存储系统负载自相似性研究综述

发布时间：2018-06-08 21:06

本文选题：存储系统 + I/O负载　；参考：《计算机科学》2013年03期

【摘要】：I/O突发是造成I/O瓶颈的一个主要原因,研究I/O负载中普遍存在的突发性并对负载进行精确合成,对存储系统设计及其性能评价具有重要意义。对实际I/O负载的研究表明,传统的泊松假定难以准确地描述长时间范围内的I/O突发行为。研究发现,I/O突发在不同时间尺度下具有相似性,即I/O负载具有自相似性,因此,自相似模型被用来刻画I/O负载中的长相关性。针对I/O负载自相似参数估计,总结了各种常用的时域和频域估值方法。着重对已有的I/O负载合成模型进行了剖析,讨论了各种自相似模型、多分形模型以及alpha稳定模型的特点。探讨了有待解决的开放性问题,并对I/O负载自相似性研究的发展趋势进行了展望。上述工作将对存储负载的自相似性研究提供有益参考。
[Abstract]:I / O burst is one of the main causes of I / O bottleneck. It is of great significance to study the sudden and accurate composition of I / O load for storage system design and performance evaluation. The research on the actual I / O load shows that the traditional Poisson assumption is difficult to accurately describe the I / O burst behavior in a long time range. It is found that I / O burst has similarity at different time scales, that is, the I / O load has self-similarity, so the self-similar model is used to describe the long dependence of I / O load. Based on the self-similar parameter estimation of I / O load, various methods of time domain and frequency domain estimation are summarized. The existing I / O load composition models are analyzed, and the characteristics of various self-similar models, multifractal models and alpha stability models are discussed. The open problem to be solved is discussed, and the development trend of I / O load self-similarity research is prospected. The above work will provide a useful reference for the study of self-similarity of storage load.
【作者单位】：西南大学计算机与信息科学学院;
【基金】：国家科技支撑计划资助项目(2012BAD35B08) 中央高校基本科研业务费专项资金资助项目(XDJK2012A006) 重庆市自然科学基金项目(2011BB2008) 西南大学博士基金资助项目(SWU111015)资助
【分类号】：TP393.06

【参考文献】