Columns

The Struggle for Interoperability

The storage industry is struggling with its toughest challenge—agreeing to standards. Your best bet as a consumer may lie in supporting CIM.

If you're a history buff, you might know that the founders of our country's first railways created havoc by introducing varying rail widths, or gauges, in their respective systems. Trains in the Northern states used rails set at a different width than those in the South; western trains later used still another, narrower gauge.

Until the situation was resolved, railroads developed some pretty kludgy workarounds—like lifting an entire train car chassis off its wheels and setting it down on a set of wheels fitting the other gauge rail, or creating trains with adjustable wheel bases. It was messy, time-consuming and bothersome.

See any similarities here with the enterprise storage management plans of leading vendors?

In my January column, I discussed one of those plans: EMC's WideSky. (See "A Kinder, Gentler EMC?") In that column, I described how in mid-October 2001, EMC made a bid to enter the lucrative enterprise storage management software market by "opening up" its proprietary storage management software called ControlCenter (which EMC provided with its own hardware) for use in managing third-party hardware platforms. The solution featured a middleware layer called WideSky that EMC said provided an "open standard" for storage management integration.

In other words, WideSky provided a common way for any hardware or software vendor wanting its products to be managed via "ControlCenter: Open Edition" to interface with the EMC solution. The vendor needed only to write its device driver to work with the WideSky middleware.

At the time, it seemed pretty obvious that EMC was making a Microsoft-like play to become the de facto standard for storage management. Implicit in the initiative's description was an airy promise to software vendors that their support for WideSky would give them traction with EMC's large number of hardware customers. That seemed like quite a carrot and one that might reverse the time-honored dictum of storage management frameworks—that software vendors are beholden to hardware vendors.

Slaves to the Industry
Storage management software developers will likely agree with Veritas Software's Roger Reich, the company's senior technology director for interface standards, who says his primary challenge "is keeping device support in [their] product up-to-date with the 17,000 disparate devices on the market." Doing so involves an expensive and time-consuming process of cultivating relationships with each and every hardware vendor. That means approaching them hat-in-hand and requesting (or in some cases, out-and-out begging for) access to the vendor's proprietary application programming interface (API), command line interface (CLI), or some "hook" into the vendor's platform that will enable the creation of a specialized device driver or agent for that platform.

Reich says that Veritas is no different than any other vendor in this respect: "We are a slave to the industry."

As a hardware vendor with many customers, EMC took a different tack with WideSky, hoping that the mountain would come to Mohammed. It hasn't.

With the predictable exception of a few of EMC's own cadre of partners, the WideSky play has attracted little interest from EMC competitors. In fact, the response from the hardware side of the industry has been widespread rejection of EMC's "standards bid."

That isn't to say, however, that the industry has come up with an objective standard in place of EMC's offering. Sure, many name-brand hardware manufacturers have recently given verbal support to an evolving object-oriented management method, Common Information Model/Web-Based Enterprise Management (CIM/WBEM), in development at the Storage Networking Industry Association. A few—including Veritas—have sought to promote their own proprietary framework.

Veritas' Adaptive Storage Architecture (ASA), announced last May at the company's Vision conference, looks a lot like EMC's AutoIS when you diagram it.

Imagine a train track with one rail representing clients and the other representing devices, and the ties in between representing middleware to bind the two rails together.

On the device "rail," ASA and AutoIS are virtually indistinguishable. All third-party storage software and hardware, as well as CIM- controllable devices, are potential candidates for management via the vendor frameworks.

ASA and AutoIS look a lot alike in their middleware functionality. The only clear difference is that EMC's has just one client: ControlCenter. All devices integrate through WideSky with EMC's ControlCenter. The Veritas scheme, however, depicts a range of clients: Some are Veritas-branded products, or OEM interfaces it has marketed to other hardware providers. However, the scheme also indicates support for CIM/WBEM "clients" (called "servers" in CIM lingo).

While both vendors have publicly embraced CIM/WBEM, which in the best of circumstances would negate the value of both companies' proprietary offerings, they continue to sing the same tune. According to Reich, "It's a chicken and egg debate. We still offer proprietary APIs for interfacing products to our management clients. We will change over to CIM when the market demands it and the industry supports it on products."

