An Efficient Data Management Scheme Based on Spatial and Temporal Characteristics in Virtual Environments

In a distributed interactive walkthrough system, there are two major bottlenecks which cause performance degradation. One is the server-side workload in the client-server architecture; the other is the network transmission delay. In this paper, we present a knowledge-based data management scheme which takes consideration of both internal (memory) and external (disk) data storage management to ease server-side workload and reduce network transmissions. Our system first analyzes users' logs to discover the spatial and temporal semantic patterns in the virtual environment. Using these patterns, we can determine the proper data layout on disk, and better improve our caching mechanism. Experimental results show good prediction rates and achieve improvements in overall system performance