From the many challenges in allocating resources for massively multiplayer online games (MMOGs), Jordan et al. present a novel load management approach to dynamically adjust resources to current demand. The approach includes two layers, global and local, to reflect user demand. The global level uses a peer-to-peer infrastructure to assign data centers to serve specific game regions, and the local level operates within a facility to maintain server instances. Two bin-packing heuristics maximize utilization of resources while providing quality of service (QoS) for the user.
After establishing a foundation for this problem domain, an MMOG-specific architecture is presented, including an effective two-part figure to illustrate the big picture. The following section provides the abstract bin model for dynamic load management, with particular detail for the local and global balancers, and gives the two specific heuristic-based algorithms. The fourth section presents detailed experimental results based on various simulation scenarios and includes multiple figures and tables to document the results. The next section gives a brief summary of related work by others.
The paper is well written and well organized, with an effective presentation that incorporates both zoning and replication to achieve workload balance and QoS simultaneously. This paper is recommended for readers with an interest in MMOG performance specifically, or distributed workload performance generally.