In-Depth
Differentiating Storage
Now that SAS/SATA has achieved the status its champions said it would, how do vendors blend the software functions with their gear and provide usable functionality at a reasonable cost?
Ask anyone who builds storage arrays for a living and they will tell you it is difficult to differentiate one product from another. Some acknowledge that this reflects the inevitable march toward commoditization and is a by-product of standards development, which, at a hardware level at least, has mostly caught up with the products in the market today. Others suggest that consumers are increasingly focused on cost per gigabyte or per terabyte as the most important discriminator of array value.
Both are correct to some degree. After more than ten years of Fibre Channel plug fests at the University of New Hampshire (UNH) and elsewhere, it is apparent that the standards for Fibre Channel haven’t delivered what a standards are supposed to deliver: interoperability between products. Today, standards such as Serial Attached SCSI (SAS) and Serial ATA (SATA) are much more advanced. After some hiccups with SAS early on, the most recent round of UNH tests suggest that the bugs have been largely worked out and that RAID controllers from a vendor will probably work with drives from all other vendors. This is how storage hardware ought to be.
The rise of SAS/SATA is a curious thing. SAS delivers every bit the value of Fibre Channel, from dual porting on drives to a deep command-queuing space for robust, enterprise-class data handling, which helps explain the initial subtle and not-so-subtle disparagement of the technology by certain Fibre Channel vendors. A SAS drive is, in fact, the same physical drive as a Fibre Channel (or parallel SCSI) drive, but with a different connector and a different bridge chip to convert the protocol back into its native parallel SCSI command format. SAS drives, however, are about 20 or 30 percent cheaper when compared to Fibre Channel.
As feared by the FC crowd, in many cases the same vendors who manufacture SAS gear, the rise of SAS is putting a crimp in their expansion plans for more lucrative FC products. Some are still talking about FC and SAS as though they are companion technologies addressing different needs at different price points; but if you ask them, off the record, which products will sell in greater numbers in the future, they'll tell you it's clearly SAS, not FC.
SAS has something going for it that FC does not: SATA. A SAS controller supports both high- end enterprise-class drives—and with less expensive, highly capacious, SATA drives—all in one kit. For a while, EMC and HP toyed with an alternative Fibre Channel array configuration comprising high-end FC drives and “low-cost Fibre Channel” (LC-FC) drives, but both companies have now acknowledged that SATA is what folks want in a “second tier” of storage. Both have announced that SATA is what they will be providing in “tier 2” arrays going forward.
Bottom line: SAS/SATA has become the fait accompli that its champions said it would be. The gating factor was always interoperability, and I had my doubts early on that the industry would be able to pull itself together and get all of its competitors “singing from the same sheet of music.”
This concern was reinforced at the first SAS plug fest, where, in order to get cooperation from all of the vendors, the organizers had to loosen the definition of SAS—which is to say that they dropped just about all of the criteria that made a SAS product a SAS product. Not surprisingly, the results of the first interoperability tests were abysmal failures.
This translated into real-world experience with SAS/SATA. About a year and a half ago, when we built some SAS arrays for testing in our labs, it was difficult to find SAS controllers that would work with the microcode on SAS drives, let alone support a mixed configuration of SAS and SATA.
Today, standards conformance has been tightened and plug-and-play interoperability is the norm rather than the exception. The only thing we haven’t seen yet is a wholesale replacement of the Fibre Channel protocol with one based on SAS. SAS has its own worldwide naming scheme and delivery method for commands and data that could readily replace Fibre Channel. No one talks about it openly, of course. It is always acknowledged by evangelists from the SCSI Trade Association with a wink and a nod. However, suggesting that the cash cow of storage, FC, might be vulnerable to new, less-expensive technology is a sure road to a new career out of storage and into the exciting world of fast food.
Some are already referring to Fibre Channel fabrics as “legacy SANs” and looking to lower-cost alternatives to provide the same return In the realm of SAS/SATA, one vendor who will be hard to beat is Promise Technology.
