Evaluating the Effects of Upgrading Heterogeneous Disk Arrays

摘要

When a set of disks with different capabilities must be configured as a RAID system in low-cost clusters, applying a block placement technique that takes advantage of disks with higher capabilities (AdaptRaid) is the key to maximize the array potential (capacity and performance). However, the AdaptRaid practical application requires adaptability mechanisms because the current applications need expandable and reliable storage systems working in a 24/7 manner. This paper presents an algorithm that offers a reliability mechanism when recovering a damaged disk, upgrade mechanisms when adding or replacing disks, and a transition mechanism from an installed RAID system to an AdaptRaid one. Focusing on the increase of parallelism and without interrupting the normal operation, the mechanisms change a RAID or AdaptRaid block placement to an AdaptRaid one to greatly improve the arrays potential and achieve a balanced load. We examine the algorithm effects on the array normal operation. The evaluation reveals that the mechanisms improve the performance and reliability of arrays producing a low overhead.