A performance comparison of multi-hop wireless ad hoc network routing protocols

As its name suggests, this paper compares 4 routing protocols for in an ad hoc network scenario: DSDV (hop-by-hop routing, periodic broadcast updates), TORA (route discovery on demand), DSR (source routing, discovery on demand) and AODV (hop-by-hop routing, discovery on demand). The evaluations are performed on a network simulator, ns, using an updated physical and data link layer model.

The experimental methodology studies routing performance with varying node mobility, packet rates, and number of transmitting sources. Nodes use a "random waypoint" movement model, with a pause time (when the node is stationary) before moving to the next waypoint, at maximum speeds of 20 m/s (~45 mph) or 1 m/s (~2.25 mph). Transmitting sources are assumed to be constant bitrate sources. Its not clear what real-life scenarios this simulation is reflective of (perhaps vehicular adhoc networks?) at speeds of 20 m/s. If the goal is to evaluate the performance on topology changes at various speeds, then maybe limiting transmission to when sources/sinks are stationary would be more realistic.

The main results are that:

• Packet delivery ratio is high (0.9-1) and relatively stable for TORA, DSR and AODV over a range of pause times (i.e. topology change rate), but is lower for DSDV at pause times less than 300s. However, at low speeds (1 m/s), all protocols have fairly similar performance.
  
• TORA performance in packet delivery ratio degrades when the number of sources is high (i.e. 30)
• Routing overhead is constant with respect to mobility for DSDV (as expected). Routing overhead scales with mobility for the other routing protocols, with TORA > AODV > DSR
• Routing overhead is constant with respect to number of transmitting sources for DSDV (again as expected). It scales with number of transmitting sources for the other routing protocols. At 30 sources, TORA >> AODV >> DSR, with an order of magnitude of difference between each of them.
• In terms of path optimality, both DSDV and DSR do very well, while TORA and AODV having significant number of packets routed through paths with 3/4 extra hops than the shortest path.

The main message is that there are no clear winners, and which protocol to use may well depend on what the operating conditions are.