U. Farooq and J. Glauert (UK)
Simulation, Scalability, Synchronisation, High Level Architecture (HLA), Time Management (TM).
Splitting and executing different parts of a system with the help of a set of dedicated servers is the most widely used technique for making it scalable. However, it introduces synchronisation issues and requires special attention. Syn chronisation is extensively investigated for parallel and dis tributed systems and generally centralised or distributed ap proaches are used to obtain a consistent view of a system. Centralised methods are easy to implement and achieve precise synchronisation but are not scalable. On the other hand, distributed methods are difficult to manage and in troduce potential communication overhead. They also in troduce longer delays due to dependencies among compo nents at different levels when used with conservative strate gies and complex hierarchical models. In this paper, we propose a fully decentralised synchronisation mechanism to maintain a consistent view of a virtual world. It greatly reduces communication overhead by restricting a server simulating a region to synchronise itself with the servers executing regions sharing boundaries (we define them as adjacent regions) with it. It maintains traditional synchro nisation constraints. The proposed mechanism is illustrated with the help of simple scenarios and an abstract model is used to compare it with traditional approaches. It is demon strated that number of intermediate points among servers and thus complexity and delay in traditional approaches in crease significantly with an increase in number of levels in a hierarchy.
Important Links:
Go Back
