Managing Bandwidth and Traffic via Bundling and Filtration in Large- Scale Distributed Simulations

摘要

Research has shown that bandwidth can be a limiting factor in the performance of distributed simulations. The Air Force's Distributed Mission Operations Center (DMOC) periodically hosts one of the largest distributed simulation events in the world. The engineers at the DMOC have dealt with the difficult problem of limited bandwidth by implementing application level filters that process all DIS PDUs between the various networks connected to the exercise. This thesis examines their implemented filter and proposes: adaptive range-based filtering and bundling together of PDUs. The goals are to reduce the number of PDUs passed by the adaptive filter and to reduce network overhead and the total amount of data transferred by maximizing packet size up to the MTU. The proposed changes were implemented and logged data from previous events were used on a test network in order to measure the improvement from the base filter to the improved filter. The results showed that the adaptive range based filter was effective, though minimally so, and that the PDU bundling resulted in a reduction of 17% to 20% of the total traffic transmitted across the network.