A Case for Redundant Arrays of Inexpensive Disks (RAID)

A Case for Redundant Arrays of Inexpensive Disks (RAID)
by David A. Patterson, Garth Gibson, Randy H. Katz

Important Points --
* The paper addresses innovation necessary to speed up I/O intensive applications.
* The fundamentals introduced by the paper are still applicable, and further improvements like RAID-DP(RAID level 6) build on top of the ideas presented here.
* A new paradigm is presented to improve performance, reliability, power consumption and scalability of disk usage.
* The metrics are appropriately defined for supercomputers and transaction processing systems, depending on the access patterns.
* The paper is clear and iteratively explains the advantage of each level, the problem in the level, and how the next level adddresses a particular problem.
* The quantitative and qualitive arguments in the paper about the advantages of the taxonomy of 5 different organizations of disks arrays are spot on.
--
Deficiencies--
* For level 3 RAID, the paper does not describe the ECC code used and thus how ECC is calculated using only one check disk per group. Though a reference to a paper [Park86] has been provided, it would have good to have a small description in the paper as well. Thus it was easy to determine why level 3 RAID is better suited to supercomputing applications than to small transaction processing systems.
* The paper could have talked more about how the ideas can be implemented by software instead of hardware.
--
Conclusion--
The paper presents the advantages of using inexpensive PC disks and advantages of 5 different levels of organization of these arrays. It is able to effectively describe the advantages of the different levels in terms of important quantitative parameters, and the kind of applications for which each level is suited. I think the paper marked a major change in data storage paradigms, applicable across the computing industry, whose fundamentals are still applicable nowdays.