AT LAST MONTH'S Storage Networking World conference in San Diego, a storage vendor told me an interesting story. It seems that one of this vendor's customers--a large and very recognizable outfit--was apparently a tad overzealous in its pursuit of a tiered storage environment and ended up with eight tiers of storage.
Now, I'm sure some storage guru in that company burned the midnight oil developing a sound justification for creating so many tiers--just the level of detail required to even think about such a complex storage environment is mind-boggling. Defining that many levels of storage and the associated data classifications is no mean trick. To then match them to storage costs and acceptable levels of risk starts to push the storage planning envelope.
Whoever put that plan together deserves a tip of the hat. And condolences, too, it seems. Because along with all those tiers of storage came the nightmarish task of actually making it work. And it didn't. Each new tier of storage means there's a potentially new set of tools to learn to manage that storage. It's not likely this eight-tier shop literally had products from eight different vendors and eight completely different tool sets, but just moving data around so many layers must have been like directing traffic in Times Square during rush hour. It's nothing short of an information lifecycle (ILM) for masochists. Good luck to those storage architects on their next project--figuring out how to fold eight tiers into a much more manageable three or four.
There was also a session on data classification at the conference. The speaker described why classification is necessary, some of the data elements related to classification and why you'll feel really good after classifying your company's data. It was a good talk, but it might have been a little off the mark for a couple of reasons. First, most storage managers know why data classification is bearing down on them like a runaway freight train--they have tons of data sitting around on expensive disks and need to figure out if it should be there. And while looking under the hood of a data classification app to see how it works might be interesting, most storage managers are still far from choosing a product. What they could probably get more use from is a list of questions to ask the business owners about how they use data, how important it is to them and to what extent they're willing to risk losing it.
Of course, data classification is at the heart of a tiered storage system and most other ILM efforts. But trying to conjure up sophisticated classification policies hardly seems to be the most ideal jumping-off point. If you dwell on solving data classification issues, ILM will probably seem more abstract--and harder--than ever.
Besides, anyone running a storage shop is already doing a fair amount of data classification. Nearly every decision that's made about stored data is an exercise in data classification: Does it need high-performance hardware? Should it be replicated? How often should it be backed up? The same goes for tiered storage. Even the most modest setup, one with a single storage system and a tape library, is a tiered environment.
So the bromide--I guess you can say the old bromide, as it's been kicking around for a couple of years now--that ILM is a process, not a product, is true. But the process is about building on what you have, not necessarily replacing parts that seem to be working just fine.
A perfect ILM system isn't necessarily defined by how granular your data classification system is or by how many tiers of storage you have. The ultimate measurement is how well the environment handles your company's information. And the key to getting there is the ability to shift the focus from looking at what's sitting on disks and tapes as data, to seeing it as information.