Mike Joyce has just come aboard at Promise as a senior director of marketing and I interviewed him last week. Joyce has a lot to work with at a company that has a long list of industry firsts (first ATA RAID cards in 0/1/10 and RAID 5, first SATA RAID 5 card, etc.), but he is also keenly aware that you can’t sit on your laurels for long in this business, even with over 15 million installations of arrays based on your RAID controllers. The dilemma he confronts—now that Promise has diversified from making component boards for arrays into building a full range of RAID arrays in form factor, speeds and feeds, and price points suitable to small firms, medium sized enterprises, and now large scale enterprises—is how to differentiate the company and its products.
The joy of solid standards such as SAS and SATA is that anyone can build arrays with standards-based components and deliver roughly the same business value. We have done speed tests on some products from Promise and have been impressed by their performance. They have surpassed products by competitors such EMC/Dell at the low end and have stood their ground admirably with IBM offerings for the midrange—always at a fraction of the cost of competitor products. Shortly, we will be testing their VTrack E-class and J-class arrays which are targeted at the enterprise.
We understand Joyce’s problem, however. Promise is capable of delivering a 2U 12-bay or 3U 16-bay SAS (or Fibre Channel) rack mount array that is every bit as good as a three-letter vendor’s box. What they haven’t done is to “add value” with extra software to do things such as snapshot, point-in-time mirror splitting, content addressing, or thin provisioning, which other vendors have offered to increase the price of their otherwise commodity array products.
Joyce believes, and so do I, that customers may be better served by buying exactly the software that they need—apart from the box—and implementing it in a way that makes sense. Competitors who blend these software functions with their gear often provide more functionality than the customer needs or will ever use, but the customer must pay for licenses and maintenance on all of the software whether they use it or not.
On the other hand, when you talk to Compellent or 3PAR or Equallogic about the value-add functionality that they have included in their wares, the explanation they give also makes a sort of internal sense. Many consumers lack the staff, the skills, or the desire to cobble together hardware and software and to customize storage to meet their exact requirements. These folks turn to the one-stop-shop vendor who takes on full responsibility for fixing any problems that develop.
This may be true, and it may be worthwhile for Promise Technology to find some partners with desirable value-add software products who can not only tightly integrate their wares with Promise array products, but who can also blend their customer support functions with Promise’s so that the “one-throat-to-choke” solution can be provided. On second thought, this is the role of a good integrator/reseller—so, Promise might invest some of its resources developing a great indirect channel.
The problem with this strategy, of course, is that many integrators no longer do storage integration. Instead, they pass through, with a mark-up to compensate them for their time, a pre-packaged solution from a brand-name vendor. Comparatively few reseller/integrators have the skills, the time, or the interest to be anything more than order takers these days.
Still, for Promise to realize the full potential of its offerings, it needs to pick a few niches where it can own the market. Strategic partnerships with, say, Caringo, could enable it to provide a content-addressing solution that is better by a long shot than EMC’s Centera, and at a fraction of the cost. Alliances with a few other software providers I can think of could enable the company to own the short term archiving space, the disk-to-disk replication space, or even the audio/video post-production and editing space.
Joyce’s real dilemma is how to map products to business needs—not just in a horizontal storage infrastructure sense, but in a vertical business process-focused way. From those niches, the company could expand outward into more and more niches until Promise is a household name.
There are impediments to this strategy, of course. One is that Promise must be sensitive to its bread-and-butter OEM channels. They already sell their RAID components to most of the name-brand vendors out there. Joyce will need to carefully manage the perceptions of this OEM channel to ensure that he does not at the same time appear to be a competitor and a partner. Adaptec finds itself in a similar conundrum with respect to its fine SnapServer product line.
The continued success and expansion of Promise’s great and growing product family will be entirely dependent upon partnerships and their wise management of them. In this, I wasn’t telling Joyce anything he didn’t already know. The good news is that he seems vigorous, engaged, and cognizant of the fact that he is backed by some great technology. In a down economy, where price takes on significant importance in determining product acquisitions, Promise can’t be beat.
We will continue to track Promise Technology and look forward to testing their latest gear in our labs shortly. We will report our findings, and you are welcome to report yours: [email protected].