1

6LoWPAN

Based On-
IP is Dead, Long Live IP for Wireless Sensor Networks
-Jonathan W. Hui -David E. Culler
 2

Presented By
- Milan Jain MT12067
- Anil Sharma MT12063
Outline
 Introduction
 Network Design And Software Abstraction
 Link Layer
 Adaptation Layer
 Network Layer
 Evaluation
 Application
 Can We Implement this in Flyport?
 References

3
Introduction
Internet of Things (IoT) and Wireless Sensor Networks
How This Started?

 Core Internet - Backbone

routers and servers.

 Fringe Internet – Laptops, PC

and local network
infrastructure.

 Internet of Things – Low

power, low-bandwidth
wireless embedded devices.

5
Internet Of Things(IoT)

 Uniquely identifiable things and their virtual representations in an

Internet like structure.

 “The Internet Protocol could and should be applied even to the smallest
devices”

 Low-Power devices with limited processing capabilities should be able to

participate in the Internet of Things.

 Idea lead us to 6LoWPAN – “IPv6 over Low Power Wireless Personal Area
Networks”

6
7
 How 6LoWPAN different from TCP/IP?

 Transport Layer
 TCP not that useful.
 Application Layer
 HTTP not much of use.

Will see in Evaluation

802.15.4 at Physical and Link Layer – Duty Cycled Operations

Adaptation Layer – Fragmentation and Header Compression
Network Layer
 IPv6
 Layer two mechanisms – Bootstraping, Discovering and Autoconfiguration 8
Primary contribution (Paper)

 Complete IPv6-based network architecture for WSNs that allows end-to-

end communication between nodes and any IP device at the network
layer.

 Software Architecture that preserves IP’s layered protocol model, defining

the services, interfaces and their interaction that can incorporate many of
the techniques developed within the WSN community.

 Implementation of a complex, efficient and production quality IPv6

solution for WSNs and show that general network architecture can
outperform existing systems that do not adhere to any particular standard
or architecture.
9
Introduction
 General argument by the Researchers in WSN research
“while many of lessons learned from Internet and mobile network
design will be applicable to designing wireless sensor applications …
sensor networks have different enough requirements to warrant
reconsidering the overall structure of applications and services”.
 Reasons IP is Dead for Wireless Sensor Networks
 Resource constraints may cause us to give up the layered architecture.
 Sheer number of these devices, and their unattended
deployment.(Unreliable)
 Limited task of sensing.
 Traditional interfaces and layers of system abstractions should not be
assumed.

10
Network Design And Software
Abstraction
Network Design and Software Abstraction for IP
IPv6 Based Network Design
 Each WSN node serves as an IP
router.
 As border routers forward
datagrams at network layer,
they don’t maintain any
application layer state.
 Peers communicate in terms of
the capabilities provided by the
layer below.
 Link must allow network to
achieve high “best-effort”
datagram delivery.
 To achieve reliable transport.

12
6LoWPAN with Traditional IP (Edge Router)

13
Software Abstraction

 Each node implements full

network stack.
 It is to respect IP’s layered
model while using the
proper mechanism to
support efficient
communication in WSNs.

14
IP Link
 IP Link - Nodes that are reachable over single hop.
 Direct Connection at Physical Layer.
 Emulated over different physical communication domains.
 Properties:
 Always-On
 Best-Effort Reliability
 Single Broadcast Domain
 Problem:
 Places too much policy in Link Layer.
 Network Layer unaware about complex link-layer dynamics.

15
Proposed Model - Avoiding IP Link Emulation

 Now, IP Link - Neighbors reachable

within a single radio transmission.
 WSNs overlapping Link-local Scopes
 This violates Single Broadcast
Domain Property.
 But provides better visibility into
connectivity
 Better Discovery and communication.
 Better control of forwarding policies.

16
 17

Link layer
6LoWPAN
Introduction
 Low power radios consume as much power when receiving or just even
listening when compared to transmitting.
 Idle-listening completely dominates system energy consumption when
radio is not duty cycled.
 Industry has not yet come to agreement on a link protocol for duty cycled
operation in Multihop Network.
 Developed duty cycle link protocol, while keeping in mind use of IPv6
network layer above (Interoperability).
 Duty-Cycled Radio – Transmitter can send packets to receiver at specific
times.

18
Goal to Design Duty-Cycled Links
 Consume minimal power.
 Provides following IP-friendly properties:
 Always-On: Nodes able to communicate without establishing connection or
requiring any existing state.
 Low Latency: Transmission delays to any neighboring node should be low.
 Broadcast Capable: Able to broadcast frames to all neighboring nodes,
regardless of node density.
 Synchronous Acks: Link should allow IP to achieve high “best-effort”
datagram delivery.

