of a packet needs buffer
– Cut through
 Mostly store-and-forward switching
– An arriving packet entirely copied into switch from input to output links
Arbitrary Buffering at Every Hop

Arbitrary Buffering
 Connection admission control with burst scale buffering
 Leaky bucket shaped sources and QoS requirements
Buffering constraints
 An arriving stream connection may or may not be admitted, if traffic is already being carried
by the link
 Problem is exacerbated for multihop links
Scenario for Arbitrary Buffering
 Voice at loc 1 destined

for loc 2 enters network
at router 1 & leaves at
router 2
 Voice at loc 1 destined
for loc 3 leaves the link
from router 1 to router 2
and enters the link from
router 2 to router 3
 Here two-hop traffic
multiplexed with data
from loc 2 to loc 3
Comments
 Traffic from a source may be well characterized at the point where it enters the network
 After multiplexing at the first hop, the flows become dependent
– This dependence is very difficult to characterize
Problem Set 1
Effect of BER on Channel Performance
Suppose that an 11-Mbps 802.11b LAN is transmitting 64-byte frames back-to-back over a radio
channel with a bit error rate of 10-7 . How many frames per second will be damaged on average?
Ethernet Framing
A 1-km-long, 10-Mbps CSMA/CD LAN (not 802.3) has a propagation speed of 200 m/μsec.
Repeaters are not allowed in this system. Data frames are 256 bits long, including 32 bits of header,
checksum, and other overhead. The first bit slot after a successful transmission is reserved for the
receiver to capture the channel in order to send a 32-bit acknowledgement frame. What is the
effective data rate, excluding overhead, assuming that there are no collisions?
CSMA/CD Backoff Algo Performance

Two CSMA/CD stations are each trying to transmit long (multiframe) files. After each frame is
sent, they contend for the channel, using the binary exponential backoff algorithm. What is the
probability that the contention ends on round k, and what is the mean number of rounds per
contention period?
Problem Set 1
Operation of MAC Addressing
Suppose nodes A, B, and C each attach to the same broadcast LAN (through their adapters). If A
sends thousands of IP datagrams to B with each encapsulating frame addressed to the MAC address
of B, will C’s adapter process these frames? If so, will C’s adapter pass the IP datagrams in these
frames to the network layer C? How would your answers change if A sends frames with the MAC
broadcast address?
Performance of ALOHA
Suppose four active nodes—nodes A, B, C and D—are competing for access to a channel using
slotted ALOHA. Assume each node has an infinite number of packets to send. Each node attempts
to transmit in each slot with probability p. The first slot is numbered slot 1, the second slot is
numbered slot 2, and so on.
a. What is the probability that node A succeeds for the first time in slot 5?
b. What is the probability that some node (either A, B, C or D) succeeds in slot 4?
c. What is the probability that the first success occurs in slot 3?
d. What is the efficiency of this four-node system?
Switch Learn-ability
Consider a network in which 6 nodes labeled A through F are star connected into an Ethernet
switch. Suppose that (i) B sends a frame to E, (ii) E replies with a frame to B, (iii) A sends a frame
to B, (iv) B replies with a frame to A. The switch table is initially empty. Show the state of the
switch table before and after each of these events. For each of these events, identify the link(s) on
which the transmitted frame will be forwarded, and briefly justify your answers.
Simulate Parity Scheme Failure

Support in INET (Channel Behaviour)
 BER & PER allow basic error modelling
 When channel decides (based on RN) that an error occurred during transmission of packet
 Sets an error flag in the packet object
Support in INET (Rx Behaviour)

 The receiver module is expected to check the flag
 Discard the packet as corrupted if it is set
– Default BER and PER are zero
Typical Example
• channel Ethernet100 extends ned.DatarateChannel
{
datarate = 100Mbps;
delay = 100us;
ber = 1e-10;
}
Failing Parity Scheme

 Need to hardcode the pattern that fails parity scheme
 The data pattern must be known
– So that a corresponding error model can be designed
Simulate ARP Behaviour

Scenario
 Client computer opens TCP session with server
 Rest of operations (including ARP) follow
– ARP has to learn the MAC address for the default router
Design Tour of INET 3

arpTest.client.eth[0].arp
Inside ARP Packet
ARP Packet Class (Generated by .msg file)

