As the scale of distributed virtual environment systems grows, scalability becomes a key issue. One of approaches for enhancing scalability is to adopt multiple server architecture. However, a non-uniform and highly skewed distribution of users over the VE would incur a significant workload imbalance among servers; consequently, users managed by heavily-loaded servers would experience low interactive performance due to long latencies of state updates at servers. Dynamic load distribution among servers can prevent the interactive performance degradation of users by transferring the workloads from overloaded servers to less-loaded ones.
This thesis proposes, implements, and evaluates a scalable dynamic load distribution scheme for multiple server distributed virtual environment systems, where users are highly skewed rather than uniformly distributed over a virtual environment. In the proposed scheme, an over-loaded server initiating load distribution selects a set of servers to be involved in load distribution by dynamically adapting to the workload d status of servers, unlike the existing approaches. Upon completion of the server selection, the intiating server repartitions the regions dedicated to the involved servers using a graph partitioning algorithm such that all the involved servers have the roughly equal workload. The involved servers then migrate their workloads with each other in a peer-to-peer manner according to the result of repartitioning.
The performance of the proposed scheme is evaluated by simulations. The simulation results demostrate that the proposed scheme performs much better than the existing local schemes in terms of effectiveness-how much a load distribution scheme reduces the overloaded users-by more than 20%, and global schemes in terms of overhead-how much users are migrated as a result of load distribution-by about 10%, when the distribution of users is highly skewed. Moreover, the proposed scheme is more scalable than the exist...