Queueing Behavior of Hashing Methods Used to Build URL Routing Tables

This study pioneers a new way to study the performance of hashing algorithms by evaluating their queueing behavior. We examine hashing as applied to URL routing used in Content Distribution Networks (CDNs). URL routing tables can be built as self-adjusting hash tables to reduce network bandwidth consumption in CDNs and at the same time take advantage of the high temporal locality of Web document requests. We show through a trace-driven simulation that the existing Aggressive hashing (based on move-to-front rule) improves upon H1 hashing (based on the transposition rule). Surprising results were found: even though the expected improvement in look-up time over H1 hashing is modest, the improvement from a queueing perspective appears to be very significant. The long range dependence in look-up time of H1 hashing contributes to its extreme queueing delay. This shows that using queueing delay as a means of evaluating hashing algorithms can be very insightful. Fig. 1 URL router forwarding or redirecting a request