19
How They Achieved This Goal? - MMC
 MMC – Arbitrate access to media between simultaneous
transmitters.
 Improve on WiseMAC by embedding addressing and timing
Information into wakeup signal.
 Techniques Used:
 Sampled Listening
 Synchronous Acks
 Scheduling
 Streaming

20
Sampled listening
 Requires two primitives:
 Wakeup signal when transmitting
 Channel Sampling
 Chirp frame – 802.15.4 compliant frame and contains
 Destination Address
 Rendezvous Time: Time remaining until actual data frame transmission.
 Overhearing cost is reduced
 Rendezvous time allows destination to power down for that time.
 Reduces receive cost to that of receiving a chirp frame and the data
frame.

21
Synchronous Acks
 Ack frame defined in 802.15.4 insufficient:
 No addressing information.
 Not Secured.
 Cannot carry payload – Useful for hop-by-hop feedback
 New Ack Frame Defined as 802.15.4 data frame:
 Addressing and Security mechanisms can be used.
 Also carry payload which can be utilized for:
 Scheduling Optimizations
 Network Layer Optimizations

22
Scheduling
 Sample period and phase in payload of each ack.

 How it is helpful?
 Node can synchronize to any neighbor after single acked transmission.
 If destination’s schedule is known, chirp duration can be reduced to small
synchronization guard time.

 Central Manager can be used: To collect and modify schedules of

individual nodes.

23
Streaming
 To increase throughput and Energy Efficiency.
 Transmitter can signal that another data frame will immediately
follow.
 Node can send data frames back-to-back without delay after sending
a single chirp frame signal.
 This allows both sender and receiver to amortize wake-up costs
across multiple frames.

24
Link Software Abstraction
 Link layer maintains neighbor table that holds link specific state
about neighbors:
 Link Addresses
 Schedules
 Frame Pending Indicator
 Link Quality Statistics – Helpful in selecting routes
 RSSI
 Link Success Rates
 Also provides feedback on each transmission and reception.
 Indicates whether transmission attempt was Acked.
 RSSI of Ack.
 Also RSSI for each received frames.

25
ADAPTATION LAYER
Fragmentation & Header Compression
Why And How?

Now Who will Fragment it

IPv6 routers unlike IPv4 never

fragments IPv6 packets.

Base header of 40 bytes and

minimum MTU requirements of
1280 bytes
802.15.4 supports 127 bytes of
payload and 80 bytes in worst
case. TCP/IP Stack
27
Answer - Adaptation Layer

• Header compression to support

additional communication
paradigms.

• Introduced header compression

technique reduces 48 byte
UDP/IPv6 header down to 6 bytes
in best case.
ADAPTATION LAYER

6LoWPAN Stack
28
Header Compression – RFC 4944 HC
 Stateless – provide nodes flexibility to communicate with any neighbor in
compact form.
 Two ways to compress header
 Making assumptions about common values for IPv6 header fields in WSNs.
 By removing redundant information about layers.
 Payload length and Interface Identifiers(IID) are derived from the link header.
 Doesn’t efficiently compress headers when communicating outside of
link-local scope or when using multicast.
 When prefix is elided it is assumed to be CGP(Common Global Prefix) or
Link-Local Prefix.

29
IPv6 Header Compression
VTF Next Hop Limit Source Address Destination Address
Header

6LP_IPHC

IPv6 Header

30
Packet Format (After Compression)
VTF Next Hop Limit Source Address Destination Address
Header
 VTF (Version, Traffic Class, Flow Label) – 1 Bit to indicate whether elided
or not.
 Next Header – 1 Bit to indicate Next Header is elided and 6LP_NHC is
used.
 Hop Limit – 2 Bits to indicate whether 1, 64 or 255 or carried inline.
 Source and Destination Address (IPv6 Addresses) : 2 Bits each
 Full 128-bit carried inline.
 64-bit IID carried inline.
 Bottom 16-bits of IID carried inline.
 Fully elided.

31
Compression Efficiency
 When prefix is elided it is assumed to be CGP(Common Global Prefix)
or Link-Local Prefix.
 Possible compressions in various cases are:
 Link-Local Unicast: 6 Bytes
 Link-Local Multicast: 8 Bytes
 Global Unicast: 11 Bytes
 Multicast within WSN: 11 Bytes
 Arbitrary IP Devices outside WSN: 25 Bytes

32
 33

Network Layer
Forwarding And Routing
Why IPv6 for WSNs?
 Much larger address space (128 Bits address).

 Autoconf and ICMPv6 – Scalability, Visibility and Unattended Operations

 Inclusion of various two layer protocols e.g., ARP and DHCP into IPv6
framework.

