An Introduction to Quality of Service (QoS)

and Quality of Experience (QoE)

Network Architecture (INWK 6115)

1
 Module 6: Quality of Service Routing

• QoS definition
• QoS protocols
• QoS architectures
• Protection schemes
• Routing over survivable networks
• Priority-aware routing mechanisms
• SLA-based routing mechanisms
• QoS-based algorithms over WDM networks
• Multicast routing
2
 IP-based Networks - Internet Today
• Internet today
– Provides “best effort” data delivery
– Complexity stays in the end-hosts
– Network core remains simple
– As demands exceeds capacity, service degrades gracefully
(increased jitter etc.)
Delivery delays cause problems to real-time applications

3
 QoS Definition
• If a request has certain resource requirements that it explicitly
announces to the network at the time of arrival, then QoS refers to
the network's ability to meet the resource guarantees for this
request.
• The goal :
Provide some level of predictability and control beyond the
current IP “best-effort” service
• Fundamental principle
Leave complexity at the “edges” and keep network “core”
simple

4
 QoS routing
• QoS routing refers to a network's ability to accommodate a
QoS request by determining a path through the network that
meets the QoS guarantee.

• network's performance is also optimized

• QoS routing cannot be completely decoupled from traffic

engineering

5
 QoS Metrics
• Request attributes
• Holding time
• Service availability
• Resource requirements (bandwidth, wavelength, node,…)
• priority
• Performance attributes
– Delay
– Delay variation (jitter)
– Throughput
– Packet loss rate
– Blocking probability
– Availability satisfaction rate
– Resource overbuild
6
• Service Level Agreement (SLA) parameters
 QoS Attributes
• Additive/non-additive attributes
• Single attributes
• Link cost function
• Distance, bandwidth, number of wavelengths, delay, hop
• Shortest path algorithm
• Multiple attributes
• Optimization problem
• Heuristics/linear programing
• Constrained-based shortest path routing

7
 Service Level Agreement
• SLA is a formal contract between a service provider and a
subscriber
• An SLA Contains technical and non-technical terms and conditions

• A set of parameters and their values that together define the

service offered to a traffic stream in a network

• SLA parameters: Source, destination, link availability, path

availability, request priority, and the number of available
wavelengths

• Negotiating SLA parameters: link attributes, path attributes, cost

functions, and requests’ attributes

8
 Service Level Agreements (SLA)

QUALITY OF SERVICE PARAMETERS

Service Level Application Priority Mapping

1 • Non-critical data • Best-effort delivery

• Similar to Internet today • Unmanaged performance
• No minimum information rate
guaranteed

2 • Mission-critical data • Low loss rate

• VPN outsourcing, e- • Controlled delay and delay
commerce variation
• Similar to ATM VBR

3 • Real time applications • Low loss rate

• Video streaming, voice, • Low delay and delay variation
videoconferencing

9
 QoS Protocol Classification
• QoS can be achieved by :
– Resource reservation (integrated services)
– Prioritization (differentiated services)
– Existing protocols with extensions (OSPF-TE)
• QoS can be applied :
– Per flow (individual, uni-directional streams)
– Per aggregate (two or more flows having something in
common)

10
 QoS Protocols

• ReSerVation Protocol (RSVP)

• Differentiated Services (DiffServ)

• Multi Protocol Labeling Switching (MPLS)

• SLA-based algorithms

• Priority-aware algorithms

11
 RSVP
- Resource Reservation
• Attributes
– The most complex of all QoS technologies
– Closest thing to circuit emulation on IP networks
– The biggest departure from “best-effort” IP service
• Provides the highest level of QoS in terms of :
– Service guarantees
– Granularity of resource allocation
– Detail of feedback to QoS-enabled applications

12
 RSVP
- Integrated Services

• Enables integrated services (IntServ)

• IntServ types
– Guaranteed : as close as possible to a dedicated virtual
circuit
– Controlled Load : equivalent to best-effort service under
unloaded conditions

13
 RSVP
- Implementation

Qo
S
and Lev
el
fic Filt
Traf ification er S
pec
c
Spe ifica
tion
H RE
PAT SV

Host A Host B

14
 RSVP
- Implementation
• Sender
– PATH message containing
• traffic specification (bitrate, peak rate etc.)
• Receiver
– RECV message containing
• the reservation specification (guaranteed or controlled)
• the filter specification (type of packets that the
reservation is made for)

