摘要： 设计并实现一个基于分级Cache的透明计算系统HCTS，在系统客户端和服务端采用两级缓存来提升I/O性能。在缓存的管理策略上，针对透明计算应用环境，以提高缓存命中率为主要目标，提出一种基于访问频率计数阈值的改进LRU置换算法LRU-AFS。测试结果表明，当网络环境中的客户主机数不断增加时，与普通透明计算系统TS相比，HCTS能够在减少网络流量的同时大幅缩短客户机启动时间，提高随机读写吞吐量。

关键词: 透明计算, 分级Cache, I/O性能

Abstract: A novel transparent computing system called HCTS is proposed in this paper. HCTS adopts the hierarchical cache strategy in client and server respectively to improve the I/O performance of the system. In order to improve the cache hit ratio according to the given environment of transparent computing, this paper presents a modified LRU replacement algorithm called LRU-AFS based on the count threshold of data accessing times, which is used to distinguish between frequently used data and rarely used data. Test results show that, with growing in number of clients in LAN, HCTS can better improve and enhance the client I/O performance compared to general TS, reduce the network traffic dramatically, shorten the boot time significantly and increase the random read-write throughput of clients effectively.

Key words: transparent computing, hierarchical Cache, I/O performance

中图分类号: 

• TP311