Scaling knowledge based virtual worlds: An efficiency gain through multi-core architectures

Abstract

This paper investigates the impact of sequential content load algorithm on regional transfer in OpenSimulator framework. It, then, presents an abstract parallel content load algorithm, which is implemented using the basic parallel for, and chunk partitioning constructs to exploit the parallelism of multi-core architectures. This work used a set of five example virtual worlds content to investigate the effectiveness of the proposed work and its impact on the reallocation process of regions in OpenSimulator. Simulation results revealed that the implementation using parallel for, reduced the timings by 47% to 57% while the implementation using chunk partitioning minimized it by 63% to 72%. Time impact on regional transfer was reduced by 39% to 51% using parallel for loop, while 52% to 66% using chunk partitioning.