In-Depth
Storage Standards, Part I: PBS or Reality TV
Are standards something we should really care about?
Someone once quipped that the nice thing about storage standards is that there are so many of them. It is a bit of whimsy that comes to mind whenever I discuss with a company how they are building and managing their storage infrastructure.
To begin at the beginning, there are two kinds of standards: de facto and de jure. De facto standards are those technologies that have come into widespread use as a result of the predominance of a particular vendor. De jure standards are codified and adjudicated by standards groups that, through a process of open peer review and voting, approve a technology (protocol, architecture, or method) as an industry standard.
Two key standards groups in storage today are the American National Standards Institute (ANSI), where Small Computer Systems Interface (SCSI) and Fibre Channel-related protocols developed elsewhere are debated and formalized, and the Internet Engineering Task Force (IETF), where the Internet Protocol suite (and storage protocols that leverage it, such as iSCSI and Remote Device Memory Access) are worked and reworked.
Most of the technology that becomes standard is contributed by vendors or by ancillary organizations such as industry associations that submit it to ANSI or IETF so it can be subjected to peer review. This crucible is supposed to produce common intellectual property that, among other things, can be implemented by vendors in their products to ensure interoperability between respective proprietary wares.
The case could be made that most standards are both de facto and de jure. Microsoft, for example, recently received the nod from ECMA, a Swiss standards body, for its OpenXML document format, popularized by its dominant word processing application, Microsoft Word. Over the years, Cisco Systems’ routing protocols and Brocade Communications’ FC fabric switching protocols have been ratified by standards bodies where those companies either chaired committees or otherwise participated on a massive scale—in part because of their dominance at a given time within the specific markets served by these vendors.
It would take more space than we have here to discuss the rarified world of standards development—and the manipulation of the standards-making process by vendors. If you want to investigate that subject, go to your local library and check out a copy of a book I wrote a couple of years ago: The Holy Grail of Network Storage Management. In this column, I just wantto establish a level playing field on the role of standards in building storage.
Standards are something that many start-up vendors build into their wares to cajole storage administrators into taking a chance on new products largely untested in the marketplace. Frequent reference to standards by these vendors in their marketing literature is intended to instill confidence in the would-be buyer that, even if the seller doesn’t survive Round B of its venture capital funding hunt, what you buy from them today will still serve you through the three to five year lifespan that most buyers expect from their storage gear.
Standards rarely take center stage in the marketing pitches of wares from established brand-name vendors. The assumption is that a company with a pedigree in the market, one with many notable customers to its credit, is delivering consistent value that may or may not embrace standard technology, and in any case isn’t going to disappear from the map anytime soon. To paraphrase the TV ad, you are in good hands with EMC, IBM, HDS, etc.
No Guarantee
Truth be told, however, standards are rarely an ironclad guarantee of interoperability—or of anything else. Parallel SCSI, for example, was a set of standards implemented differently by different vendors in different products. Storage builders have always needed to test host bus adapters with arrays and array controllers with disk drives to ensure that a group of components would hold together as a unified system.
We have been seeing the same phenomenon with Serial Attached SCSI products in our test labs. Just because a SAS specification has been approved as a standard doesn’t mean that Fujitsu or Seagate SAS drives will work with Adaptec or LSI Logic SAS controllers. Firmware implemented on controller boards and drive electronics, which are proprietary to their manufacturers, determine whether the components will work and play well together. Also, many vendors use the term “drive signing”—which is often a code phrase for a special formatting of disk—to limit the alternatives available to consumers for “driving” their arrays. Even if a certain Seagate drive is pretested and certified for use in an array, the array vendor often wants the consumer to source drives only from the vendor itself (and often at an obscene price mark-up) and uses special drive signing or formatting to preclude you from buying the identical drive directly from the drive manufacturer or an approved distributor.
Fibre Channel standards present another dimension of the standards “reality show.” While standards exist in the Fibre Channel world around components such as switches, they have been written with sufficient wiggle room to allow multiple switch vendors to create “standards compliant” equipment that will not interoperate with their competitor’s “standards-compliant” wares. Alternatively, connecting FC switches together with only their standards-based features activated (and their proprietary features turned off, if this is an option) reduces the capabilities of the switch and the value-add for which the buyer paid so much money.
In a conversation last year with a forthcoming marketing person (odd combination that) for a second-tier storage systems vendor, I asked what had become of his emphasis on “standards-based architecture” in his product literature. His response was telling: “Standards are important until you gain market share. Then they aren’t so important anymore.”
I have gradually reached the conclusion that, in the real world, standards have become marketing tools more than anything else. They are useful to vendors only as they contribute to making a sale. Vendor engineers might give me an argument on this point, but never very loudly in my experience. While “pure” technologists seek standards, competitive vendors eschew them as the commoditizing force they might be. If all storage were standardized, there would be little to differentiate one product from another. This has occurred, arguably, in the server and in the network worlds. Storage is a hold-out.
I was reminded about this a couple of weeks ago as I reviewed e-mailed press releases and editorials about the Storage Networking Industry Association’s Storage Management Initiative Specification (SMI-S) adoption “success” and “adoption” by the International Standards Organization (ISO). SMI-S is a method developed at SNIA for passing information about storage components and their status from devices equipped with an SMI-S provider to a monitoring and management console.
On paper, it looks pretty good. In the real world, however, it is somewhat lacking.
I just returned from a site where the consumer had made SMI-S compliance a criterion in the purchase of its storage infrastructure components—something that might happen more frequently now that ISO has approved the spec. EMC had done a fairly effective job of implementing a provider on its DMX and Clarion gear that the IT manager had on the floor. However, Hitachi Data Systems required the purchase of a Software Developer’s Toolkit (SDK), at additional cost, to make its gear manageable via the technique.
When the company made all of the necessary purchases, they discovered that providers were not equal and that the visibility afforded by the provider on the EMC wares into the disposition of that equipment was very different in terms of granularity than what was provided by the SMI-S hook in his HDS hardware.
From the storage management software companies I have spoken with recently, that variability is exactly why SMI-S does nothing to threaten their value proposition as management utilities. It simply has not been implemented uniformly across any gear, so storage management is not yet a standards-based commodity play.
You might take a different view if you just read the hyperbole-filled commentary from SNIA. Its current chairman boasts that the SMI standard has achieved market preeminence because providers have now been implemented on 450 storage products. Considering that nearly 11,000 new storage products were introduced to the market in 2006 alone, the meaningfulness of this assertion is dubious. Moreover, the uneven implementation of SMI providers on gear that have done the work further calls the value of the standard into question.
I spoke recently about this with one reseller who told me he was hopeful that a “herd mentality” would drive sales of his SMI-S based storage solutions. Now that the SMI standard had been ratified by ISO, he believes that it will become a checkbox item for companies that have embraced ISO standards whole cloth, whether or not it actually delivers any value at all.
Frankly, I am starting to question the efficacy of standards in guiding the acquisition of storage products. I have put out feelers to the industry to obtain input on this subject from different perspectives. In the next column, I will provide a summary of these interviews. Until then, your comments are welcome: [email protected]