Distributed shortest-path computation

Abstract

We are proposing a technique to distribute the computational load of online route planners like Google/Bing/Nokia Maps between backend servers and clients. We are presenting an implementation integrated in the ToureNPlaner online routing service and its Android client that increases throughput by more than a factor of 8. Based on Contraction Hierarchies we are decreasing the amount of per request computation on backend systems enabling a leaner and less energy hungry infrastructure to be used that scales with IO instead of CPU power.