A Performance Model of Disk Array Storage Systems

The distribution of data across multiple, parallel disks, known as data striping, is a popular method of data storage as it allows for faster access. In this paper, we develop an analytical model that estimates the mean response time, throughput, and mean queue length of data requests when data striping storage techniques are used. Our model is shown via simulations to provide a good approximation of the actual access time under a variety of system and workload parameters designed to emulate real-world conditions. Since analytical results can be readily computed (as opposed to empirical and simulation models), our model should prove useful to le and storage managers in their decisions with regard to the optimal placement of data on storage devices.