// file: ARPPacket.msg
message ARPPacket
{
fields:
int opcode enum(ARPOpcode);
MACAddress srcMACAddress;
MACAddress destMACAddress;
IPAddress srcIPAddress;
IPAddress destIPAddress;
};
Packet Queue (Contains IP Packet)

ARP Cache Build-up
ARP Variants
 ARP Broadcast-unicast behaviour
 Proxy ARP
 Gratuitous ARP
 Reverse ARP
Performance
 No of broadcast attempts
 No of successes
 Effect of network size
 Multihop performance
Output Analysis on WireShark

Wireshark
 A packet capturing & analysis tool
― Work in promiscuous mode
 Presents output in Binary, Hex and ASCII
 Saves files as .pcap
Packet Capture Process

Wireshark Interface
Example
inet/examples/inet/tcpsack
 Sets up a flow between two hosts with TCP Sack
 Outputs files in multiple formats,
 Including the pcap format
Simulate Switching vs Routing

Why compare!
 Routing is inter-network phenomenon
― It is pre-forwarding
 Switching is intra-network
― It is forwarding
 Apparently no comparison
 Comparison at the device level
― Router vs switch
Router vs Switch
 Routing process
― Forwarding process
 Switching process
― Port-based MAC learning
 ID-based behaviour
― Unicast
― Broadcast
Basis of Comparison
 Cost
― All router
― All switch
― Hybrid
 Isolation
― Traffic
― Domain
 Speed
 Complexity
Parameters
 Output queue lengths
 Output queue length distribution
 Output queue length Vs time plots
 Number of packets generated and received by hosts
 Packet size distribution
 Hop count distribution
 End to end delay
A Router (or Switch) Package
Overview of Access Technologies

Broadband Access
 Broadband is longhaul (backhaul)
– Shared medium
– Long distance
 Vs access side (baseband)
 Lastmile (first mile)
– User-connecting technologies
Taxonomy of Packet Technologies
Taxonomy of Wireless Technologies

WiFi
WLAN Protocol Stack
The Hidden (Exposed) Station Problem

RTS CTS Mechanism
 Sender sends request to send
 Receiver acknowledges as clear
– Overhearing neighborhood cautioned
WLAN Configuration
WiFi Operations
Operations
 Synchronization
 Authentication
 Association
 Data Transmission
 Handoff
 Power management
Scanning for APs

Mobility in the Same IP Subnet
 H1 moves from BSS1 to BSS2
 Keeps its IP address
― And all of its ongoing TCP connections
Mobile IP
Degrees of Mobility
Mobile IP Standard
 RFC 3344
 Elements
― Home agents,
― Foreign agents,
 Foreign-agent registration
 Care-of-addresses
 Encapsulation (packet-within-a-packet)
Elements of Mobile IP System
Procedures
 Agent discovery
 Registration with home agent
 Indirect routing of datagrams
Indirect Routing
Packet Cable Networks
Background
 Packet broadband cable network
― Built on existing broadcast cable TV (CATV) networks
 Hybrid fiber coax (HFC) cable networks
― Deployment of optical fiber
― New amplifier technology
 Alternative to DSL
Architecture
 Tree topology
 One-way broadcast
 Headend and cable modems
Headend
 Operational center of a CATV cable access network
 Connected to many distribution nodes via trunk cables
― Coax cable or fiber
Components
Functions of Headend
 Receiving broadcast signals from satellite or microwave dishes
 Mixing local or recorded TV programming
 Assigning channel frequencies to all signals destined for cable distribution
Functions of CMTS
 Controlling bandwidth allocation for data traffic to each modem
 Enforcing bandwidth allocation policy
 Assigning a time slot to each cable modem for transmitting upstream messages
 Enforcing QoS policies such as traffic shaping and policing (packet classification based on
QoS classes)
Cable Modem Network Configuration
 Cable Model Systems accommodates two way communication
 DOCSIS (data over cable service interface specification)
WiMax
Background
 IEEE 802.16 is an emerging wireless MAN technology
 Originally designed to provide wireless last mile/first mile deployment in a MAN
 Also end-user access an alternative to 802.11 family
 Mobility support provided
Introduction
 Worldwide Interoperability for Microwave Access (WiMAX)
 Many basic ideas of 802.16 borrowed from DOCSIS/HFC applied to the wireless setting
 Good analogy : Wi-Fi : Ethernet :: WiMAX : DOCSIS/HFC
Architecture
 Line-of-Sight(LOS) and tens of Ghz spectrum
 Severe atmospheric attenuation
– Suitable in operator network between two nodes with high bandwidth
Many base stations deployed at elevated positions
Components
Digital Subscriber Line
Background
 A family of technologies for broadband last-mile solution using existing copper wires
Introduction
 Based on two premises
– Discrete multitone (DMT) line code
– Widely deployed twisted pair
 Provides upto 7 Mbps (suitable for Internet)
 Flexible bandwidth allocation per user demand
 Dedicated vs CATV
Architecture
 Enterprise CPE includes an integrated access device (IAD)
 Or connected through Feeder Distribution Interface
DSL Family
Wireless Personal Area Networks

