基于实时流式计算系统的数据分类节能策略

摘要

To solve the energy efficiency problems of data storage in real-time stream computing,data classification strategy (DCS),based on the idea of classification storage of in memory,was put forward.RAM storage was divided into real time area and duplicate file area.Aiming at the data of real time area,the peak voltage of PRE zone was reduced,and a large number of data nodes were turned off to transform part of data in PRE zone into unpopular zone.The repeated data of duplicate file area was filtered by Bloom filter.Experimental results show that DCS can not only save energy by 27.2% in the field of storage,but also improve the efficiency of the system.%为解决实时流式计算中数据存储的能耗问题,采用内存分类存储的思想,提出数据分类策略(data classification strategy,DCS).将内存存储划分为实时文件区与重复文件区;针对实时文件区中的数据,通过降低预存储区(precharge fast powerdown,PRE zone)的峰值电压,关闭大量的数据节点,使服务器预存储区的部分数据进入不常用文件区,通过布隆过滤器(Bloom filter)过滤重复文件区的重复数据.实验结果表明,该策略在存储方面有效节能约27.2%,提高了系统的效率.