DiskSeen預(yù)取算法的分析及優(yōu)化研究
發(fā)布時(shí)間:2018-09-04 08:35
【摘要】:計(jì)算機(jī)存儲(chǔ)層次結(jié)構(gòu)是一種典型的金字塔形結(jié)構(gòu),以平衡計(jì)算機(jī)對(duì)存儲(chǔ)系統(tǒng)的兩方面需求,即高速處理數(shù)據(jù)和大的存儲(chǔ)容量。然而隨著信息技術(shù)的飛速發(fā)展,計(jì)算機(jī)處理器和磁盤之間的速度鴻溝持續(xù)擴(kuò)大,因而磁盤訪問便成為一個(gè)影響計(jì)算機(jī)系統(tǒng)性能的瓶頸問題。近幾十年來,如何減小磁盤訪問延遲對(duì)整個(gè)計(jì)算機(jī)系統(tǒng)性能的影響,一直是存儲(chǔ)領(lǐng)域的熱點(diǎn)研究問題。預(yù)取技術(shù),通過提前預(yù)測(cè)I/O請(qǐng)求并將數(shù)據(jù)讀入緩存中,以對(duì)上層應(yīng)用程序隱藏I/O延遲,是緩解這一瓶頸問題的重要技術(shù)手段。DiskSeen是一種塊級(jí)預(yù)取算法,通過分析磁盤塊的位置和訪問時(shí)間的聯(lián)系來提高磁盤訪問的順序性和總體的預(yù)取性能。針對(duì)DiskSeen算法,文中主要做了以下幾方面工作:首先,分析DiskSeen算法的不足之處,據(jù)此提出動(dòng)態(tài)控制預(yù)取粒度和二次匹配激活歷史預(yù)取方法,以優(yōu)化效率;然后,實(shí)現(xiàn)了DiskSeen算法及改進(jìn)后的算法;最后,在模擬仿真實(shí)驗(yàn)環(huán)境下對(duì)算法進(jìn)行了性能對(duì)比測(cè)試。實(shí)驗(yàn)結(jié)果顯示,DiskSeen算法能夠明顯提高緩存命中率并減少平均響應(yīng)時(shí)間,而優(yōu)化后的DiskSeen算法則可以進(jìn)一步提升上述兩方面的系統(tǒng)性能。
[Abstract]:Computer storage hierarchy is a typical pyramid structure to balance two aspects of computer storage system requirements: high-speed data processing and large storage capacity. However, with the rapid development of information technology, the speed gap between computer processor and disk continues to widen, so disk access becomes an impact. In recent decades, how to reduce the impact of disk access latency on the performance of the entire computer system has been a hot research topic in the storage field. Prefetching technology, by predicting I/O requests in advance and reading data into the cache to hide I/O latency from the upper application, alleviates this bottleneck problem. DiskSeen is a block-level prefetching algorithm to improve the order of disk access and the overall prefetching performance by analyzing the relationship between the location of disk blocks and the access time. The prefetch granularity and the secondary matching activation history prefetching method are designed to optimize the efficiency; then, the DiskSeen algorithm and the improved algorithm are implemented; finally, the performance of the algorithm is tested in the simulation environment. The experimental results show that the DiskSeen algorithm can significantly improve the cache hit rate and reduce the average response time, and is superior. The modified DiskSeen algorithm can further enhance the performance of the above two aspects.
【作者單位】: 華中科技大學(xué)武漢光電國家實(shí)驗(yàn)室;
【分類號(hào)】:TP333
本文編號(hào):2221566
[Abstract]:Computer storage hierarchy is a typical pyramid structure to balance two aspects of computer storage system requirements: high-speed data processing and large storage capacity. However, with the rapid development of information technology, the speed gap between computer processor and disk continues to widen, so disk access becomes an impact. In recent decades, how to reduce the impact of disk access latency on the performance of the entire computer system has been a hot research topic in the storage field. Prefetching technology, by predicting I/O requests in advance and reading data into the cache to hide I/O latency from the upper application, alleviates this bottleneck problem. DiskSeen is a block-level prefetching algorithm to improve the order of disk access and the overall prefetching performance by analyzing the relationship between the location of disk blocks and the access time. The prefetch granularity and the secondary matching activation history prefetching method are designed to optimize the efficiency; then, the DiskSeen algorithm and the improved algorithm are implemented; finally, the performance of the algorithm is tested in the simulation environment. The experimental results show that the DiskSeen algorithm can significantly improve the cache hit rate and reduce the average response time, and is superior. The modified DiskSeen algorithm can further enhance the performance of the above two aspects.
【作者單位】: 華中科技大學(xué)武漢光電國家實(shí)驗(yàn)室;
【分類號(hào)】:TP333
【相似文獻(xiàn)】
相關(guān)期刊論文 前1條
1 劉立;陳明宇;包云崗;許建衛(wèi);樊建平;;一種基于頁面級(jí)流緩存結(jié)構(gòu)的流檢測(cè)和預(yù)取算法[J];計(jì)算機(jī)研究與發(fā)展;2009年10期
,本文編號(hào):2221566
本文鏈接:http://sikaile.net/kejilunwen/jisuanjikexuelunwen/2221566.html
最近更新
教材專著