In-Depth

SMI: Will The Spec Ever Get Respect?

The SMI spec offers considerable benefits: improved efficiency and potentially huge management cost savings as well as improved troubleshooting capabilities. Yet SMI's poor outreach to end users has been a serious blunder. What the SMI should do next.

• By Jon William Toigo
• 05/13/2004

A reader in Belgium e-mailed me in response to my recent columns on the Storage Management Initiative-Specification (SMI-S) to offer his two Euros. He wrote that, in his view, SMI-S and its core Common Information Model (CIM) are “critically important” to solving the storage management issues frequently raised in this column. He couldn’t understand why we weren’t more positive in our treatment of the technology.

My response is, quite simply, that SMI needs more folks like this reader. Over the past three years of writing columns for Enterprise Systems, I have dedicated a lot of ink to SMI-S and CIM—at least nine columns by a cursory count. Some of these have criticized the hype around CIM and the process by which the spec was being developed, but I have always endorsed the concept itself.

Version 1 of the SMI spec was officially announced in early April at Storage Networking World in Phoenix, AZ. It was, in fact, the only real news of that particular show. And, I must admit that I was awestruck by the unprecedented level of cooperation shown by the vendor community in developing the specification.

Version 1 isn’t a complete job, but it’s a start—and a good one—that describes how SMI can serve as middleware between storage devices and management software and the storage administrator, providing a way for storage consumers to simplify and streamline storage capacity management. What you don’t get from the document, and shouldn’t, is a sense of the political machinations that were required to pull it off.

For one thing, some overly enthusiastic advocates hyped CIM early on as a panacea and claimed it could do things that it clearly couldn’t, doing more damage than good. These “disruptive revolutionaries” were reined in, but not before they alienated many vendors from participating in the SMI-S process.

Secondly, using SNIA as the arena for developing the spec was problematic at best. A lot of software folks see SNIA as a hardware vendor-dominated organization and treat the organization as an anathema. Even Microsoft has kept the SMI-S development effort at arm’s length.

Thirdly, the engineers involved in SMI have focused first on cultivating the support of hardware vendors and are now turning their attention to software vendors in an effort to get CIM “providers” built into storage products entering the market. This bias has been at the expense of the effective cultivation of end-user appreciation of the business value of the technology—a huge mistake in our view, but again a reflection of SNIA’s vendor, rather than consumer, focus.Why is it a huge mistake to ignore end users? Simply put, if there is no grassroots support for SMI-S, there is ultimately little incentive for vendors to build CIM providers into their gear. We have received a lot of pushback on this point from SMI backers, who cite the labor savings value to vendors if CIM support becomes real. That said, a common reason cited by vendors for their failure to add providers to their products is that no one in the consumer world is clamoring to buy it—and certainly no one seems willing to pay extra for it.

SMI advocates acknowledge that (1) a generally lousy job has been done to educate the end user about the business value of SMI and CIM, and (2) the best that can be hoped for is that SMI-S will become a “check-off” item in the selection criteria used by consumers in making product selections.

On the first point, I earlier noted the poor job done by SNIA in SMI education. I am told that they are trying to get their act together to fix this deficit. Frankly speaking, educational sessions around CIM from SNIA have been downright boring. The combination of PowerPoint eye charts and monotone explanations of object-oriented programming and meta-modeling techniques had the tendency to suck the life force from even the most starry-eyed CIM acolytes. Without a significant improvement in consumer outreach, I seriously doubt whether CIM/SMI will achieve even the status of a “check-off” item in consumer product selection.

“Check-off” is a code phrase, by the way, that communicates both a subtle wisdom and a warning. Translated from SMI-speak, it means that vendors should not expect consumers to pay extra for CIM providers that they engineer into their products. The subtle warning part is that CIM support is something that vendor products will need to have, or consumers won’t buy their products—period. SMI advocates are saying that CIM-based management is somehow inevitable and everyone had better join in or be left behind.

Vendors who haven’t jumped on the SMI bandwagon, or who became discouraged with the SNIA process and jumped off the wagon, don’t appear to be intimidated by this line of reasoning. It is less a matter of inevitability than a simple case of cost/benefit analysis: SMI-S may be the best thing since sliced bread, but it won’t be executed if consumers don’t demand it.

You would probably see SMI executed by everyone if it had the enthusiastic support of end users, and to some extent, of Microsoft. With the SNIA trying to figure out how to sell itself as both a vendor pitchman and a user ombudsman without sounding like someone peddling a floor wax that is also a desert topping, end user support is slow in coming.

Bottom line: There are obvious advantages to an SMI-instrumented storage environment that go to improved efficiency and potentially huge management cost savings as well as improved troubleshooting capabilities that accrue to the SMI view of storage. We believe in the spec, but we are dubious of SNIA’s ability to sell it to consumers. Perhaps the SMI crowd needs to look outside SNIA to build the end-user momentum they need to succeed.

Let me know what you think: [email protected]