Figure 1
Figure 1: The EMC approach to interoperability shown here envisions a storage management console such as the company's own ControlCenter. In that scheme, EMC's Open Edition controls and monitors storage hardware and software (from EMC and other vendors) through a middleware layer called WideSky. Companies that interface their products to WideSky, via application programming interfaces and other software "hooks," hope that cooperating with the EMC approach will give its products an edge in IT shops where EMC technology predominates.

CIM: Ready or Not?
Reich is sensitive to criticism that Veritas isn't making bold moves to CIM. He notes that, contrary to public perception, his company isn't the "800-pound gorilla" it's made out to be. "We're more like the two-pound gorilla. We're as much a hostage to vendors as anyone else, and we have to negotiate with the vendors."

Reich says that for the past two years, Veritas, as a co-founder of SNIA, used its influence to "lock the big guys together in a room" to reach a standard specification for device discovery in a heterogeneous Fibre Channel SAN. He cites the recently announced "Bluefin" standard specification as proof that Veritas is working toward CIM's goals. Bluefin provides a means for storage devices from competing platform vendors to discover and interact with each other in the same fabric. It's based on CIM and extends the SNIA model.

While the argument could be made that vendors, sensing the reluctance of customers to deploy homogeneous SANs, would have gotten around to a Bluefin-like standard in any case, Reich casts the outcome in more triumphant tones. Bluefin shows that Veritas is solidly behind CIM, he says, adding, "If the device vendors would start showing up at our door with a CIM Managed Object Format (MOF) file, it would make our lives a lot easier."

Mark Carlson, a storage network architect at Sun Microsystems Inc., and co-chair of the CIM/WBEM working group at SNIA, understands the plight of Veritas and EMC with respect to CIM. Clearly, says Carlson, a standardized MOF would contribute to universal storage management. It would also simplify the work of developers who are "spending a lot of time and resources creating adapters."

These advantages, however, might be offset by lost sales of proprietary products the vendors already offer. "It's going to take some risk-taking for established vendors to embrace CIM. I suspect that start-ups will beat established vendors to the punch, since they have less to lose in existing markets."

"If consumers are waiting for EMC or Veritas to do CIM [in their storage management software products]," Carlson says, "they will probably wait for at least two years. I know of several start-ups that are almost ready to come to market with CIM-based management products." When I asked him to name one, he demurred. "Most of them are in stealth mode right now. But I will say that customers who are looking for something to solve the management problem today need to be willing to try out these new CIM-based management tools when they're announced."

Figure 2
Figure 2: CIM/WBEM relies on a common object naming schema that can be customized by individual vendors. A neutral client console is forthcoming from SNIA.

Carlson says that he's looking forward to the next CIM demo at the Fall Storage Networking World trade show, which he says will provide a compelling demonstration of how far CIM has come. He says that Sun, Hewlett-Packard, IBM and Hitachi Data Systems have all promised to show platforms instrumented for CIM/WBEM-based management.

Though available product roadmaps from the vendors named by Carlson offer no indication of whether, or when, CIM support will be provided, he says he is "sure that CIM support is bundled into Sun's Storage One initiative." He adds that if there is any dearth of CIM-ready devices at the demo, "it will be because we require that the product must already be shipping."

He further expects the demo to provide another look at a new and improved "neutral client" from SNIA: An open-source console for managing storage using CIM. Two years ago, then start-up Tek-Tools offered its pre-release Storage Profiler management console for use as a client console in the Spring SNW CIM Demo. This spring, however, Tek-Tools was displaced by a neutral client "because Storage Profiler has begun shipping in the market and is therefore considered a proprietary product." Carlson says such decisions reflect a desire to demonstrate the value of CIM without linking it to any specific product in the market today.

Carlson's dedication to "openness" is reflected in his view of Simple Network Management Protocol (SNMP), an existing network management standard that's touted by iSCSI advocates as a natural fit for IP SANs. "SNMP is a valuable tool," he says, "but it only gives you status information. It doesn't give you any way to automate responses to error conditions that it reports. Going forward, most vendors should support both a CIM agent and an SNMP agent. Writing a CIM wrapper around an SNMP management information base (MIB) will give you the best of both worlds."

The development of CIM appears to be taking on momentum. Carlson says the next step will be to create a management forum at SNIA and to raise money to "push CIM into the customer consciousness." Customers, he says, "need to understand its value. They need to embrace CIM, then buy only those products that support it."

Must Read Articles