15
 RSVP
- Conclusions
• Reservations are “soft”
– Periodic refresh is necessary
• It is a network (control) protocol
– Works in parallel to TCP and UDP

• APIs are required to specify flow requirements

• Reservations are receiver-based
• Has to maintain a state for each flow
• Multicast reservations
– Merged at replication points, difficult to understood algorithms have to be
used though

16
 DiffServ
- Prioritization
• Description
– Applied on flow aggregates
– Services requirements are classified
– Classification is performed at network ingress points
– A predefined per-hop behavior (PHB) is applied to every
service class
– Traffic is smoothed according to PHB applied

17
 DiffServ
- Traffic Classes
Two traffic classes are available :
– Expedited Forwarding (EF) - 1 codepoint

• Minimizes delay and jitter

• Provides the highest QoS

• Traffic that exceeds the traffic profile is discarded

– Assured Forwarding (AF) - 12 codepoints

• 4 classes, 3 drop-precedences within each class

• Traffic that exceeds the traffic profile is not delivered with such
high probability

18
 DiffServ
- Implementation

Classifier Conditioner

Maps DSCPs to Applies the

PHBs defined PHB
Marker Meter
(scheduling)

Maintains Accumulates
DSCP statistics
mappings and
associations
with local
policies

19
 DiffServ
- Implementation
• DiffServ codepoints (DSCPs) redefine the Type-of-Service (ToS) IPv4 field

• Precedence bits are preserved

• Type-of-Service bits are NOT

0 1 2 3 4 5 6 7 0 1 2 3 4 5 6 7

DSCP CU Precedence Type of Service MBZ

Class Selector Unused RFC 1122 RFC 1349

Must
Be
Zero
Differenciated Services
Codepoint (DSCP) IP Type of Service (TOS)

20
 DiffServ
- Conclusions
• Traffic classes are equivalent to IP precedence service
descriptors
– DiffServ unaware routers pass-through DiffServ traffic

• Easy to be implemented / integrated even into the network

core.
• Proper classification can lead to efficient resource allocation
and though improved QoS

21
 MPLS
- Label Switching
• Used to establish fixed bandwidth routes (similar to ATM
virtual circuits)
• Resides only on routers and is protocol independent
• Traffic is marked at ingress and unmarked at egress
boundaries
• Markings are used to determine next router hop (not priority)
The aim is to simplify the routing process …

22
 MPLS
- Implementation
• The 1st hop router, using the header information (destination address
etc.) attaches a label and forwards the packet

• Every MPLS-enabled router uses the label as an index to a table defining

the next hop and label

20 3 1 8

Label Value Exp . S TTL

20-bits : Label value used for lookup 3-bits : Reserved 8-bits : Time-To-Live
1-bit : Bottom of Label Stack

23
 QoS Architectures

Host A Host B

Application Application
QoS-enabled
Presentation Presentation
Application
Top-to-Bottom QoS

Session Session QoS API

Transport Transport RSVP

Network Network DiffServ

Data Link Data Link SBM

Physical Physical

SBM

RSVP DiffServ and MPLS RSVP

End-to-End QoS

24
QoS routing algorithms for
survivable shared mesh
WDM networks

25
 Transport network concept
End
use
rs
End
use
rs

Intermediate
End s
use

nodes
r

sers
End End u
use
rs

26
 Network survivability
• Modern optical networks

– The ever-increasing bit rate makes an unrecovered failure a

significant loss for network operators

– Fiber cuts
– No network-operator is willing to accept unprotected networks
anymore

• Restoration: function of rerouting failed connections

• Survivability: property of a network to be resilient to failure

• Physical redundancy and restoration protocols.

