Patterson, Gibson and Katz, University of California, Berkeley, Didn’t Invent RAID !?

Following our recent article “RAID Is 30-Year Old“, Storage.Newsletter.com exchanges surprising email with Tom Gardner who has excellent historical background, being treasurer and webmaster at IEEE Silicon Valley History Committee and volunteer at Computer History Museum. In the past, he worked at SyQuest, Shugart and Memorex.

First email by Tom Gardner:
Your article on RAID doesn’t quite live up to your normally excellent journalist standards.

As you state the term RAID is thirty years old but the authors didn’t invent anything since the five redundacy techniques were all known and practiced before the Berkeley paper was published.

You don’t have to believe me; here is what Katz has to say: “We were not the first to think of the idea of replacing what Patterson described as a slow large expensive disk (SLED) with an array of inexpensive disks. For example, the concept of disk mirroring, pioneered by Tandem, was well known, and some storage products had already been constructed around arrays of small disks.” [Randy H. Katz (October 2010). RAID: A Personal Recollection of How Storage Became a System]

Mirroring (renamed to RAID-1) was practiced by DEC, IBM, Micropolis and Tandem well before the RAID paper. Parallel ECC renamed RAID-2 goes back to stretch of the 1960s. See in Wikipedia for some more history of redundancy in disk storage.

Until the mid-1990s most of the industry used the term ‘disk array’ rather than RAID. Probably the most successful array company EMC did not use the term RAID to describe its products until about 1995, well after the publication of the paper.

RAID turned out to be a useful taxonomy but it had little to do with the transition from SLEDs to arrays.

Editor’s answer:
In the article you mentioned on Wikipedia link is written: “The term RAID was invented by David Patterson, Garth A. Gibson, and Randy Katz at the University of California, Berkeley in 1987.“

Furthermore, in our paper, we specify that “the technologies of the five levels of RAID were used in various products prior to the Berkeley paper’s publication“, especially at DEC, IBM and Tandem.

Tom, does it mean that the research from the three students of University of California, Berkeley, does not merit the following recompenses: 1998 SIGMOD Test of Time Award, 1999 Allan Newell Award for Research Excellence, 1999 IEEE Reynold B. Johnson Information Storage Award for outstanding contributions in the field of information storage, entrance into the ACM SIGOPS Hall of Fame in 2011 and the 2012 IFIP WG10.4 Jean-Claude Laprie Award in Dependable Computing?

Gardner’s second email and answer:
Thanks for the reply.
 
Sorry, somehow I missed the second half of the article which goes thru the history of redundancy prior to the RAID paper – it reads like the Wikipedia article. Was it in the original posting and I just missed it? Again my apologies.
 
To answer your question, yes, the paper does not justify the awards given to the authors – what ‘research’ did they do and what did it lead to? 
 
The whole paper is flawed since it is based on retail pricing which then for IBM in particular had nothing to do with cost. So their fundamental premise, ‘inexpensive’ turned out to be wrong which is why the industry replaced it with ‘independent’ even though the disks in a RAID array are not independent. SLEDs it turns out were replaced by Redundant Arrays of Expensive Small Disks – RAESDs. Even now most enterprise RAIDs use expensive SAS devices. In the end all the authors provided was an incomplete and inflexible taxonomy which turned out to be adopted by the industry but which IMO doesn’t deserve such recognition.
 
There is a further question about how the concept came into being. The three authors were participants in an industry group sponsored by DEC and including most array storage vendors. The group may have met multiple times before the paper published. Here is what one of the participants, Dave Gordon of Array Technology describes it as it existed before the paper: “My marketing guy, Phil Ingalls, somehow got wind of the fact that Berkeley was studying disk arrays. And so, I don’t remember exactly who called, but I was invited to attend one of their meetings. And their meetings were, kind of, run by Dave Patterson and Randy Katz. And they had a storage architecture committee, mainly people working on their PhDs in storage, and their objective was to use these small drives for large systems, so a similar kind of thing that I was trying to do. So they had developed the concept that was later published in a RAID paper by the same group. But they met couple of times a year with industry, and the objective was for us in industry to comment on the research that the students were doing …”

Oral history of Dave Gordon, Interview by Tom Gardner, Computer History Museum, January 31, 2008, page32

Full disclosure: I interviewed Dave and we had worked together at Memorex. In addition the DEC sponsor of the meetings has stated to me that they preceded the paper.
 
Finally at the time of the paper SLEDs were mainly a mainframe phenomenon; minis had gone to small disks and PCs never used SLEDs. The technology that truly enabled the replacement of enterprise SLEDs was the emulation of CKD track format which is not even mentioned in the paper but was pioneered by Moshe Yanai at EMC. Even today IBM’s mainframe OSs use CKD emulation on RAID arrays.