Last time, we discussed the problem of assigning value to data within environments that run according to service-level agreements. We identified two countervailing forces: the first, that every manager's data is critical - to the manager at least; and the second, that in an IT environment of limited assets, budgets, personnel and so forth, it is in everybody's best interests to assign value to data relative to the rest of the data in the enterprise, and to do this before the SLA is drawn up. In this way, data resources can be assigned based on corporate priorities, and IT doesn't waste cycles servicing unnecessary files while valuable data sets languish.
We should all agree that SLAs are a good thing, but only when they are implemented in a sensible fashion. Of course all my readers are sensible, so clearly there is no need to belabor this point. (If however, you are sensible but still not up to speed on SLAs, take a look at http://www.slm-info.com/, a newly inaugurated site sponsored by Enterprise Management Associates focusing on service-level management. This month it is offering a free download of a paper by Elliott Kass, director of marketing at iCan.)
Alas, enlightened though my readers are, they have friends in the business who are laggards when it comes to defining service levels. With no SLAs to help their site define the relative value of data, what mechanism do these poor souls use? As it turns out, they typically do one of three things. And these are all bad things.
When the managers at these sites have to value data, the major decision mechanism used to define how valuable a file is often has something to do with data aging. This idea that data changes in value over time has some validity: as data gets older, its value usually does change. The problem is with the basic assumption that aging data gets stale, diminishes in value, and should therefore be moved down to lower performing or lower availability storage.
It's not just my graying hair that makes me think that what gets older doesn't necessarily get less useful.
Truth be told, most data does get less valuable, but not necessarily all of it.
The key concept here is that data does not change in value because of time, but rather because of how it is being used. It is not a time-value relationship, but rather is a function of usage and value.
Here is an example of what I mean. Monthly sales reports for example, certainly would seem to diminish is usefulness after the end of the month. At some companies however, while there is a change in how the data is used it may be questionable in which direction the value is changed. Groups that use this data for trending still receive great worth, and may in fact be producing research that shapes the future of their companies. The significant point here is not that the data has become less useful, but rather that the way it is used - and the way it is accessed - has changed.
Those sites that do move data according to its age should at least apply rules of information lifecycle management (ILM). ILM helps migrate data from device to device, going from high availability devices to progressively less expensive environments as the data becomes less useful. In this case however, data movement is governed by policies which management sets, and not by the near-arbitrary method of assigning a value to time.
If ILM is of interest, you may want to check out the offerings from Legato, Mercury Interactive and StorageTek, among others.