基于分形整形器的GPS系统性能确定上界研究

Deterministic upper bounds on performance of generalized processor sharing based on fractal regulators

为了精确求解以自相似业务作为输入的网络性能确定边界问题,在GPS(generalized processor sharing)系统入口处利用分形整形器对自相似业务进行调控,提出了基于分形整形器的GPS系统模型,基于确定性网络演算理论推导了使用自相似业务作为输入时该GPS系统的缓存队列长度、延迟、延迟抖动以及有效带宽等性能确定上界模型。数值结果表明,基于分形整形器的GPS系统性能确定上界模型对自相似业务具有较好的性能。

In order to get exact bounds problems encountered in networking with self-similar input traffic, a fractal regulator was introduced into the ingress of a generalized processor sharing （GPS） system for regulating the self-similar traffic, and a model was presented based on fractal regulator and GPS. The deterministic bounds on queue length, delay, delay jitter and effective bandwidth of the system with self-similar input traffic were derived using network calculus. Numerical simulations demonstrate that the bound models of a GPS system based on fractal regulators using network calculus are suited for the networks with self-similar input traffic.