At the Crossroads: Reengineering Storage from the Bottom Up

How much is storage compromised by simply adding more of the same?

Among storage industry CEOs, Rob Sims, at Crossroads Systems in Austin, TX, strikes me as one of the brighter people in the group—both technically and in terms of business savvy. A knee injury laid him up last week long enough to give me an opportunity to chat with him about industry trends.

Readers of this column know that I am already favorably disposed toward Sims and company. They have transformed themselves from a provider of Fibre Channel to Parallel SCSI bridges in the early days of FC fabrics into a forward-looking provider of router-based software engines that provide discrete functions for storage infrastructure ranging from database access control to storage security key management to read/verify functions in tape backup operations.

Basically, the company’s router platform is a Swiss Army Knife that is useful in hosting important data provisioning and protection services in a much more efficient manner than hosting these services on servers, switches, or storage arrays themselves.

Sharing some of my conversation with Rob supports my recognition of him as one of the storage industry’s top CEOs in 2007, a moniker that some might not agree with given Crossroads’ status as an over-the-counter player rather than a NASDAQ company. Frankly, I think this man has vision.

Sims thinks that the time has come for companies to consider reengineering storage from the bottom up as they seek to scale to new requirements. He says, “Companies cannot afford to keep building storage using big devices that work only with themselves.” He is quick to point out that some applications might need big access systems, but that this is certainly not the case for all applications.

Read that last paragraph again, closely. It is a loud shot fired across the bows of the big vendors in the industry who Sims charges with “adding to their own structures without listening to their customers.” He says that selling “double-speed drives and shiny new robots” will be less strategic for customers going forward than “abstracting layers of control and management” that will enable consumers to “dial in the right kinds of services as required by the data itself.”

The problem with storage monoliths goes beyond acquisition costs, Sims argues. He paints a picture of the enterprise IT shop where IT users need to have multiple certifications for the many platforms that the company fields in its data center. “Training and certifying IT personnel in all of these [stovepipe] technologies is expensive, and it is an investment that the company loses every time an IT person changes employers to realize a two-to-three dollar-per-hour increase in his or her salary.”

Correlating Costs and Services

Sims says that this company money could be better spent by tasking IT personnel to “dial in services to meet requirements of applications and data.” As things currently stand, he argues, there is no discernable correlation between the costs of IT operations and the services that are being delivered. To align cost with service, “control must be returned to the granular level. We need to understand what’s coming across the wire, where it comes from, [and] where it is going, and we need to monitor and measure this activity so we can constantly modify system resource allocations and services according to business change.”

That is what I would call real SOA—service-oriented architecture—not the stuff I have been reading in ghost-written whitepapers and pay-per-view analyst reports from the major storage vendors. Furthermore, Sims stands behind his words. Crossroads’ SurePath ShareLoader is his company’s first foray into the architecture he envisions.

ShareLoader is an appliance, nested on the company’s vaunted router platform, that delivers data protection services in a distributed computing environment. His brochure says that ShareLoader “seamlessly integrates into existing backup environments, automatically protecting ‘edge’ data [on PC, laptop, and departmental servers]. ShareLoader delivers a simplified, sharable, and secure alternative to traditional backup and recovery by combining the benefits of disk combined with the long-term advantages of tape media for archiving and disaster recovery.”

Sims asks why companies are still hardwiring applications to equipment, rather than to services. He notes that this practice is viewed by a growing number of companies as a fact of life imposed by the Fibre Channel pipe. He cites backup as a prime example of the problems with the strategy.

“Backup applications are inefficient,” he argues, “because hosts talk directly to devices. The SCSI protocol was built on that premise. The backup software imposes Unitarian control over the process. ShareLoader is an example of an approach for breaking that mode.”

ShareLoader appliances are placed in the IP network where they each emulate the functionality of a unique drive and small library to each connected host system. Each ShareLoader emulates a traditional tape library that is sharable between 100 machines. ShareLoader accelerates restores and provides nearly immediate access to data, drastically reducing recovery time. With the use of lossless data compression, the cost of storage media is further reduced through better capacity utilization.

In short, ShareLoader is an example of how a service might be delivered in the network infrastructure that is both highly manageable and flexible in the face of demand change. It is, in Sims view, a low-cost investment that blows the socks off of expanding a $50,000 LTO library that doesn’t scale with the business.

ShareLoader is just one of a number of products coming out of Crossroads that solve different problems by providing centrally managed service adjuncts to monolithic or fabric-based storage. Sims is on the right path as he seeks to leverage his core router platform to solve a lot of one-off problems in the contemporary storage infrastructure. Whether deliberately or inadvertently, he has also discovered an innovative method for building storage that might save companies a lot of money.

His biggest challenge will be to compete for customer mindshare against powerful vendor marketing campaigns. At a time when storage array makers seek to prop up the margins on their arrays by adding proprietary value-add features and functions to otherwise commodity boxes of disk drives, Sims is gently nudging consumers to consider an alternative: to deconstruct monolithic arrays and to deploy services on a centrally-managed-but-point-basis so they can be shared more effectively and applied more intelligently in support of application demands.

He says it more simply, “Just follow the right thing for your business.” We couldn’t agree more.

Your comments on ShareLoader or Sims’ view of storage architecture are welcome: