Saturday, October 4, 2008

A Performance Comparison of Multi-Hop Wireless Ad Hoc Network Routing Protocols

This paper provides an analysis on the comparison of a variety of multi-hop wireless ad hoc network routing protocols. The ad hoc routing protocols evaluated are:
  1. DSDV (Destination Sequenced Distance Vector) - It is a hop-by-hop distance vector routing protocol that requires each node to periodically broadcast routing updates. This guarantees loop free routing. DSDV maintains a routing table that lists the next hop for each destination. Each route update is tagged with a sequence number and route R is updated only if the update message has a higher sequence number than old update message.
  2. TORA (Temporally-Ordered Routing algorithm) - This is a distributed routing protocol based on link reversal algorithm. It discovers routes on demand, provide multiple routes to a destination, establish routes quickly and minimize the communication overhead by localizing algorithmic reaction to topological changes.
  3. DSR (Dynamic Source Routing) - The header of each packet carries the complete ordered list of nodes through which the packet must pass.
  4. AODV (Ad Hoc On-Demand Distance Vector) - This is a combination of both DSR and DSDV.
These four protocols are compared based on:
  1. Path Optimality - Number of hops a packet took to reach its destination compared to the length of the shortest path in the network
  2. Packet Delivery Ratios - Packet loss rate seen by the protocol.
  3. Routing Overhead - Total number of routing packets transmitted during the simulation
The paper has made a lot of observations in their comparison analysis. It has been observed that both DSDV-SQ and DSR use routes very close to optimal while TORA and AODV-LL have a significant tail, TORA is not designed to find shortest path, DSDV performs well in low node mobility, AODV performs as well as DSR.

Things I would like to take away from the paper is the ns network simulator extension. I agree with Stephen about the point that performance should have been evaluated with several models.

No comments: