The encyclopedia · Engineering & Operations · Technical decision · 1987–1988
Berkeley's RAID paper made cheap small disks outperform one big drive.
Patterson, Gibson and Katz showed arrays of PC disks with parity beat single mainframe drives on speed, cost and reliability.
University of California, Berkeley
the move
By the late 1980s, mainframe-class disks were fast but cost so much per megabyte that storage budgets were enormous, while smaller PC disks were cheap but slow individually. The 1988 SIGMOD paper, A Case for Redundant Arrays of Inexpensive Disks (RAID), argued the industry was solving the wrong problem.
David Patterson, Garth Gibson and Randy Katz showed that arrays of small disks could outperform a single large drive: striping data across the array multiplied bandwidth, and redundancy provided fault tolerance that a single point-of-failure drive did not have.
The Computer History Museum credits the paper with catalyzing the industry: RAID made high-performance, reliable storage affordable by building it from commodity parts, and the name itself changed how storage was designed.
why it works
- Striping turns many slow disks into one fast aggregate.
- Parity lets the array survive a disk failure without a second full copy.
- Small drives' lower cost per byte changed the economics of capacity.
- The taxonomy of RAID levels gave engineers a shared language.
what transfers
When one expensive component is the bottleneck, question the component, not the workload: many cheap parts with a little redundancy can outperform the premium single part.
what came after
RAID moved from a research proposal to the standard architecture of server storage within a decade, and the term survived even as drives grew. Every major storage vendor shipped RAID products, and the paper remains one of the most cited in systems research.
references
- A Case for Redundant Arrays of Inexpensive Disks (RAID)
- 1988: U.C. Berkeley paper catalyses interest in RAID
spotted an error? The archive wants to know.