On the Empirical Evaluation of Multicasting with Minimum Delay Variation

Abstract

In certain types of Internet applications, multicasting with Quality of Service (QoS) optimizations are required on both the end-to-end delay from the source to the destinations, as well as the difference in the end-to-end delay between the destinations. Examples of these applications include video-conferencing, online games, and distributed database replication. These applications require that the messages should reach all the destinations within a specified period of time and all the destinations should receive the message from the source almost at the same time. The solutions to the Delay and Delay Variation Bounded Multicast Network (DVBMN) problem involve generating a network which spans the source and the group of destinations such that the end-to-end delay from source to each destination is within a bound and the delay variation among all the destinations is minimum. The Chains heuristic proposed in the literature provides such a multicast network which achieves the tightest delay variation with minimum execution time. To test the performance of multicasting using the multicast network generated by Chains heuristic on the Internet, we have implemented a prototype of a multicast network using Chains and experimented multicasting with the prototype on PlanetLab. PlanetLab is an overlay testbed which connects more than 1000 nodes all over the world. We have analyzed the performance of multicasting using the prototype on PlanetLab. Results show that the measured value of maximum delay variation from the prototype approximately matches with the value calculated by Chains for a small number of destination nodes. However, as the number of destination nodes increases, the measured value of maximum delay variation from the prototype quickly increases..