Affiliation:
1. University of Connecticut
Abstract
RAID has long been established as an effective way to provide highly reliable disk subsystems. However, reliability in RAID systems comes at the cost of extra disks and somewhat lower performance. In this paper, we examine some mechanisms to reduce this cost in the context of integration with backup processes. These methods are most useful in storage systems where complete data protection or availability is not necessary such as in desktop personal computers, laptops, and other mobile storage devices. We will in particular investigate strategies with disk and flash that trade off between availability and reliability, snapshotting tradeoffs of reliability between time and space, and user directed redundancy.
Publisher
Association for Computing Machinery (ACM)