Windows 2000 Performance Guide

Enter RAID

In a landmark paper on RAID, Patterson, Gibson, and Katz of the University of California at Berkeley suggested the reliability concerns with disk striping could be solved by adding redundant data used for automatic error correction to the array. They described five different approaches to building redundant disk arrays, which they named RAID Levels 1 through 5. The taxonomy of redundant disk arrays that Patterson et al developed has become the standard terminology for describing RAID disk architectures.

A decision to implement RAID disks is usually based on overriding disk reliability concerns. Hard drive failures are catastrophic events in the life of a Windows 2000 system administrator. Even with effective backup and recovery procedures, a hard drive failure leads to an extensive computer outage and a loss of service. There are also a number of performance issues associated with different RAID configuration choices. This chapter is mainly concerned with those performance issues. We try to provide a conceptual understanding of the cost/performance design trade-offs that are explicit at each RAID level, and the advantages and disadvantages of each option.

The material in this chapter is organized into three sections. The first section describes RAID Levels 1 through 5 in some detail. We also discuss RAID Level 0, as disk striping came to be known. The second section discusses hardware and software RAID options for the Windows 2000 and Windows NT platforms. Finally, the ...