Analysis of Server Provisioning for Distributed Interactive Applications

Increasing geographical spreads of modern distributed interactive applications (DIAs) make distributed server deployment vital for combating network latency and improving the interactivity among participants. In this paper, we investigate the server provisioning problem that concerns where to place servers for DIAs. We formulate the server provisioning problem with an objective of reducing the network latency involved in the interaction between participants. We prove that the problem is NP-hard under several scenarios. We analyze the performance of the classical k-median server placement for DIAs and propose a new greedy server provisioning heuristic for DIAs. Theoretical analysis shows that the approximation ratio of the proposed greedy algorithm is much lower than that of the k-median placement. Experiments using real Internet latency data also show that our proposed algorithm significantly outperforms the k-median and other baseline server placements.