34
Network Layer - Introduction
 Network Layer includes three services:
 Configuration and Management
 Forwarding
 Routing

 ICMPv6 is used to configure and manage a large number of nodes in WSN.

 Neighbor Discovery
 Auto-configuration

35
Neighbor Discovery
 To detect each other’s presence and to find routers

 Routers periodically multicast RA(router advertisement) to announce their

presence (operate over a single IP link, not multiple)

 Here RA is propagated to multiple hops with some network parameters

 Trickle algorithm is used to maintain freshness of information

36
Auto-configuration
 IPv6 support both stateless and DHCPv6 autoconf methods to assign
unique address

 Stateless:
 Disseminates parameters to all nodes (like using RA message)
 Low cost (single RA message)
 Global parameters ?
 Address generated by concatenating prefix with IID

37
Auto-Configuration – DHCPv6

 DHCPv6:
 It selectively assign parameters to individual nodes (means behave like a
central server)
 Nodes request to central server for the unique address
 Every WSN router act as a relay agent

38
Forwarding - Introduction
 Separate from routing.

 Responsibilities
 Receive datagrams from an interface.
 Next hop lookups in a forwarding table.
 Submit message to appropriate interface.

 Primary Goal for forwarder design:

 Energy Efficient
 High end-to-end success rates.

39
Unicast forwarder
 Applies three orthogonal mechanisms:
 Hop-By-Hop Recovery
 Ensures that datagram will reach to destination
 Link layer ack indicates whether or not network layer was able to accept the
message.
 So no need to rely on broadcast or continuous snooping.
 Streaming
 When submitting datagrams to the link, forwarder indicates whether other
packets for the same next-hop destination follow.
 If first transmission succeeds, it is likely that remaining transmissions will
succeed.
 Reduces average transmission cost.
 Congestion Control
 Congestion can cause queues to become full and decrease energy efficiency
due to forwarding failures.
 Uses feedback to adjust transmission rates using an additive-increase,
multiplicative decrease control.

40
Quality of Service
 Previous three things may induce higher latency due to forwarding delays
 QoS allows upper layers to select forwarding policy.
 Three mechanisms:
 Upper layer tag datagrams as latency tolerant:
 Upper layer tag datagrams as high priority
 Queue reservation for different traffic classes.
 Information about datagram is placed in an IPv6 Hop-by-Hop Option
header.

41
Multicast Forwarder
 Simple controlled flood using Trickle.

 Trickle’s sequence number is included in an IPv6 Hop-by-Hop Option

Header.

 Nodes buffer single datagram for continuous retransmission until

maximum transmission period is reached.

42
Routing - Introduction

 Why Routing ?
 To establish reachability, forming paths, minimize routing metric

 Nodes are having limited resources like memory, processing power

etc.

 Routing protocol for WSN

 RSSI or (Link Quality information to detect PRR)
 These measurements have high variance
 Send control messages to compute PRR
 Require more time, energy.

43
Contd..

 Because of limited resources

 nodes have next-hop information
for a limited set of destination
 default route for all others

 Border routers maintain host-

routes to every node in WSN

 Routing responsibility on border

router reduces the control
message overhead
44
 Default routes

 Four main tasks

a) Discovering Routes
 maintains a routing table (separate from forwarding table)
 includes
 Sender’s distance in number of hops
 Estimated transmissions (ETX)

b) Managing the Routing Table

 Router inserts potential route into the routing table
 Selection by sorting the path cost and the link quality estimate

45
 Contd..
c) Selecting Default Routes
 Top entry in the routing table is selected as the default route
 Router randomly selects a small fraction of forwarded datagram for link quality
estimate (uses hop count information)
 No need to generate explicit control packets

d) Maintaining Route consistency

 Can create loops or suboptimal path
 Uses IPv6 hop-by-hop option
 in case of inefficient routes RA Trickle timer is restarted

46
Host Routes

 Border router maintain host route to every node in WSN

 Border router routes the datagram to WSN node by inserting IPv6 routing
header which contains path to the destination
 WSN node provide the default route information by including a RRO
(Record Route Option) in IPv6 header.

 Inside RRO
 contains a list of addresses, identifying the hosts that have forwarded the datagram.
 RRO requires 2 byte per entry
 Because all have same prefix

47
Routing overhead
 Routing protocol configures both default and host routes

 Any communication overhead ?

