版权所有:内蒙古大学图书馆 技术提供:维普资讯• 智图
内蒙古自治区呼和浩特市赛罕区大学西街235号 邮编: 010021
作者机构:Zhejiang Gongshang Univ Sch Comp & Informat Engn Hangzhou 310018 Zhejiang Peoples R China Chinese Acad Sci Inst Comp Technol Beijing 100080 Peoples R China Zhejiang Univ Technol Hangzhou 310014 Zhejiang Peoples R China
出 版 物:《MATHEMATICAL PROBLEMS IN ENGINEERING》 (Math. Probl. Eng.)
年 卷 期:2013年第2013卷第1期
页 面:1-13页
核心收录:
基 金:Natural Science Foundation of China National Basic Research Program of China [2004CB318205]
主 题:CACHE memory COMPUTER algorithms CLOUD computing SIMULATION methods & models CLIENT/SERVER computing
摘 要:Data congestion and network delay are the important factors that affect performance of cloud computing systems. Using local disk of computing nodes as a cache can sometimes get better performance than accessing data through the network. This paper presents a storage cache placement algorithm-C-Aware, which traces history access information of cache and data source, adaptively decides whether to cache data according to cache media characteristic and current access environment, and achieves good performance under different workload on storage server. We implement this algorithm in both simulated and real environments. Our simulation results using OLTP and WebSearch traces demonstrate that C-Aware achieves better adaptability to the changes of server workload. Our benchmark results in real system show that, in the scenario where the size of local cache is half of data set, C-Aware gets nearly 80% improvement compared with traditional methods when the server is not busy and still presents comparable performance when there is high workload on server side.