Performance Comparison of Multihop Wireless Ad-Hoc Routing Protocols

This paper compares several different routing protocols (more accurately, it compares optimized versions of several different routing protocols) in a simulated wireless environment. It is interesting in that it is a first study of the relative performance of these protocols; I don't have too much criticism except that I'm not necessarily convinved that the protocols were "equally" optimized by the group. That is, the one that seemed most optimized is the protocol that performed best overall--- DSR.

Still, this is at least partially due to the use of source routing which seems to be a better fit than per-hop routing that some of the other algorithms use. Combined with eavesdropping, which allows state updates with less overhead, this gives each node an idea about the accessibility and route to each other node; with DSDV etc, the route information at each node is still about routes to all other nodes, but only keeps the next hop information.

The eavesdropping and caching employed by DSR are the reasons it performs best; both of these are optimizations to the original DSR algorithm.

In addition, I thought it interesting that TORA's short-lived routing loops created during route reversal lead to many packet drops; in hindsight, this seems like a poor decision in a routing algorithm (creating loops). Finally, the last interesting thing in this paper is that large packets are undesirable for the scenario used because of unfairness in MAC protocols; thus wireless routing should use smaller packets than those used in wired networks, where large packets can be desirable for amortizing the bandwidth-latency product.