分布式系统中主机负载预测的一种普适性方法

A Popularly Adaptive Method on Host Load Prediction in the Distributed Systems

分布式系统中为了获得高效的动态负载均衡,需要对主机负载进行有效的预测,这区别于网络流量的预测。论文从分析主机负载的特性入手,依据负载动态变化的规律,提出了一种具有普遍适用性的主机负载预测方法—线性时间序列模型法,并按该方法建立了HLPS(HostLoadPredictionSystem)模型,提出了HLPS的实现策略。

In order to gain effectively dynamic load balancing,it is necessary to predict host load accurately,and this is different from network flow prediction.The paper starts to develop from analyzing the properties of host load,and propose a popularly adaptive method on host load prediction—liner time series models based on the regularity of load's dynamic varieties.According to the method,we construct the HLPS(Host Load Prediction System)model and put forward HLPS's implementing strategy.