PernixData just announced a new release of its core FVP server-side-storage management product as well as a number of product extensions. Of course, you can get all the specific details from its press release but it did make me consider how a company like PernixData is truly a manifestation of a significant change in the storage world. It represents that change by talking a lot about “decoupling” storage performance and capacity. While that’s technically true, what companies like PernixData really reflect is something bigger than that.
For years “storage” has been just that – a singular, monolithic, and only-somewhat-malleable entity. You get more or less of it, and it is more or less capable to deliver some level of performance. The original IBM term of “DASD” – Direct Access Storage Device (how quaint that sounds in today’s mobile and networked world eh!?) – gave way to “disks” and indeed in many parts of the IT world the terms “disk” and “storage” were actually synonymous….despite the existence of alternatives. But, anyhow, for decades we muddled along with what we had. It was the least-worst option in many respects – and this is said with no lack of respect for the brilliant engineering that drove HDDs from MBs to GBs to TBs. It has been quite a ride.
However the problem with essentially one “storage” type is that it could all too easily get confused with an assumption that what is being stored – data – is itself somewhat homogenous. It’s all just a big mass of “stuff” sitting on some “spinning rust.” But of course we know that isn’t really the truth. Now, we can take it to extremes and point out that there are an infinite number of gradations and permutations in what is demanded of storage in any given situation. But let’s keep it really simple….the “decoupling” that PernixData is addressing really requires us to acknowledge the decoupling of storage itself into two major types – that which is highly active and that which is not. In other words, we should not just talk about “storage” but, on the one hand, about “active I/O” or “working data” and on the other “retention” or maybe “stacks” (if, like me, you remember real, physical libraries!).
When we only had – despite its myriad small variants – essentially one storage tier we began to apply the identical semantic descriptor to all “data storage.” Decouple the word that describes what you’re trying to achieve in terms of I/O and applications – let’s just go “active” and “retain” for ease - and you’ll see what something like PernixData FVP does for you. It is all about an augmentation – no rip and replace required – to optimally serve, manage, and share the relatively-expensive storage media (today that’s mainly server-side DRAM and flash) that you are obligated and/or willing to deploy for your active data. It was not possible to do this in your grandfather’s day (or, for that matter, your father’s) and so we have yet to get used to really thinking this way. But now we can – and should – do both.