I'm tired of hearing that information lifecycle management (ILM) saves money. In almost every piece of storage literature that old song is replayed. The lyrics say chopping a large homogenous pool of storage into many smaller pools of disparate technology saves money. Storage vendors and IT media have sufficiently hyped the use of ILM to control storage costs, without much discussion about the consequences of doing it inappropriately. For several years now we VARs have been feeding off the frenzy and rushing to the scene, advising eager customers to deploy five tiers of storage so that they can cash in the on savings. The truth is that if done incorrectly, tiered storage can actually increase costs. Lets discuss several ways that storage tiering can go wrong.
Too few storage tiers
Ok, I agree; running the entire enterprise on the most expensive disk will drive costs well above what's necessary. It may also be true that running critical applications on a single storage tier of unprotected SATA JBOD could spell disaster. The bottom line is that incorrectly aligning business applications to technology can waste money or put the business at risk. Obviously some storage tiering is required; the trick is not to go overboard.
Too many storage tiers
Storage white space is expensive but necessary. One of the most challenging truisms facing storage managers is that it is never acceptable to run out of disk. A server can operate at 100% CPU utilization and network pipes can be full without spelling disaster, but when an Oracle database runs out of disk space, things come to a screeching halt. Storage managers must keep free (white) space in reserve for growth and surprise projects. Each storage tier on the floor requires its own pool of free capacity for safety margin; and more tiers spell excess waste in the form of unused disk. See my storage inventory management tip for help.
Keep data moving
Migrations take time and time is money. What benefit is ILM without the ability to promote and demote application data to the appropriate tier of storage? Applications naturally become misaligned with their storage over time for several reasons: Data loses importance and business priorities change, pulling applications in and out of the spotlight. Keeping the storage infrastructure in line with these changes means that data must move. Even with virtualization, migrations take effort. The quantity of migrations required will be proportional to the number of tiers installed. Fewer tiers mean fewer migrations, less time and lower cost.
Best-of-breed challenges
Supporting heterogeneous technology is hard. Deploying an effective ILM strategy allows customers to choose best-of-breed technologies in each tier, perfectly balancing their portfolio of risk and cost. The problem is that each distinct technology will require new skills and procedures, tools and integration points. The added complexity is difficult to quantify in hard dollars but certainly dilutes the cost savings derived from putting old data on cheap disk.
How many storage tiers are enough?
I wish there were a simple heuristic to establish the optimal number of tiers for a given client. But like most things the answer is subjective. A few simple guidelines will help find the best fit for a client.
Generally speaking, the more volatile the growth is, the more challenging it will be to get real rewards from a highly tiered infrastructure. Fast-growing shops should focus on reducing change rate and maintaining stability despite the rapid growth. Help the customer understand that, while everything else seems out of control, keeping things simple is the best course of action. For the fastest growing data centers, two tiers of storage are usually sufficient.
Once the growth rate is arrested, and the organization's focus turns to sharpening the saw, more tiers may be in order. Up to five tiers may be appropriate only if the customer isn't outgrowing storage arrays before they depreciate, or the business changes are slow enough that data can remain on one tier for years at a time.
The trick to real value-added reselling is to be responsible and consider the true needs of the customer. Negative consequences may result from "one-size-fits-all" methodologies. Crafting an appropriate solution that accounts for the customer environment will reward everyone.
About the author: Brian Peterson is an independent IT Infrastructure Analyst. He has a deep background in enterprise storage and open systems computing platforms. A recognized expert in his field, he held positions of great responsibility on both the supplier and customer sides of IT.
