Author: Willis Whittington, Senior Manager, Interface Planning,
It was clear that parallel SCSI was running out of steam when the transition from Ultra 3 (160 MB/sec) to Ultra 4 (320 MB/sec) brought with it a new terminology to articulate the technical challenges faced by the speed increase. Engineers were now forced to debate the parts and merits of a variety of “…ations” such as pre-compensation, post-compensation, equalization, packetization and skew (how did that get in there?). It was obvious that an interface technology based on the parallel transmission of data bits down a wide cable was facing a technological Armageddon and Ultra 640 SCSI would never see the light of day. Thus the SAS story began.
Six storage provider companies got together in Houston TX to discuss and float a proposal for a new serial interface to replace parallel SCSI. This new project needed a code name that would be easy to remember and eminently suitable to the new endeavor, so looking no further than the tasty local brew at their elbows, this intrepid group of interface pioneers settled unanimously on the appellation “Shiner Bock.”¹
The proposal for the new interface was given an airing to a wider audience at the Fall Comdex show of 2001. By January of 2002, twenty five companies contributed specialist engineers to a Working Group to develop a specification for the new interface which had now relinquished its code name and rejoiced in the new assignation of Serial Attached SCSI. Just four months later on the 29th of April, a preliminary specification (at Rev. 16), and further responsibility for the development of SAS, was handed over to the INCITS T10 technical committee and the original Working Group was disbanded.
Not only was T10 charged with creating a new serial interface to replace the trusted 20-year-old workhorse, but new features and enhancements had to be incorporated into the design to improve performance (full duplex, wide ports), reliability (dual port, point-to-point architecture), interface speed (initially 3 Gbits/sec with a roadmap to 6 and 12 Gbits/sec) and scalability. With a view to bridging the gap between enterprise and desktop storage, the specification called for electrical and mechanical connector compatibility with Serial ATA (SATA), the new serial interface successor to Parallel ATA. This would allow mission critical and business critical systems access to high capacity, low-cost storage in the same subsystem, where lower performance, reliability and duty cycle could be traded for fewer dollars per gigabyte.
By the end of 2002, the specification was far enough advanced to allow ASIC designs to be formalized and submitted for fabrication and evaluation, a process that normally takes about 18 months. In parallel with the ASIC development, the specification continued to mature, and “breadboard” prototypes were used to test the new protocol and its multiple state machines. A change/test/prove development cycle ensued and continued until the industry met the next major milestone event. In March 2004, less than two years after the handoff to T10, the SCSI Trade Association (STA) held the first SAS plugfest at the University of New Hampshire where 16 companies from various disciplines in the storage infrastructure got together to test the design and interoperability of their SAS wares and build confidence in their products. This plugfest spurred the need for more and the attendance grew at the follow-on events in June and November of the same year.
By the end of 2004, the industry had shipped 26,000 SAS drives into development sites throughout the US to support design verification testing of cables, connectors, mid-planes, expanders, controllers, systems, and of course, the drives themselves. Two more plugfests were held in 2005 and drive shipments in that year swelled to 816,000, as the interface design matured and SAS systems began to appear.
In 2006, SAS will start to make its mark with an estimated 3< million drive shipments, representing 12% of the SCSI market (SCSI, SAS, FC) and 9% of the total enterprise HDD market which includes both PATA and SATA. Although the SAS demand is initially expected to be oriented toward Direct Attached Storage (DAS), there is already some modest activity in the NAS/SAN direction which could account for upwards of 15% of SAS shipments by 2009.
The banner year for SAS is forecasted to be 2007 when shipments are expected to approach the 12M mark and SAS becomes the leading interface in the enterprise arena. With the demise of the parallel interfaces all but complete in 2009, the compatible SAS/SATA duo is on track to reign supreme with 85% of the 54M unit enterprise drive market.
From Bock, the original idea and proposal, to Bonanza, the leading interface in only five years: in the history of storage, an unprecedented adoption rate for a storage interface.