In-Depth

Storage Management: Headed for the Switch

Sun's must-read white paper articulates its vision of switch-based storage management as well as anyone could today. The question is, can we get there from here?

• By Jon William Toigo
• 07/24/2003

A name you haven’t heard mentioned very often in this column is Sun Microsystems. The darling of the Internet Era, Sun’s Web server business, together with its stock value, plummeted when the dotcom bomb exploded at the end of the 1990s. Truth be told, most industry watchers stopped watching Sun when they tried unsuccessfully to 1) introduce their fifth high-end array to market in as many years back in 2000 (a re-branded platform from MaxStrat, the high-end array manufacturer they acquired in 1999), and 2) make the world over in Java and Jini as part of Project StoreX (a strategy later abandoned)—all to no avail.

One thing that has kept Sun in the storage game for the past couple of years has been its leadership in developing the Common Information Model (CIM) within the Storage Networking Industry Association (SNIA). The company can also be singled out as the first industry player to incorporate CIM support into its own storage management solution.

Beyond this, Sun has stayed pretty much under the radar. While everyone else was loudly pursuing SAN-this and NAS-that, the core development effort at Sun appears to have been the improvement of storage management capabilities and their integration directly into the operating environment software stack. The company’s colorful CEO, Scott McNealy, has been quoted to remark that if 90 percent of customers need a function, it should be part of the operating system.

Apparently, this rationale has driven Sun to add fabric multi-pathing, snapshot support, and remote replication directly into its Solaris operating environment stack. Archrival Microsoft appears to be following suit by adding lots of storage bells and whistles to both the current and the next generation of its Windows server operating systems. If this trend were to continue, a lot of the software currently purchased from third parties would be unnecessary.

Interestingly, however, despite Sun’s pragmatic work on enhanced storage-related OS functionality and a strong, on-going development effort in the midrange storage space, the company still has the heart of a pioneer and visionary. If you need proof, take a couple of minutes to read their white paper on the Intelligent Storage Platform at http://www.sun.com/storage/white-papers/intelligent_storage_platform.html. The company eloquently states what I believe most of the industry has already concluded. Storage management is eventually going to be located on the switch.

Perhaps folks at the company remember when early networks used Sun server boxes, running software applications, as routers or firewalls. These functions were eventually integrated directly into switches. According to the white paper, this is what Sun thinks will happen to many of the single-purpose tin-wrapped storage software appliances that are proliferating in the market today.

It seems that everyone with a bright idea is taking their one bright idea, wrapping it in a commodity Linux server, and selling it as a storage appliance. While the strategy has the merit of showing off the capabilities of the software that inventors, usually newbies in the business, would otherwise need to describe less convincingly with a PowerPoint slide deck, it makes for a messy environment. Sun sees a time in the near future when the best of these appliances will be integrated into a large, multi-gigabit cross-bar switch augmented with management processing engines. Not surprisingly, the logo on this intelligent storage switch platform in one white paper illustration is Sun’s.

If you are wondering what all this talk about the “intelligent core of the storage infrastructure” really means, Sun gives a clear statement of the vision and I encourage you to read it—whether or not you are a Sun shop. Reading the paper also gives you a surprisingly accurate, if somewhat sugar-coated, overview of what’s wrong with the management of storage area networks today. The author describes the current situation in order to contrast it with the visionary future and his critique is dead on. What the author understates is the enormity of the challenge in getting from here to there.

The subtext of the story is that the future intelligent storage infrastructure will build on the best of the hodgepodge of technologies that are in their experimental stages today. In an enormous leap of faith, the reader is expected to assume that consolidation of storage management functionality around a common switch platform is inevitable, just as it was in the TCP/IP networking world.

The reality is that unlike the Ethernet and TCP/IP experience, there are no common standards for storage or any other point of reference on which all vendors must fundamentally agree or with which all of their products must comply. The record in storage has been one best characterized as “every man for himself.” As intelligence is added to switches, there is absolutely no guarantee that Brocade’s intelligent switch and Cisco’s intelligent switch will be able to co-exist and interoperate in the same network to any greater degree than each vendor’s not-so-smart switches do today—which is to say, hardly at all.

Some point to the T-11.5 subcommittee at ANSI as the harbinger of greater standardization of switches, but Cisco and Andiamo, failing to get their VSAN technology the desired nod at T-11, are currently trying to do an end run at IETF’s IP Storage Working Group. In June, they submitted a Request for Comment to try to initiate standards development on Fibre Channel Management Information Bases (MIBs) that mostly pertain to managing VSAN technology, which is not a Fibre Channel standard, but with standard MIBs might appear to be one.

Bottom line: Sun has articulated the vision of switch-based storage management as well as anyone could today. In so doing, they have underscored the problems that plague FC SANs—problems that most vendors in the industry prefer to claim do not exist. The only problem with the Sun vision is that they underestimate the hurdles that will need to be surmounted before the intelligent storage infrastructure becomes a reality. It doesn’t mean that it exists just because more and more vendors talk about it as though it is already here.