As a result of some of the longish posts, comments, and replies about the architecture and functionality of DL4000s with deduplication that I engaged in with Curtis Preston and others, I got to thinking.
My feeling was that Curtis was wrong, but not for any specific technical reason.
Sure, some of his points were wrong in my estimation. Some were right.
But even accounting for that, I had the feeling that the sum of his argument was more wrong than right, and more wrong than I could account for by simply saying that this or that specific point was wrong.
Put another way: the whole of his argument was less than the sum of its parts.
But why exactly he was wrong is, I think, interesting.
Mr. Preston is wrong because the discussion is not relevant. Not cogent to why somebody might choose to buy a particular virtual tape library, or deduplication device. So ultimately, it comes down to not so much that he is wrong, but that the whole discussion fundamentally misses the point, and is not relevant.
It doesn't pass the "so what?" test.
If you say that a DL4000 with deduplication has seven pools of storage (it doesn't, but lets just make a hugely uncharitable supposition for the sake of this discussion), then the next question is: "So what?" Why does that matter?
And what does matter? Well, what matters is operational behavior. How easy is it to manage? How easy to deploy? What does a backup admin need to do to manage the device, vs. a storage admin? (And the answers to those two questions are, hopefully, little and nothing respectively.) Arguing nits about what storage does what and how many pools there are and what is cache is a red herring.
What else matters? I think that what matters is:
- What is the use case?
- What is the TCO?
- Does it reduce the management complexity of backup?
- Does it meet these objectives in a way that few if any competitive products can?
- Does it reduce risk, or increase data security?
- Does it increase the reliability of my restores?
- Does it improve service levels, and meet my RTOs?
These things are important. And discussions about what is under the covers of a given product are only relevant insofar as they help us answer one of these questions.
In particular, issues about the number of pools and all that are only really interesting as they relate to performance and cost. And performance, is, at root only interesting to us insofar as it determines our ability to meet SLAs and backup windows and recovery windows and RTOs.
Cost is only interesting when we have a sound understanding of our TCO (Total Cost of Ownership) of backup, and what impact of deduplication or virtualization will have on that TCO. Or when we are staring at the price for an (allegedly) identical competitive product and can do a direct "apples-to-apples" comparison. Seven pools of storage, or twenty seven, or two, the storage will have a price. And is that price sufficient to help me reduce my TCO? Is it competitive?
And once I figured that out, I understood why Mr. Preston's arguments troubled me: not because they were wrong but because they weren't relevant to the buying/ownership decisions for virtual tape and deduplication.