An Active Self-Optimizing Multiplayer Gaming Architecture

Multiplayer games are representative of a large class of distributed applications that suffer from redundant communication, bottlenecks, single points of failure and poor reactivity to changing network conditions. Many of these problems can be alleviated through simple network adaptations at the infrastructure level. In this paper, we describe a model in which game packets are directed along the edges of a rooted tree connecting the players, aggregated during the upstream flight and multicast from the root to the leaves. This tree is constructed based on a heuristic, and can dynamically adjust itself in response to changes in network conditions. This gaming infrastructure is built and maintained using active networks, which is currently the only open architecture suitable for these types of applications. We have designed and implemented a prototype using ANTS that performs these adaptations for unmodified DOOM clients. We present analytical and simulation results that illustrate the reduction in communication overhead, and show that the multicast tree can quickly adjust to changing network conditions. The overhead of the active network-based middleware is acceptable, especially in wide-area networks.