To Tier or Not to Tier

If you read the previous post on Shedding Tiers, you already know my opinions here. Tiered storage is just one more version of HSM and ILM that is great in concept, but awful in execution.

But the day after we posted this article, I received and email proclaiming the benefits of tiered storage from IBM. I have nothing against IBM, and think they do a fine job with their overall solutions.

But I also saw this from another email newsletter:

ILM seems to be MIA, but why?

It is clear we are far from a consensus here. We'll see how it plays out...

Comments

[Tory Skyers]

Hello Gary, I just red your comment on my ILM blog post, I've also read your article on tiering.

I disagree (obviously hehe) that tiering has to be complex or expensive. I do agree that poorly implemented tiering will absolutely be complex and more expensive than no tiering.

In your article you refer to a sinlge storage pool, which is the key to effective tiering in my opinion. The complexity of moving data up and down the tiers should not be exposed to the users. The user should input lowest common denominator rules into the system and the system should handle the rest based on those rules. While I realize this is a very idealistic view of ILM or HSM, there are vendors doing it now. I'll mention Compellent again because they are doing right at the storage level, resenting one "pool" of storage to the storage consumer and handling the busy work of moving it up and down the tiers on the back end.

Software vendors have tools that do similar things but not quite to the level of simplicity that Compellent has gotten it. I have no doubt that software vendors are working on storage virtualization, encryption and dedupe all wrapped up in one product to sit in front of heterogeneous storage and present it as a single pool coupled with simplicity and ease of use (if they aren't you heard it here first! :)

The tools are not perfect and still rely on good planning to get the rules right, but there again we have consulting companies with deep pools of experience in this expertise to help.

My point was with the current state of storage and data management technology coupled with the frequency of legal challenges surround data management topped off with increasing economic pressures it would seem like automation and consolidation would be a key cost saving, risk reducing method of managing data.

[Gary O]

Tory, Thanks for the comments. I used to work at Compellent. They are definitely the standout exception to my comments. But 99% of the time, large vendors promote their software intelligence to move data up and down tiers. That activity often requires user-intervention and that is where the HSM, ILM, tiering approach often breaks down.