A standard that leaves out the good stuff?

Vendors tend to offer the lowest common denominator of functionality

A specification meant to bring together storage device management under one umbrella instead continues to be criticized for its innate inability to enable interoperable functionality.

The Storage Management Initiative Specification (SMI-S) is a lightning rod for controversy among both its supporters and detractors. It has been approved by the ISO international standards organization and the International Electrotechnical Commission (IEC) as a standard for interoperable storage management technologies. Yet detractors say SMI-S was overhyped and will never fulfill its original promise of integrating storage device management across the industry. Supporters continue to say SMI-S is opening storage resource management (SRM) doors that were previously locked by proprietary vendors.

SMI-S, now in Version 1.02, grew out of the Storage Management Initiative (SMI), which was created by the Storage Networking Industry Association (SNIA) in 2002. SMI-S is criticized for not being more open to smaller SRM vendors and failing to help end user IT organizations establish centralized control over their heterogeneous storage infrastructures because larger vendors will not port the most sophisticated application programming interfaces (API) to the standard's framework.

That criticism is countered by SNIA and many large members of the vendor trade group who say that even though SMI-S implementations have lagged behind the rapidly evolving standard, both vendors and users are benefiting from it.

The basic concept behind SMI-S is simple: Vendors translate product specifications into XML code for their specific storage management software, which creates a universal API that can be accessed by any other SMI-S-compatible products. SNIA says that about 450 products from 30 vendors have so far passed the SNIA Conformance Testing Program for SMI-S. Product types include storage networking components (such as arrays, switches and host bus adapters) and their associated management software, as well as client software.

Slowing down the development process?

Marty LeFebvre, Nielsen Media Research's vice president of technology and a member of the SNIA End-User Council governing board, says the problem with SMI-S is broad and one of vendor reluctance for interoperability.

"There's a certain amount of feeling among end users that if we had a more direct conduit to [vendor] product managers, we might be able to make more progress toward that," he says. "But part of it also is [that] instead of each vendor trying to solve the whole thing front to back, if there was a little of collaboration to where individual vendors could solve a pieces of it and then there was interoperability between the software, just as we're talking about interoperability between hardware, that might help advance that."

LeFabvre said when end users talk about SRM, it encompasses a broad range of functionality, such as provisioning, capacity planning, automated response to performance issues and load balancing. "To try to solve that front to back is a tall order," he recognizes.

(Read Computerworld's entire Q&A with LeFabvre.)

Jon William Toigo, chairman of the Data Management Institute and CEO of Toigo Partners International LLC, is a frequent critic of SNIA and its SMI-S standard. He refers to SMI-S as a "golden dream" that will never come to pass and says that it is only being implemented on a small percentage of the 15,000 new storage systems released in 2006. According to Toigo, the unwillingness of vendors to share their most sensitive APIs is a major stumbling block to SMI-S.

Toigo also believes that smaller storage management software vendors in particular are feeling the lack of cross-platform functionality as they try to plug their products into larger vendors' consoles. Other industry observers have said this problem is so bad that up to 50% of the capacity-related fields in the SMI-S framework are empty because vendors have not filled them.

"That's absolutely true," Toigo says. "They tried to come up with a much more complicated version of the Storage Management Initiative, and major vendors balked because they did not want to reveal that much of the internal goings-on inside their platforms, so what they did was create a generic, very inconsequential amount of status monitoring that would go into every SMI implementation and then opened the door to do customized additional builds."

Bob Laliberte, an analyst at Enterprise Strategy Group Inc., agrees that SMI-S has put restrictions on smaller SRM vendors, saying the situation slows down the development process and forces them to only support one or two vendors to start. They are continuing to add more vendors as they get access to APIs and test equipment, he says.

"But this same challenge also becomes a competitive differentiation for these companies as they grow and add new devices," Laliberte declares. "Other independent software companies like Symantec and Tek-Tools that have been around longer and have broader support use this to their advantage."

Smaller vendors weigh in

Rick Clark is president and CEO of Aptare Inc., a private vendor that sells a Web-based storage monitoring tool, says that four or five years ago, many SRM vendors were hoping for technology that would provide homogeneous, centralized management, but even in the wake of SMI-S, "monolithic SRM is dead." Clark blames that situation on the inability of SRM proponents to create anything beyond a very thin veneer over the top of heterogeneous storage systems.

As a result of that situation, the storage industry looked to large vendors in the hopes that they would open up their proprietary APIs. The large vendors were reluctant to comply because doing so would undercut their bottom lines, Clark says. "If you open up and expose the same interfaces that you're using to manage your arrays," he notes, "all of a sudden, your revenue associated with your management software are going to diminish significantly, and all of a sudden, your competitors are going to start managing your arrays."

Bryan Semple, vice president of marketing at Onaro Inc., a software vendor that extends data center automation to storage, says that most customers -- even if they have single-vendor storage environments -- don't want to rely on single-pane provisioning because most of it is still script-based and they don't trust it. That being the case, he says that he doesn't know if SMI-S will get those customers past that point of reluctance.

Nonetheless, Semple states, "We're big fans of SMI-S here because it will make our lives easier." As he explains it, Onaro "interrogates" storage switches and arrays in real time to come up with the most accurate, heterogeneous picture of the services delivered by the storage environment back to an application. Toward that goal, Onaro's products use a variety of interfaces, including command-line interface, HTTP, SNMP and, where possible, SMIS.

"The broader SMI-S gets adopted, and the more robust the standard becomes, the less we will need to maintain these other interface mechanisms into the server and switch arrays," Semple says.

SNIA chief says improvements coming

Vincent Franceschini, chairman of SNIA, says that the problems experienced by smaller SRM vendors who attempt to plug into SMI-S-compliant consoles of larger vendors are well-known to SNIA, adding it is important to note that the leading storage vendors have invested for years in proprietary interface technologies, and when SMI-S is not ready to fulfill all the management requirements, alternatives are developed with those vendors' unique approaches.

"Smaller and/or new storage software players don't have to deal with such issues and would love to see SMI-S fields utilized consistently across all platforms," Franceschini says. "The SNIA Storage Management Initiative is attempting to gradually improve this situation by prioritizing so the SMI-S issues can be fixed with feedback from developers, product managers and IT users. SNIA SMI is also looking at the introduction of vendor-unique elements in SSM-S in future versions -- possibly Version 1.4. This open access to management capabilities across platforms is also an important topic at SNIA."

As a former two-time SNIA chairman, Wayne Adams -- who is currently senior technologist and director of standards at EMC Corp. -- is intimately familiar with SMI-S. According to him, SMI-S is not a magical key to SRM success. "It's just an enabling technology to get your information from the devices, so the value-add or secret sauce of storage management still comes from vendor innovation," he says. Take away that proprietary innovation, he adds, and vendors lose their unique competitive advantages.

According to Adams, part of the problem with SMI-S can be attributed to the gap between the relatively rapid writing of new SMI-S versions and the slower development of the enabling technology. "The development has been working at a conservative level to ensure that whatever has been defined is implemented and tends to be more bug-free than just implementing something because that's what's been written this time around," he says. "So I would say in a rearview-mirror perspective, in some feature sets, it may be two cycles of the standard before there are enough vendors who have implemented it and deem it to be stable."

Regarding the frustration of smaller SRM vendors who find it difficult to plug their products into larger vendors' SMI-S-compatible consoles, Adams calls it a "vendor-by-vendor adoption decision," saying he will not try to "sugarcoat" the fact that the smaller vendors lack the installed base necessary to create leverage with their larger counterparts.

Hey, EMC supports it

Adams says that EMC has been a leader in SMI-S implementation, noting that the company has been among the first vendors to implement the most recent versions of the specification. That has included using SMI-S to enable performance metrics and some of the new block interfaces that manage larger devices through block services. As evidence of EMC's proactive attitude toward SMI-S, he pointed to the company's April announcement of v1.2 support for 36 models of its Symmetrix and Clariion storage arrays. "I would say that we've been out there at least at the 95% level saying this is what's ready to be tested and demonstrated, and then we have implemented it."

So far, SMI-S has not been viewed as a direct aid to users, although Laliberte says that on a scale of 1 to 10, he would give it a 5. According to him, many companies are providing SMI-S support, which provides a baseline of support for a variety of different vendors. This, Laliberte says, creates some visibility into heterogeneous devices and accelerates the introduction of more products that can work cohesively in a data center with multivendor storage devices.

Laliberte elaborates by stating, "The ability to leverage SMI-S to accelerate getting products to market will provide more competition and help drive innovation because companies can spend more time developing advanced analytics or automation instead of support for different arrays and switches."

Franceschini acknowledges that SNIA has not gathered a lot of information relating to how SMI-S has helped IT shops in user organizations gain more control of the heterogeneous storage infrastructures. According to him, SNIA intends to improve its industry feedback in this area, adding that SNIA's requirements-gathering process for future SMI-S versions already involves product managers and the SNIA End User Council, a group that meets regularly to provide feedback and direction on storage technology.

"As SMI-S is deployed in production environments, it helps approaching the overall management requirements differently by enabling the setup of a broader framework to discover and manage heterogeneous resources," Franceschini states. "It is particularly true when introducing new resources and for getting support with storage management applications. SMI-S has allowed IT organizations to be more focused on storage management requirements."

Despite his sometimes strident criticism, even Toigo pays a complement to SMI-S, saying, "The specification itself is great. If it was fully implemented in full flower by every vendor, you would have all the information at your fingertips not only to monitor status, but also be able to handle the configuration and management of the capacity of the array or the device in question. It would be a boon to every consumer on the planet."

Copyright © 2007 IDG Communications, Inc.

Shop Tech Products at Amazon