Master of Science (MS)
We present extensive simulation and analysis on a traditional, simple, efficient dynamic hot potato routing algorithm on a multi-dimensional torus network. These simulations are performed under a more recent network model than previous, more limited studies, with dynamic (rather than batch) models, no flow-control, and extended high dimensional scenarios. We collect more comprehensive statistics on system performance, and empirically show that the system can recover from worst-case scenarios to quickly re-achieve its standard steady-state delivery rates, with expected delivery time for a packet of O(n), where n is the initial packet distance from its destination. Experiments also show that for our model, the constant multiplier hidden in the O() notation decreases with higher dimensions.
Chi, Raymond Y., "Randomized Greedy Hot-Potato Routing on the Multi-Dimensional Torus" (2009). Master's Projects. 43.