UCT CS Research Document Archive

A Storage System Workload Analyzer

Sikalinda, Paul, Lourens Walters and Pieter S Kritzinger (2006) A Storage System Workload Analyzer. Technical Report CS06-02-00, Department of Computer Science, University of Cape Town.

Full text available as:
PDF - Requires Adobe Acrobat Reader or other PDF viewer.
Other (DVI Format)

Abstract

The analysis of storage system workloads is important for a number of reasons. It is necessary to understand the usage patterns of secondary storage to enable architects to understand and build a new, or improve upon an existing storage system. It is also important for a storage administrator to understand the workload profile when configuring and tuning a system. Furthermore, the analysis of workloads is necessary to understand the storage requirements and behavior of application software. A different reason for workload analysis, is the need to come up with adequate models for performance evaluation. For the evaluation, based on simulation or otherwise, to be reliable one has to analyze, understand and adequately model the workloads.

In this paper we describe a methodology to go about storage system workload analysis and illustrate it with a general tool we call the Enterprize Storage System Workload Analyzer or ESSWA. Although such analyzes are not new, we believe our proposals have a number of advantages over previous analyzers described in the literature. We experimented with ESSWA by analyzing workloads represented by three sets of publicly available workload traces.

EPrint Type:Departmental Technical Report
Keywords:Storage System, I/O Trace, Statistics, Simulation, Probability Distribution Functions, Synthetic Workloads
Subjects:UNSPECIFIED
ID Code:310
Deposited By:Arnab, A
Deposited On:12 July 2006