Introduction to LR-WPANs
 Low-rate low-power wireless personal area networks
– Types of wireless sensor networks
 Applications
– Industrial control & monitoring
– Environmental & health monitorinG
 Home automation, entertainment & toys
 Security, location and asset tracking
 Emergency and disaster response
Comparison
 IEEE 802.15.4
– A new MAC for LR-WPAN
 IEEE 802.11: an “overkill technology”
 Bluetooth: High data rate for multimedia applications
 Small size network
 High power consumption
ZigBee vs Bluetooth
 Smaller packets over large network
 Mostly Static networks with many, infrequently used devices
 Larger packets over small network
 Ad-hoc networks
IEEE802.15.4
Features
 Channels
– 16 channels in 2450 MHz band
– 10 channels in 915 MHz
– 1 channel in 868 MHz
 Over-the-air rates of 250,40& 20 kb/s
 Addressing
 16 bit short
 64 bit extended
 Allocation of guaranteed time slots (GTSs)
 CSMA-CA channel access
 Fully acknowledged data transfer
 Low power consumption
 Energy detection (ED)
Link quality indication (LQI)
Topology Models
Radio Frequency Identification
Introduction
 Presence known if within a certain radius
― Object identified
 Do not know exactly the position
Application Areas
Architecture
Traffic Flow

CS432 Handouts PDF

Uploaded by

Copyright:

Available Formats

CS432 Handouts PDF

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

CS432 Handouts PDF

Uploaded by

Copyright:

Available Formats

Booklet of Network Modeling and Simulation

Dr. Ali Hammad Akbar

Video lectures of short

duration (5-7 minutes)

Mid-term and final

Need for NeMS

The model definition could be

The computer models can be described into various types

Simulation Building Process

Inappropriate levels of details

Improper programming language

Improper initial conditions

Short run times

Poor random number generators

Inadequate time estimate

Incomplete mix of essential skills

Inadequate level of user participation

Inability to manage simulation project

Over reliance on link budget methods for abstraction

Overly simplistic modeling of radio layers

Development of Systems Simulation

A “Still I am not dead yet!” scenario

Available Not available

/* Height of an object moving under gravity. */

Data collection & analysis

Model validation, verification, & calibration

Model validation, verification, & calibration

Life cycle of Simulation Development

Recommended Text and References

NeMS contents cover

NeMS contents don't cover

Getting a free copy

Compile & Install

Minimum GNU environment for Windows

Compile & Install

Running first time

 Simple modules can be grouped into compound modules and so forth

No. of hierarchy levels not limited

 Message; tuple (time stamp, arbitrary data, ...)

Internal architecture of OMNET++

OMNeT++ simulation programs possess a modular structure.

Simulation Kernel & SIM Class Library

Envir, Cmdenv and Tkenv Libraries

Typical Ingredients of NED description

Simple Module Declaration

Let us create a topology called My_Network using Graphical Editor

More About NED Language

MobileHost compound module

Design time IMobility

Configuring OMNET++ Simulations

Separation of Model and Experiments

INI File Syntax

INI File Editor

INI File Editor

Using GUI Project Builder

What is Simulation Run?

OMNET++ IDE Features

Run queuenet Launch

Animation and Tracing

Object inspector Timeline