Minimizing Routing State for Light-Weight Network Simulation

We introduce a routing mechanism, referred to as algorithmic routing. It is a viable routing alternative for network simulations with minimal space complexity – O(N). In theory and for simulations size of the Internet, algorithmic routing has the potential of reducing memory requirement by several orders of magnitude. In practice and through ns2 simulations on random topologies, we find memory consumption of algorithmic routing exhibits a similar scaling property. However, routes computed by algorithmic routing are not all the shortest. Although we find the relative difference is below 10% for more than 80% of the routes, we are cautious about its applicability to general network simulations. With further discussion on impacts of the distortion, we derive a set of guidelines and recommend users to apply this technique only when suitable.