* SCSI's longevity shows it's no longer the 'small' computer storage interface
Once, long ago, in a galaxy not so far away, storage interfaces were pretty much proprietary things. Eventually the world split into storage devices dedicated to mainframes and storage devices devoted to everything else – “everything else” in this case being workstations, minicomputers and, by the mid-1980s, personal computers. It is from this second category that we got SCSI (small computer storage interface), a set of commands used for accessing disk, tape and other storage devices.
Once, long ago, in a galaxy not so far away, storage interfaces were pretty much proprietary things. Eventually the world split into storage devices dedicated to mainframes and storage devices devoted to everything else – “everything else” in this case being workstations, minicomputers and, by the mid-1980s, personal computers. It is from this second category that we got SCSI [small computer storage interface], a set of commands used for accessing disk, tape and other storage devices.
Today of course even the very largest machines use some aspect of SCSI, so the term “small” no longer is an appropriate descriptor. SCSI commands in fact show up in the darndest places. On one extreme, SATA translates SCSI commands for its read and write operations, as well as for several other key management functions. On the other, the cognoscenti among you know that the SCSI command set is fundamental to Serial Attached SCSI (SAS), to iSCSI, and even to what some folks take to be SCSI’s mortal enemy, Fibre Channel.
SCSI has been an adaptable standard. It began at a transfer speed of 5M byte/sec in the mid-1980s (on a parallel 8-bit bus, you may recall), then doubled and redoubled its speed until it reached 320M byte/sec in the last few years. This was judged to be the practical limit for a parallel interface.
Now SCSI is undergoing a fundamental change: after a quarter century it is evolving from a parallel bus to a serial architecture. The new SAS serial architecture now runs at 300M byte/sec, will double its throughput in a few years, and can accommodate literally thousands of devices when used with expanders. The result is that, although the path has clearly been evolutionary, what we see today is not at all your father’s SCSI. In its new incarnation, SCSI has the legs to carry us all well into the 21st century.
Longevity like this is almost unheard of in the computer business, and SCSI’s long run is surely a tribute to the original planning and continuing work done by the INCITS T10 technical committee and, following that, by the SCSI Trade Association (STA). The early movers contributed a lot, as do those who today continue to contribute to the standard’s continued development.
Two of the early movers in the SCSI world are old colleagues of mine from my days at NCR and Symbios Logic (now LSI Logic): John Lohmeyer has provided leadership and sweat equity to the T10 group since the early days, and Harry Mason has lead the STA since the first bits began to trickle down the first SCSI bus.Nice going guys, and nice going to Marty Czekalski, Bill Ham, and the literally hundreds of others across the industry who have made SCSI a success.
It may well be that no one has ever loved SCSI, but even SCSI’s competitors in the marketplace must admit that the thing can be relied upon to do what it is asked to do.
SCSI celebrates its 25th birthday this year. Happy birthday SCSI. Someone should buy you a beer.