 Broadcast from routers
 Unicast from leaf nodes to border router

 Can we reduce this overhead ?

 by piggybacking on already existing traffic

48
 49

Evaluation
Results Presented in the Paper
Hardware Description
 All the discussed techniques were implemented on TinyOS 2.x and
TelosB platform

 TelosB
 16-bit TI MSP430 MCU
 48KB ROM, 10KB RAM
 2.4 GHz, 250 kbps TI CC2420 IEEE 802.15.4 radio
 AES-128 authentication

 Evaluation of the Power Analysis of the IPv6 based network

architecture

50
Experimental Setup
 The power model is validated in a real world Home Monitoring
Application
 Data collection
 Temperature, Humidity
 Network Statistics
 Routing topology
 7 nodes within 1-hop of border router
 Other within 2 or 3.
 Continued for 4 weeks
 Computed Average duty Cycle and success rate for each node

51
Evaluation Metrics

 Link Energy Cost

 Total power consumed by the node

 Network Energy Cost

 Average Power Draw in maintaining the connectivity of the network

 Application Energy Cost

 Average Power Draw of a node in an application environment

52
Link Energy Cost
 Total power was modeled using listen, receive, transmit costs

 How it was measured ?

 Values of interest …

Any effect of
chirp length ?

53
Current Signature

54
Understanding the Graph

 Transmission with chirp is more costly

 Why Transmit without chirp are more costly than Receive with chirp ?

55
Take Aways
 Chirp length effects transmission

 Increasing sample rate reduces power draw when node is only listening

 Transmit with chirp is more costly than without chirp

 Transmit without chirp is more costly than Receive with chirp

56
Network Energy Cost
 Average power draw for a node maintaining network connectivity

 What is used to maintain connectivity of nodes ?

 RA (Router Advertisement) – Broadcast
 RRO (Record Route Option) – unicast

 So what is the cost of transmission and reception ?

N = Number of Neighboring
Routers
D = Number of Descendants
Frr (RRO unicast)
Fra (RA broadcast)

57
Understanding the Graph

 In Case (a) - Transmit cost is more ?

 Because RA broadcast scales with the sample period (D = 5, N = 5).
 But not in case (b) (D = 0, N = 5).
 Because no broadcast by host.

58
Take Away – Network Energy Cost
 Transmit is more costly than other operations

 Router nodes draw more power than Hosts

 Due to broadcast of RA information

 Power draw increase when number of Descendants were increased

 Independent to number of neighbor nodes

59
Application Energy cost

 Average Power draw of a node in an application environment

 Both host-only and router nodes source UDP datagram to a data server
through border router

 What was the overall cost ?

N = Number of datagrams by a
node
D = Number of nodes to forward

60
Understanding the Graph

 Why transmit power for Router Node is more than Host Node ?

61
How Proposed System Is Better?

 Results were compared with prior deployments

 RP – Report Period
 DC – Duty Cycle
 Latency – latency of the radio
 DRR – Data Reception Rate
62
Goodput and latency
 Goodput - Application Level Throughput,
 Number of information bits, delivered by the network to a certain destination,
per unit of time
 Lower than UDP
 Because TCP requires communication in both directions

63
 64

Application
LessTricity – 6LoWPAN based Building Energy Saving and Management
System
Lesstricity – Building Energy Savings and Management
LessTricity system developed by consortium of companies in UK.

AIM: Increase efficiency of energy

usage and better management in
commercial buildings and
businesses.

Uses wireless control and

management technology based on
6LoWPAN to enable management
of large facilities and even remote
buildings.
Image Source:
http://www.iebmedia.com/index.php?id=7177&parentid=63&themeid=255&hft
65
=59&showdetail=true&bb=1&PHPSESSID=p6us7l21pgpr2e7sven8ig3e13
Network architecture
Small deployment consists of
clusters of about 50 LessTricity
power controllers(LPC).
Each LPC transmits measured power
through a 6LoWPAN network to
LessTricity Network Interface(LNI-
Edge Router).
Readings are sent to a central
database.
A graphical web interface to monitor
and manage energy consumption in
the LPCs and to manage network
and LPC themselves.
66
Benefits from 6lowpan

Easy Deployment.
Compatible with building’s existing
IT Infrastructure.
Energy Usage information can be
made available locally or over
Internet for remote monitoring.
Ex: Corporate Energy Usage
Monitoring.
End User Focus: Reduce energy
footprint of their operations, both
from cost saving and environment
responsibility point of view.
67
Can We Implement 6LoWPAN?
6LoWPAN
 Can we implement 6lowpan?

IPv4
Network

Edge Router
Sensor Act

Wireless Sensors having 6LoWPAN

Stack 69
Reference

Book Based on 6LoWPAN

– Zach Scelby
- Carsten Bormann

70
Thank You!!

71
QUESTIONS?
FEEDBACK/COMMENTS/SUGGESTIONS
6LoWPAN