In-Depth
When Standards Aren’t
Now you see it,now you don't. New products bear a startling resemblance to what you may already have.
The person from Maxtor conducting the press conference for the SCSI Trade Association at SNW Europe in Frankfurt, Germany a few weeks ago seemed genuine enough as he extolled the virtues of the coming wave of Serial-Attached SCSI and Serial ATA Version II (SAS/SATA II) storage solutions. He showed the four or five journalists gathered in the room roughly the same slide deck that had been floating around the industry for the last couple of years.
The familiar pitch began with a description of the transition from parallel to serial interfaces within the storage industry. Back in the 1990s, IBM had serialized SCSI in the form of both Fibre Channel and Serial Storage Architecture (SSA) protocols. While Big Blue opted for SSA on some of its own products, its Fibre Channel protocol was tossed out into the standards world where, in 1999, it was joined at the hip with the concept of storage networks, producing the now infamous and oxymoronical Fibre Channel SAN.
In 2003, Parallel ATA succumbed to Serial ATA (SATA). Driven initially by a desire to thin out the cable clutter inside servers and PCs (thin wire serial cables replace broad flat and thick parallel cables for better airflow through the cabinet), SATA had taken on a life of its own, and products based on the technology proceeded to become the stars of storage in 2004.
Basically, SATA enabled a generation of cheap, capacious, single-ported arrays that targeted primary storage in small and medium-sized business environments and “second tier” storage in larger enterprises. Talk about hitting the market at the right time—SATA came to the fore when regulatory requirements were pressuring big companies to store a lot of stale data for a long time, and as small- to medium-sized companies started buying a lot of storage. EMC’s acquisition of Clariion from Data General was in part predicated on a need to find a platform to serve these burgeoning customer requirements.
The SATA phenomenon, the spokesman said, was small potatoes compared to what was coming. Very soon, he noted, it would be possible to build an array using virtually the same enterprise-class disk drives as are used to build current Fibre Channel or parallel SCSI arrays—but at a lower cost.
Serial-Attached SCSI (SAS) drives, like their Serial ATA cousins, are “90 percent identical” to their parallel peers in terms of hardware. New connectors on the tail ends of the drives provided the only external clue about the differences between SAS drives and their nearly identical Parallel SCSI and Fibre Channel peers.
In the case of SAS drives, a close examination of the electronics wedded to the drive (or on the array controller) would also reveal a new chip set that replaced the Parallel SCSI chips and Fibre Channel to Parallel SCSI bridge chips. You would basically be buying the same hardware you bought with traditional SCSI and FC, but for less money. (He was fuzzy about the actual price differential.)
Things got interesting when he was discussing the set-up of a SAS array. Despite the fact that devices called “SAS Expanders” could be used to create massive storage fabrics combining both SAS and SATA II drives, where component drives were overlaid with a sophisticated protocol for assigning world wide names (WWNs), the fellow went to great pains NOT to make the obvious point out loud: that SAS was poised to usurp rival FC by delivering most of the important capabilities sought from FC arrays.
He went on to note that SAS (given its three-meter cabling limitation) was not being viewed by the industry as an external interconnect (like FC), but only as an internal array interface. Between the lines, it was clear that SAS-SATA II arrays were being groomed to provide a “fabric in a box” solution, not unlike the “multi-tier FC SAN in a box” designs offered by leading enterprise FC array vendors today.
If this was not the intended purpose of SAS, asked one smart French journalist, why did we need SAS at all? The question produced only a bland disclaimer that the SCSI Trade Association was not championing one technology over another, but I could swear that I detected a fleeting smile on the face of the speaker as he gave this pat response sotto voce.
Clearly, his answer was politically correct doublespeak that reflected, more than anything else, that most members of the SCSITA produced BOTH Fibre Channel and SAS components. It was not in their best interest, at least for now, to threaten their more lucrative FC product revenues with a lot of SAS talk.
I listened in earnest for additional validation of this perspective and didn’t have long to wait. Another subtle comparison between FC and SAS was made, however inadvertently, as the fellow described interoperability tests being conducted at the University of New Hampshire. The speaker stressed that SAS interoperability testing was being done BEFORE products were released into market. Whether intentional or not, the comment was a dig at the FC community, which continues to labor over interoperability testing and “plug fests” today, a full five years after their “standards-based” products have been released—and which still appears to be no closer to a uniform product interoperability story today than they were in the late 1990s.
All in all, it seemed that the handwriting was on the wall: SAS arrays will replace kludgey Fibre Channel arrays in all but the most niche applications. Moreover, arrays combining both SAS/SATA version II disk in the same cabinet would soon take over the market from purveyors such as HP and EMC of Fibre Channel/Low-Cost Fibre Channel array configurations. With little more than price to separate the technologies, it is likely that the market will move to the less expensive wares.
That said, however, there are a few caveats worth mentioning. First, SATA II is not becoming the same creature as first generation SATA. The SATA II standard has been dumbed down to enable any vendor to call his product SATA II without having to conform to much of anything in terms of standards-based functionality. I learned this from the presenter off-line, when I asked him about products shortly to appear in the market bearing the monikers “SATAextended” and “UltraWideSATA.” He directed me to visit www.sata-io.org, which fronts for the group that has inherited the mantle of responsibility for SATA testing. A detailed perusal of the site, including a participant agreement for testing in organization plug fests, offered no criterion for justifying a product’s claim to be SATA II compliant. About the only thing that is required is an executed non-disclosure agreement: apparently participants cannot be trusted to keep mum regarding the problems experienced by competitors during testing. Nobody wants their pre-release product performance information showing up in a competitor’s press release or marketing foo.
Without a set of “must have” requirements for SATA II standards conformance, vendors can call any of their SATA products SATA II. This is encouraging balkanization of the standards basis for the most successful product introduction in recent years: SATA disk. Moreover, a balkanized SATA II will compromise the value proposition of SAS itself, even if SAS products are heavily scrutinized and tested for their standards compliance, and could impair the market success of both technologies.
That may be just what the enemies of SAS want. Fibre Channel array makers do not want a competitor just now and will likely leverage any interoperability problems that arise in SATA II to make the case that SAS/SATA II remains an inferior array technology. Meanwhile, a lot of SATA I folks believe that they can make more money by eschewing SAS/SATA II altogether, preferring instead to make arrays with dual path backplanes that offer redundant SATA I disks inside the same cabinet. Such a strategy has the additional value of selling twice the number of well-established SATA disk drives.
Time will tell whether SATA II and SAS will have the impact of either FC SANs or SATA I arrays going forward. We can only hope at this point that SAS/SATA II equipment will be plug-and-play by the time it hits the streets late this year or early next year. Worst case scenario: products will come to market featuring the same Yellow Pages-sized interoperability guides that make Fibre Channel the bane of our existence today—case in point: what happens when standards really aren’t standards at all.
Your comments are welcome at [email protected].