Accelerated processing of historical BGP events for testing new BGP heuristics

This paper describes a technique for artificially accelerating 'real time' when testing new BGP protocol enhancements using historical real-world data. We show how months of BGP advertisement data may be processed in hours, yet generate outputs that appear to reflect months of actual operation by a network of fully featured BGP speakers. Using Quagga (an operational open-source implementation of BGP) we characterise the performance trade-offs of our technique, and show how 'accelerated time' benefits researchers who are exploring modifications to BGP's dynamic, timer-based behaviours. We consider the impact of our technique when multiple instances of Quagga run on a single host or are distributed across multiple hosts.