Routing Protocol Implementation

1
Routing Protocol Implementation

• Hoon Oh

2
Background

• Many different protocols proposed
• Just few demonstrated in practice
• Reasons
• are too complex
• need the computation of parameters that are not
  readily available in practical systems

3
ABR (Associativity Based Routing)

• Self-organizing
• On-demand
• Source-initiated routing protocol
• Can be implemented using COS hardware and new
  innovative software

4
System Components

• IBM, Compaq laptops running Linux
• Lucent Technologies WaveLAN Wireless PCMCIA
  adapters
• TCP/IP/Ethernet protocol suite on Linux, enhanced
  with the ABR protocol to support ad hoc wireless
  networking

5
Linux

• A variant of UNIX for x86 PC
• Capable of multi-tasking and multi-user operation
  
• Allows many users to have access to the same
  machine, running multiple processes at the same
  time
• Includes
• Most of the required development softwares
  (compilers, libraries, tools, etc.)
• A convenient UNIX and X window system
  environment
• A full suite of TCP/IP networking software
• Open architecture allows easy incorporation of
  the ABR protocol into TCP/IP
• Its widespread use has resulted in easy
  availability of audio, video and network device
  drivers for use in various experiments

6
Laptops

• An Intel Mobile Pentium II processors
• 32 MB of memory
• 2.5 inch 5.1 GB of hard disk drive
• An active 13.3 inch TFT color display with 1024
  by 768 resolution
• A standard I/O interfaces (serial, parallel, USB,
  diskette, keyboard/mouse, docking interface,
  audio I/O, external monitor connector)
• An internal 56K modem
• Two Type I/II PCMCIA card slots

7
Radio Adapters

• 2.4 GHz WaveLAN PCMCIA cards by Lucent
  Technologies
• Implements CSMA/CA media access protocol
• A fast and reliable solution for wireless
  last-hop access to services residing in the wired
  networks
• Only wireless adapters are used to provide
  point-to-point and multihop wireless
  connectivity, without using base stations

8
Data Specification for WaveLAN PCMCIA Adapter

• Data Communications Performance
• Data rate 2 Mbps
• Media Access Ethernet CSMA/CA
• Bit Error Rate Better than 10-8

9
RF Specification of WaveLAN PCMCIA adapter
DQPSK Differential Quadrature Phase Shift Keying
10
ABR Protocol Implementation on Linux

• Software Layering Architecture
• Implementing ABR Packet Headers and Beaconing
• Implementing ABR Outflow and Inflow
• Implementing ABR Routing Functions

11
Software Layering Architecture

• ABR is a sublayer between the IP and Ethernet MAC
  layers
• Packets are able to bypass the ABR module
• Ability to handle both non-ABR and ABR traffic,
  both wired and wireless

12
The Software Architecture of TCP/UDP/IP/ABR
protocol Implementation
13
The Principles of Packet Relaying in ABR Protocol
14
Implementing ABR Packet Headers

• ABR packet headers can be divided into two
  components
• The base header (the common header)
• The type-specific header
• Every packet has the same base header but a
  different type-specific header
• The types are BQ, LQ, RD or RN

15
ABR Base Header Format
16
Implementing ABR Packet Beaconing

• Supporting the associativity ticks within the ABR
  sublayer
• Processed separately from the regular ABR traffic
  flow of control and data packets
• Each node will transmit a beacon periodically
• Whenever a node receives a beacon, the
  associativity tick counter corresponding to that
  neighbor is incremented

17
Implementing ABR Outflow

• Outflow source sends data to a receiver
• ABR needs to trap output from the IP to the
  device driver to process packets
• ABR software checks to see if destination is an
  ABR host
• If it isnt, packet proceeds as normal
• If it is, software redoes the MAC header and
  inserts an ABR header between the Ethernet MAC
  and IP headers

18
Implementing ABR Inflow

• Inflow packet arrives at an ABR host
• Each packet is registered with an ABR packet type
  
• When these packets arrive, they are passed up the
  protocol stack
• ABR control packets are processed in the ABR
  sublayer
• Data packets have the ABR header stripped and
  passed up to the IP layer for further processing
• ABR sublayer appears transparent to IP and higher
  layers for data packets
• Enables ABR to support TCP/IPbased applications

19
The ABR Route Discovery Protocol State Machine
20
The ABR Route Reconstruction Protocol State
Machine
21
The ABR Route Delete Protocol State Machine
22
Experimentation and Protocol Performance

• Experimental setup of an ad hoc mobile network in
  office
• Control Packet Overhead
• Route Discovery Time
• End-to-End Delay
• Data Throughput
• Effects of Beaconing on Battery Life

23
Experimental setup of an ad hoc mobile network
in office
24
Control Packet Overhead

• An IP header is 20 bytes (assuming no IP options)
  
• An ABR data packet is 24 bytes
• Ethernet maximum transfer unit (MTU), minus MAC
  header, is 1500 bytes
• The packet overhead of ABR is (2420)/1500 2.9
  
• The packet overhead of IP is 20/1500 1.3
• ABR basically replaces IP to support ad hoc
  routing gt the IP header is not necessary
• IP header is replaced with ABR gt total overhead
  is 24/15001.6
• IP is retained so that ad hoc routing is
  transparent to existing TCP/UDP/IP based
  applications

25
Route Discovery Time

• Three different test scenarios in an indoor
  environment
• Hop count ranged from one to three
• The Ping application was used to initiate sending
  data from one ABR host to another
• From the time when a BQ packet is sent by the SRC
  until the time when a route is added to the
  routing table
• In each case, the route discovery process is
  performed 100 times
• Route discovery time accounts for the entire
  round-trip time (RTT) from SRC to DEST

26
Route Discovery Time
27
End-to-End Delay

• Data sent 100 times, RTTs are recorded
• The ping packet size is gradually increased,
  default is 64 bytes
• For every ping packet received at the DEST, a
  ping status packet was returned to the SRC,
  revealing the RTT
• The end-to-end delay is one-half the RTT

28
Data Throughput

• The average throughput is calculated by dividing
  the packet size by the average end-to-end delay
• The actual throughput is calculated by dividing
  the total number of bytes received by the total
  end-to-end delay

29
Effects of Beaconing on Battery Life

• Periodic beaconing of mobile hosts is used to
  gather associativity information
• Laptops used had several power management
  features
• Standby mode
• Suspend mode
• Hibernation mode
• The beaconing interval on the mobile hosts was
  varied and their battery life was noted at
  regular periods
• Increasing the beaconing frequency did not result
  in the significant reduction of battery life

30
Conclusions (1)

• Route discovery occurs at the ABR sublayer
• Data packets originate up at the application
  layer
• A few milliseconds of processing time as the
  packets traverse the protocol stack
• Route discovery is order of a few milliseconds,
  shorter than the data round trip time
• The throughput performance for multi-hop routes
  is satisfactory can support many existing
  applications

31
Conclusions (2)

• The RRC times indicate partial RRC can be
  established quickly
• ABR is scalable
• The route discovery time for one hop is about 6.5
  ms gt discovering a 10-hop route takes less than
  a second
• Increasing the beaconing interval does not
  significantly change the battery life when power
  management is turned on
• Implementing ABR to support multi-hop networking
  is both feasible and practical

32
Summary

• Implementation is straightforward with
• COS hardware
• communication software
• Only minor modifications has to be made to
  TCP/UDP/IP/Ethernet stack
• to accommodate the ABR sublayer to support ad
  hoc wireless networking
• Software architecture is capable of supporting
  existing UDP/TCP/IP based applications
• The real challenges are
• to decide where the ABR sublayer should be
• how to interface it with the existing IP and
  Ethernet layers

33
Questions(1)

• Toh ?? ?7??? ??? SPEC? ????, 3-hop ?? ??? 100 MB
  ??? ???? ??? ?? ?? ????
• ?? SPEC? ???? ????
• ??
• NOTE ???? SRC?? DEST? ???? ??.

34
Questions(2)

• ????? ??? ??? ?? ??? ??? ? ? ?? ??? ????.