27
 Optics in the Internet

Data
SONET
Center
SONET

DWD
M DWD
M

SONET
SONET

Metro Long Haul Metro

Access
Access
 Optical Network: a Layered vision

Layer
3 Layer
IP
2
ATM
1 IP
2/3
MPLS
SONET
0 Thin SONET
Inter-
Packet Packet
IP/MPLS
Opti working Smart
Optics cs Optical Optical 0/1

Multi-physical layers Fewer physical layers

• multi & legacy services • IP service dominance
• robustness, QOS • lower cost

1999 2001 2002

 Protection and Restoration in Internet

• A well defined set of restoration techniques already exists in the

upper electronic layers:
– ATM/MPLS
– IP
– TCP
• Restoration speeds in different layers:
– BGP-4: 15 – 30 minutes
– OSPF: 10 seconds to minutes
– SONET: 50 milliseconds
– Optical Mesh: currently hundred milliseconds to minutes
 Why Optical Layer Protection
• Restoration in the upper layers is slow and require intensive
signaling
– On contrary 50-ms range when automatic protection
schemes are implement in the optical transport layer.
• Purpose of performing restoration in the optical layer:
– To decrease the outage time by exploiting fast rerouting of
the failed connection.
• Main problem in adding protection function in a new layer:
– Instability due to duplication of functions.
– Need the merging of DWDM and electronic transport layer
control and management.
 Why Optical Layer Protection?
• Advantages.
– Speed.
– Efficiency.
• Limitation
– Detection of all faults not possible.(3R).
– Protects traffic in units of light paths.
– Race conditions when optical and client layer both try to
protect against same failure.
 Protection Technique Classification
• Restoration techniques can protect the network against:
– Link failures
• Fiber-cables cuts and line devices failures (amplifers)
– Equipment failures
• OXCs, OADMs, eclectro-optical interface.
• Protection can be implemented
– In the optical channel sublayer (path protection)
– In the optical multiplex sublayer (line protection)
• Different protection techniques are used for
– Ring networks
– Mesh networks
 Shared Mesh Optical Networks
B E

Shared
link
Ingress Egress
A C D C
LSR LSR

F E
Session 1

Session 2

Working Path
• Backup fibers are used for
protection of multiple links
• Assume independent failure and
handle single failure
• The capacity reserved for
protection is greatly reduced Backup Path 34
 1+1 Protection

• Traffic is sent over two parallel paths, and the destination

selects a better one.
• In case of failure, the destination switch onto the other
path.
• Pros: simple for implementation and fast restoration
• Cons: waste of bandwidth
 1:1 Protection

• During normal operation, no traffic or low priority traffic is

sent across the backup path.
• In case failure both the source and destination switch onto
the protection path.
• Pros: better network utilization.
• Cons: required signaling overhead, slower restoration.
 Shared Protection

Normal Operation

1:N Protection
In Case of Failure

• Backup fibers are used for protection of multiple links

• Assume independent failure and handle single failure.
• The capacity reserved for protection is greatly reduced.
 Objectives
Service provider
• Earn more revenue
• Meet SLA requirements
• Focus more on service management than network management
• Maintain reliable connections
• Allocate minimum resources

Customer
• Request via SLA
• Provision QoS-based traffic
• Better picture of current status of network
• High network performance
38
 Definitions (3)
Link availability
𝑀𝑇𝐵𝐹
𝐴! =
𝑀𝑇𝐵𝐹 + 𝑀𝑇𝑇𝑅
Path availability
𝐴" = * 𝐴!
!∈$

Primary/backup path availability

𝐴"&! = * 𝑎 ',!
',! ∈&! "#$%&#'("&)*

𝐴)&! = * 𝑎(',!)
',! ∈&!( -&./0"("&)*)

Joint Path Availability

𝐴$, = 1 − 1 − 𝐴$ 1 − 𝜃𝐴,

39
 Configuring Constraints—
LSP 1 with 40 Mbps
Follows the IGP shortest path to D since sufficient
bandwidth available
:
LSP1 bps 10.0.1/30
40 M 0 .2 .2 .1 .1 10.0.
Router B Router C 2 4 /3 0
3
.16/
192.168.0.1 192.168.2.1
.2
10.0 .1
.1 Router D
.1 10 192.168.24.1
.0
.0 30
/ 30 2/
Router A 0. .1
192.168.16.1 1 0.
.2
.2 .2
Router E
10 192.168.5.1
.0 .1

30
.1

1/
5 /3

3
10

.0.
0 .0.

