Violin's (and other pure flash array vendors) post IPO struggles continue

There’s a story on The Register right now about Violin Memory losing its CTO. But that’s not the real interesting story. In the article, Chris Mellor does a pretty good job of laying bare the issues around Violin.

There are several different threads running through this. First, they don’t have much real software IP. Their hardware IP is a different story, but fundamentally, we’ve found that its best to have a very simple and effective hardware design, coupled with intelligent software. That latter part is what they appear to lack. What’s interesting to me is that I had a conversation with a large potential customer recently, where we ran into “how are you different from and better than Violin”. I’ll argue that this is a softball question for us … the difference between an array vendor and a tightly-coupled computing and storage appliance maker is huge, and easy to articulate. The customer got it, but also asked about features that Violin lacked, and are mentioned in the article that are actually not terribly useful in a flash context for performance (which is where you want to deploy flash). Flash is a very different beast than spinning rust, and several of the things that make sense on spinning rust are completely wrong concepts for flash. Flash is about performance, massive parallelism. Anything that detracts from this should be eschewed. Flash is also not about long term (many year) archival storage … or … enterprise flash is not about that, and consumer grade flash ought not to be used for that. Spinning rust is the new tape. Flash is the new temporary (for values in small numbers of years being temporary) disk. Flash is a high performance, nearly memory … platform for storage. You use it for very high performance big data computing, analytics, etc. when the cost of data motion is so large that it makes more economic sense (many more runs per day) to keep it local. Moreover, that same point, the data motion, is why flash array vendors are mostly following similar curves. Many of them are looking at the array as the product, and not the potential value it brings. But they are not positioned to realize that value without a massive re-engineering of their systems. Look at Fusion IO stock, or Violin Memory stock. Look at the acquisitions of Virident, Texas Memory Systems. Etc. I have a sense that the market has too many Flash array vendors, and not enough use cases or consumers of flash arrays. But I also don’t think that flash arrays are a large market in general. End users are very price sensitive, and this is true whether or not they are financial services firms, or educational institutions. Huge price tags for flash (and we got walloped with an increase recently) are going to work strongly against its adoption. What we’ve run into is

While this is the case from the consumer flash world, these are the volume production tent-pole driving most of the world wide flash production. The enterprise parts are much smaller runs, and the volatility is there (and magnified heavily). This is likely weighing heavily upon the array vendors. We see it in our costs on the SSD side. The price gyrations are tremendous in some cases, and causing us and some of our customers grief as we scramble to fill orders. But the public flash array vendors (and a fair number of the startups with absurd valuations) are vying for a market with very expensive kit. Which has an overall depressing effect upon revenues. I think this is why we see the array vendors profits dropping and losses widening. And why we see some flash card builders sell out (TMS, Virident), or go under (OCZ). There are still a few out there (Fusion, Kove, …), and I’d argue that the array vendors are a simple extension of the card vendors. I suspect that there will be more consolidation in this market. I had thought Dell would buy Fusion IO, but this hasn’t happened yet. I think its pretty hard to argue that an array has value at this point, as you get technologies that enable end users to build very large and powerful storage infrastructures out of tightly coupled components. This wasn’t mentioned in the article. But I think (and we are betting on it) it will be the more important factor, in which the tightly coupled appliance architectures are going to be playing very prominent roles.