SPEK: a storage performance evaluation kernel module for block level storage systems

11th IEEE/ACM International Symposium on Modeling, Analysis and Simulation of Computer Telecommunications Systems, 2003. MASCOTS 2003. Pub Date : 2003-10-27 DOI:10.1109/MASCOT.2003.1240646

Ming Zhang, Qing Yang, Xubin He

{"title":"SPEK: a storage performance evaluation kernel module for block level storage systems","authors":"Ming Zhang, Qing Yang, Xubin He","doi":"10.1109/MASCOT.2003.1240646","DOIUrl":null,"url":null,"abstract":"In this paper we introduce SPEK (storage performance evaluation kernel module), a benchmarking tool for measuring and characterizing raw performance of data storage systems at block level. It can be used for both DAS (direct attached storage) and block level networked storage systems. Each SPEK tool consists of a controller, several workers, and one or more probers. Each worker is a kernel module generating I/O requests to lower level SCSI layer directly. Compared to traditional file system I/O and disk I/O benchmarking tools, SPEK is highly accurate and efficient since it runs at kernel level and eliminates file system overheads. It is specially suitable for accurately measuring raw performance of data storages at block level without influence of file system cache or buffer cache. Using SPEK, a user can easily simulate realistic workloads and produce detailed profiling data for networked storage as well as DAS. We have built a prototype on Linux and our experiments have demonstrated its accuracy and efficiency in measuring block level storage systems.","PeriodicalId":344411,"journal":{"name":"11th IEEE/ACM International Symposium on Modeling, Analysis and Simulation of Computer Telecommunications Systems, 2003. MASCOTS 2003.","volume":"10 1","pages":"0"},"PeriodicalIF":0.0000,"publicationDate":"2003-10-27","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"11th IEEE/ACM International Symposium on Modeling, Analysis and Simulation of Computer Telecommunications Systems, 2003. MASCOTS 2003.","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/10.1109/MASCOT.2003.1240646","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"","JCRName":"","Score":null,"Total":0}

引用次数: 0

Abstract

In this paper we introduce SPEK (storage performance evaluation kernel module), a benchmarking tool for measuring and characterizing raw performance of data storage systems at block level. It can be used for both DAS (direct attached storage) and block level networked storage systems. Each SPEK tool consists of a controller, several workers, and one or more probers. Each worker is a kernel module generating I/O requests to lower level SCSI layer directly. Compared to traditional file system I/O and disk I/O benchmarking tools, SPEK is highly accurate and efficient since it runs at kernel level and eliminates file system overheads. It is specially suitable for accurately measuring raw performance of data storages at block level without influence of file system cache or buffer cache. Using SPEK, a user can easily simulate realistic workloads and produce detailed profiling data for networked storage as well as DAS. We have built a prototype on Linux and our experiments have demonstrated its accuracy and efficiency in measuring block level storage systems.

查看原文本刊更多论文

SPEK:用于块级存储系统的存储性能评估内核模块

在本文中，我们介绍了SPEK(存储性能评估内核模块)，这是一个用于在块级别测量和表征数据存储系统原始性能的基准测试工具。它可以用于DAS(直接连接存储)和块级网络存储系统。每个SPEK工具由一个控制器、几个工作器和一个或多个探测器组成。每个worker是一个内核模块，直接向底层SCSI层生成I/O请求。与传统的文件系统I/O和磁盘I/O基准测试工具相比，SPEK非常准确和高效，因为它在内核级别运行，消除了文件系统开销。它特别适用于在不受文件系统缓存或缓冲区缓存影响的情况下，在块级精确测量数据存储的原始性能。使用SPEK，用户可以轻松地模拟真实的工作负载，并为网络存储和DAS生成详细的分析数据。我们在Linux上建立了一个原型，我们的实验证明了它在测量块级存储系统方面的准确性和效率。

本文章由计算机程序翻译，如有差异，请以英文原文为准。

求助全文

约1分钟内获得全文求助全文

来源期刊

11th IEEE/ACM International Symposium on Modeling, Analysis and Simulation of Computer Telecommunications Systems, 2003. MASCOTS 2003.

自引率

0.00%

发文量