Reply to post:

Shot version: Pernix sold two product: FVP and Architect

FVP is server-side-caching. You buy an SSD (or multiple) and put them into the server. FVP then copies frequently read data blocks into the SSD to your apps can read them faster. It also buffers writes to the SSD so your writes happen faster and then slowly drains them back to disk. If clusters this across multiple servers so that if one server goes splork all your writes waiting to be drained are still stored elsewhere.

Architect is the thing that looks at your workload and makes sure that all your writes will fit onto the SSDs without causing a flush cascade and crippling your entire infrastructure. (I.E. your writes don't fill the SSDs up faster than they can drain over the course of an arbitrary period defined in part by the size of the SSDs and the speed of the storage you're trying to accelerate.)

To make FVP work you need to buy expensive SSDs + hella expensive software + OTHER expensive software (Architect). This is because it is way - way - easier to cause a flush cascade than Pernix people will admit.

The end result is that Pernix, while not a bad idea and actually decent software, simply cost too much. Why would you pay more to accelerate your servers' access to their storage than it cost to simply buy new storage in the first place, or to move to a hyperconveged infrastructure?

In many - but not all - cases it would have been cheaper to simply toss some SSDs into existing systems and enable VSAN (or Maxta) as an all-flash setup for demanding applications than to use Pernix. (Remembering that I could still use my slow storage for non-demanding applications in a VSAN setup.) By segmenting the workloads in such a manner I also eliminate the risk of flush cascades entirely.

And on, and on, and on, and on. The arguments can go up one wall, down another, 'round and 'round and 'round and 'round. Suffice it to say that while Pernix software works, and works well, you absolutely need to know what you're doing with it, you need to be pretty good at architecting solutions based on some pretty in depth understanding of storage and, well....

...along the way a bunch of easier solutions with lower knowledge requirements came along at the same or lower prices. Pernix as a product couldn't survive as it was positioned or priced.

But Pernix as a pair of features will be a powerful and enticing addition to Nutanix's offerings.