10
8 /3
0

.2
.1

.2
Router F Router G
192.168.8.1 .1 10.0.13/30 .2 192.168.12.1
 Configuring Constraints—
LSP 2 with 70 Mbps
Insufficient bandwidth available on IGP shortest path

:
LSP1 bps 10.0.1/30
40 M 0 .2 .2 .1 .1 10.0.
Router B Router C 2 4 /3 0
3
.16/
192.168.0.1 192.168.2.1
.2
10.0 .1
.1 Router D
.1 10 192.168.24.1
.0
.0 30
/ 30 2/
Router A 0. .1
192.168.16.1 1 0.
.2
.2 .2
LS Router E
10
.0
70 P2: 192.168.5.1
.1

30
.1 M

1/
5 /3 bp

3
10

.0.
0
s .0.

10
8 /3
0

.2
.1

.2
Router F Router G
192.168.8.1 .1 10.0.13/30 .2 192.168.12.1
 Configuring Constraints—
LSP 3 with 50 Mbps
Exclude all Bronze links

:
LSP1 bps 10.0.1/30
40 M 0 .2 .2 .1 .1 10.0.
Router B Router C 2 4 /3 0
3
10.0
.16/
192.168.0.1
.1 Bronze 192.168.2.1
.2
.1 Router D
.1 10 192.168.24.1
.0
Router A M bps .0
/ 30 2/
30
: 20 ze 0.
0. .1
LSP3 de Bron
192.168.16.1 1
.2
u
Excl
.2 .2

ze
LS Router E

on
10 70 P2: 192.168.5.1

Br
.0 .1

30
.1 M

1/
5 /3 bp

3
10

.0.
0
s .0.

10
8 /3
0

.2
.1

.2
Bronze
Router F Router G
192.168.8.1 .1 10.0.13/30 .2 192.168.12.1
 Preemption
• Defines relative importance of LSPs on same ingress router
• CSPF uses priority to optimize paths
• Higher priority LSPs
– Are established first
– Offer more optimal path selection
– May tear down lower priority LSPs when rerouting

• Default configuration makes all LSPs equal

 Components of a QoS-based Routing Mechanisms
1st Component: Negotiation infrastructure

• Requests required to be modified or revised

• Customers have a chance to choose proper providers

2nd Component: QoS parameters

• Path attributes: Internet

• Path availability
• Holding time
• SLA violation risk

Negotiating
Attributes
ISP1 ISP2
• Provide a better picture of the network status
Pr ing
o
Al visi is ion ms
go on ov ith
rit in Pr lgor
hm g
s A

3rd Component: QoS-based routing algorithms

N
In ego n
fr a ti tio
s t a ti tia ure
ru on e go ruct
ct
ur N ast
r
e
Customer Inf

• Priority-aware algorithms
• SLA-aware algorithms 44
 Dynamic SLA Negotiation Infrastructure
Problem:

• Requests may require to be modified

Goal:

• Give the customer a chance to choose another provider or to comply with the
provider’s network capacity

Benefits:

• Help service providers to control the network resource assignment, and help
customers to negotiate vital and desirable SLA parameters

• Vital and desirable SLA parameters can be negotiated before placing the order
and establishing the requests

• To keep the options open for employing dynamic mechanisms and algorithms
45
 Dynamic SLA Negotiation Infrastructure (Cont’)
Intra-AS negotiation Inter-AS negotiation
• OSPF-TE: Link attributes • EBGP-TE: Path attributes
• Type-11 opaque LSAs • New sub-TLV
• IBGP-TE: Path attributes • Path Attribute field in UPDATE message

es
dat LVs
Up T
TE rray
- a
GP A a te s
IB th P E Upd TLVs
i P-T AS2
ray
W PA(1,j) EBG PA ar
t h
Wi

PA(2,j)

PA(i,j)
j
TL SAs

PA(m,j)
Vs
LA L
th TE
wi PF-

1 AS1
OS

i
m

EB
2 Wi GP-
th T
PA E U
arr pda AS3
ay te
TL s
Vs
s
pdate
-TE U y TLVs
IBGP ra
PA a r
With 46
 Conclusion and Advanced Topics
Conclusion

• To achieve service provider’ goals: 3 components of QoS-based routing mechanism presented

• Dynamic SLA negotiation infrastructure was proposed
• Challenges in priority-aware mechanisms were identified
• Challenges in SLA-based mechanisms were identified
• Novel path attributes as negotiating parameters were introduced and effective cost functions
were presented
• Several types and classes of traffic
• Considerable improvement in performance of proposed mechanisms

Advanced topics

• Energy-aware routing algorithms

• Green networking
47