Performance Evaluation of Using a Dynamic Shortest Path Algorithm in OLSRv2

MANET routing protocols are designed to scale up to thousands of routers with frequent changes of the topology. In preference, MANET routing protocols should also support constrained low-power devices. One of the bottlenecks of scalability in link-state routing protocols is the performance of the shortest path algorithm (e.g. Dijkstra). In this paper, we investigate the in-node performance of OLSRv2 and, in particular, study the benefits of using a dynamic shortest path (DSP) algorithm for this routing protocol. A DSP algorithm is an algorithm that adds or removes edges in the routing tree incrementally and calculates shortest paths, also incrementally. The performance in OLSRv2 with classic Dijkstra vs. DSP is evaluated, by comparing the CPU time for calculating paths in a large emulated network. Additionally, it is demonstrated that frequent topology changes due to mobility in MANETs lead to frequent routing table recalculations with only few routes updated each time. This property of MANETs makes the use of a DSP in OLSRv2 appropriate.