In-Depth

Self-Aware Storage: Innovation or Retrenchment?

The concept is sweet and seductive, but is this just the same old proprietary engineering game?

• By Jon William Toigo
• 04/10/2003

Recurrent themes I have been hearing are “complexity equals cost” and “self managing” equipment. There is also a lot of messaging around something called “auto-provisioning.” Let me give you some definitions before the trade press becomes totally inundated with these messages and various interpretations make them even less meaningful than they already are.

"Complexity equals cost" became a recurrent theme at this year’s Forbes/Gilder Storewidth Conference in Southern California a few weeks ago. George Gilder and new partner Forbes Magazine tried to breathe new life into this flagging conference, which was a showcase for “start-ups with promise” only a couple of years ago. Back then, the message was “disruptive technology,” and almost every vendor went to great pains to explain why its latest (or first) market entry was going to revolutionize everything.

The conference fell on hard times last year when the economy went sour and no one much wanted to disrupt his or her environment. That lesson learned, vendors began singing a different tune this year with “complexity equals cost.”

By the end of the first day of the two-day event, it was clear that times had changed and that the blush was off the technology rose. Many presentations were not delivered by start-ups at all, but by players such as Veritas, IBM, and HP. It wasn’t the same old “if you build it, they will come” conference of a few years back.

IBM/Tivoli made an interesting presentation. They did not seek to dazzle anyone with their storage resource management software, but instead communicated a distinct impression of a company that had seen the battle for universal management of the storage domain and had concluded that, for now at least, we were pretty much stuck with the blunt tools that we have. Neither CIM nor any other panacea management solution was just around the corner. The not-terribly-subtle subtext was that vendor infighting and proprietary product architectures were introducing too much complexity for a universal management solution effort to cope with or absorb. Even Veritas’ spokespersons sounded battle-weary.

In the telephone briefings I have had with vendors before and since the show, a second theme has emerged: “self-managing hardware.” All excited, various vendors of tape libraries and disk arrays have informed me that their next products would have mini-operating systems in silicon that would monitor and manage their internal components while sensing the status of and adjusting their operation to the condition of the network, fabric or other interconnect that linked them to servers.

It seemed at first a logical next step in platform architecture. Make storage products self-aware. Let them help out with the management burden they were introducing in an already overly complex topology that the SAN crowd had convinced large enterprises to buy into over the past three years. Instead of one more platform for already overextended IT staff to manage, make the storage device a capable and intelligent adjunct to the IT staff’s effort.

The message is sweet and seductive. Kind of like the soft, breathy tones that emanated from the Talking Barbie doll a couple of years ago, before the feminists got mad and Mattel pulled it from the store shelves: “Math is hard. Let’s go to the mall.”

Put together, the messages are: “Proprietary storage is complex. Let’s let the storage devices monitor themselves.” I admit that the idea had appeal. Kind of like the notion of auto-provisioning.

Auto-provisioning is the mantra of the Grid Computing crowd. Borrowing language from web server load balancing and High Performance Computing Center of the University of New Mexico’s experiments in superclustering and hyperclustering, vendors embracing auto-provisioning envision a time when the entire IT infrastructure will automatically provision an application with whatever it needs (CPU cycles, memory, network bandwidth, and storage capacity) whenever it needs it. When that happens, we can all go to the mall and forget the math.

In the storage realm, auto-provisioning smacks of last year’s panacea technology, virtualization. Remember when the virtualization vendors promised to aggregate storage LUNs (physical disks and/or array partitions assigned their own logical unit number) to create dynamically scaling volumes, saving storage administrators the hassle of resizing volumes the old fashioned way (through the operating system)? If you do, you will remember that this proved to be a lot more challenging than it looked on paper. Today, the “v” word is hardly spoken aloud in polite conversation at storage conferences.

In 2003, self-managing platforms seems to be the message that will stick. Anything that adds management to the infrastructure is good. However, the cautious side of my brain—the side that reminds me to trust no vendor—also worries about the outcome of such technology.

Will the new self-managing platforms introduce new incompatibilities between devices? Will they introduce new and sinister dependencies upon vendors—the hallmark of proprietary technology generally? Will self-managing features be used as yet another mechanism for boosting device costs well beyond the reasonable price of the underlying commoditized storage components? Will the industry try to hide this price differential behind the cloak of Total Cost of Ownership claims?

Watch this space.