One Decoding Step

1
Split-n-Save Path Multiplexing in Wireless Ad
Hoc Routing

Meeyoung Cha and DK Lee Advisor Sue
Moon (Korea Advanced Institute of Science and
Technology)
IEEE INFOCOM 2005 Student Workshop
2
Motivation of Our Work

• On-demand routing is favored in resource
  constrained environments of MANET. (e.g., AODV,
  DSR)
• Recent routing schemes use multi-path for fast
  path fail-over. (e.g., AOMDV, MDSR, NDMR, MP-DSR,
  SMR)
• Do existing multipath routings perform good in
  terms of the number of active nodes over a time
  period or workload balancing throughout the
  network?

3
Multipath Routing Basics
Following figure shows a typical multipath
routing.
2. failure in default path
X
1. using default path
3. fast transition to one of the candidate
multipaths
Destination
Origin
4. using new path
Candidate paths are used only after the main
path is no longer available. Lets use multiple
paths more actively!
4
Multipath Routing Basics
Description of a typical multipath routing
1. using default path
Destination
2. failure in default path
default path
3. fast transition to one of the candidate
multipaths
Origin
candidate paths
4. using new path
Intuition Disjoint overlay path gives maximum
robustness against single link or router failures!
5
Split-n-Save

• We propose Split-n-Save
• Added feature to an existing multipath routing
  protocol
• patch to AOMDV ns-2 code
• Use path multiplexing
• split-n-save changes paths per every k packets
• we call k, the frequency of multiplexing.
• Expected benefits of Split-n-Save
• Inherit all benefits from the underlying routing
  protocol
• Improved routing performance
• number of active nodes over a time period
• well-balanced workload throughout the network

6
Routing Performance Metrics

• To measure the efficacy of Split-n-Save we use
  following traditional metrics.
• Packet delivery ratio
• Average end-to-end delay
• Routing overhead
• We also use two other metrics.
• Node satisfiability
• ratio of the number of forwarding packets
  generated by itself and that by the other nodes
• Network survivability
• number of active nodes over a period of time

7
Simulation Settings

• CMU wireless extensions to ns-2

8
Simulation
There exists a specific k value that performs
best for each routing performance metric.
9
Simulation End-to-end Delay
surging points
We find several surging points in end-to-end
delay when k 0 and 1.
10
Simulation Node Satisfiability
node satisfiability
of packets sent by itself of packet forwarded

Node satisfiability improves when k gt 0.
11
Preliminary Results

• For network survivability, temporal info alone
  was not enough. We plan to use spatial info as
  well.
• - track down the location and causality of node
  failure
• For other metrics, path multiplexing clearly
  improved the routing performance with carefully
  chosen freq of multiplexing.

12
Conclusions

• We propose and evaluate Split-n-Save, which
  exploits path multiplexing in multipath routing.
• We propose two other metrics
• node satisfiability and network survivability
• Frequency of multiplexing can be used to cope
  with path fail prediction and path optimality.
• Path multiplexing should reflect network
  dynamics.

13
Future Works

• Methodology
• Probabilistic model for Split-n-Save
• Other multiplexing idea
• Revisit routing performance metrics for MANET
• Suggestions on simulation in MANET
• Good guideline for simulation setups
• More practical movement pattern model
• Grid movement, communication on the road, campus
  network