Evolving pub/sub subscriptions for multiplayer online games: demo

Abstract

We identify a class of content-based pub/sub applications with highly dynamic subscriptions. This includes location-based notification systems, predictive stock trading, and multiplayer games. The traditional method of handling subscription changes by engaging in expensive re-subscription protocols is inadequate when the workload is sufficiently large. We propose evolving subscriptions as a technique to alleviate the overhead of subscription churn when handling regular patterns of subscription changes. Subscribers express as a function of time the evolution of its subscriptions. The pub/sub brokers can then independently modify the evolving subscriptions without requiring further communication with the subscribers. In this demo, we present our design and implementation for supporting evolving subscriptions in the context of a multiplayer online game. We first show how evolving subscriptions are expressed to the pub/sub system. We then demonstrate its usage with a game client, which displays real-time performance measurements between the regular system and our enhanced version. We show the relative impact of using evolving subscriptions by running different experiments with our game engine.