* SCSI history
When SCSI first appeared on computer systems in the mid-1980s, it was precisely what its name indicated. SCSI – for those of you with an interest in etymology but who lack a sense of history – stood for “Small Computer Systems Interface,” and was designed to go into the early engineering workstations being built by HP, SGI, Sun, and any number of other vendors now among the dear departed. (Remember Apollo, Ardent, Dana, Data General, for example? How about Stardent?)
SCSI was designed to serve the high performance data needs of the CAD and engineering markets. In those days we had 8-bit and 16-bit connections that moved data at 5M-bytes and then at 10M-bytes per second, and SCSI began to creep out of the engineering division and onto the IT floor, where it first appeared in early RAID implementations hanging off superminis from Prime Computer and others.
By the mid-1990s storage needs increased, disk drives had grown and I/O requirements got much more demanding. What had previously been “Fast SCSI” (10M-byte) was no longer seen as being fast at all, and as SCSI-based RAID were being connected to some very large machines, the meaning of the first “S” in SCSI had long-since ceased to apply. SCSI was now on large systems too.
Important things were happening. The ANSI T10 technical committee (http://www.t10.org) defined the evolving SCSI standard, driving the performance specification from 10M- to 20M-bytes (Fast 20), then to 40M-bytes (originally Fast-40, then renamed Ultra2 as by that time the old definition of “fast” became obsolete at almost every T10 meeting).
Also in the mid-1990s chip makers like Adaptec and NCR Microelectronics (now LSI) got together with disk vendors (Quantum was the early technology leader in those days, followed by Seagate) and makers of cabinets and connectors to provide interoperable solutions and to help drive the evolving standard through the committee. They also rolled out the first products and helped found the SCSI Trade Association (https://www.scsita.org/), a vendors’ organization that markets the SCSI concept to the world. I remember that a lot of my work at NCR (by that point rebadged as Symbios Logic) was in support of all this.
Right now, although Fibre Channel seemingly gets most of the press, the preponderance of enterprise data stored on spinning disk is stored via SCSI.
By the first decade of this century, the specs were in for Ultra160 and Ultra320. It was also plain to see that when I/O reached 320M-bytes per second, SCSI had run out of headroom and had no way to grow. Thus SCSI, previously a parallel interface, was re-engineered for serial connectivity.
Serial Attached SCSI (SAS) can support transfer rates in excess of 1G byte/sec, far faster than either parallel SCSI or Fibre Channel, and can provide space on the bus for far more drives as well. And now, at last, products are starting to emerge from the pipeline.
IBM announced SAS products earlier this year. Last week in San Jose, SAS raised its head again as four companies got together to roll out more products. Next time we’ll take a look at these, and discuss why you should care.
