An Enhanced Genetic Algorithm for Server Placement in Distributed Interactive Applications

Recent years have witnessed the enormous popularity of distributed interactive applications (DIAs), which allow participants that are distributed in the network to interact with each other concurrently. The rapid growth of DIAs has raised stringent requirements on providing realistic sense of interaction between participants, whose quality is heavily influenced by network latencies. Although network latencies cannot be eliminated due to geographical spreads of participants, it is possible to reduce them by a smart selection of the locations where the servers of the DIAs are placed. The locations of servers affect not only the inter-server latencies but also the latencies from participants to servers, both of which are involved in the interactions among participants. Thus, the placement of servers is an important factor to the interactivity performance of DIAs. We formulate the server placement problem, and propose to solve it by an enhanced genetic algorithm, whose genetic operators are specially designed based on the nature of the problem. Experimental results using various datasets show that our algorithm leads to appreciable improvement of the interaction quality in DIAs.