Whenever there is a new idea, a new movement, a new music style, any new thing in any aspect of life, we can observe more or less the same timelines during its lifecylce. These days, I’m starting to see the same traits in the Flash PCIe market.
At the beginning there is the explosion of a new market. Someone launches a new technology (or it can be a new music genre, this example works in different areas), and suddenly a completely new market has born. They had an idea, they identified a new need from customers, or the same need that could have been solved in a different and better way. In this case, the bottleneck created by old bus connectors like SATA and SAS were preventing the complete use of insanely fast Flash memories. By connecting the same memories directly into the PCIe bus, that bottleneck was removed and customer could experience better performances.
Immediately, a great number of followers enter the same market, with more or less the same solution: there are for sure differentiating features in each of them, new ideas brought into the same technology, but at the end they all compete in the same market. Venture capitalists see a new business opportunity, and during the hype of the new technology whoever has an idea regarding the same solution is quickly and easily funded and can start delevoping that idea and try to make it become a business. In this case, there has been a incredible amount of vendor creating and selling Flash PCIe devices.
At some point however, the market reaches its peak. Also at this stage, there are some traits that can be identified. I’m not a business expert, but those are the ones I usually see:
– no more new vendors
– many of those startups go bankrupcy
– only few of them become independent corporation
– most of them are acquired or they merge
Sum all of these traits, and the final result is a market that maybe is still growing by the numbers, but the list of vendors in this market is shrinking and the market shares owned by each of them grows.
Is this one of those symptoms?
The latest news in the Flash PCIe market has arrived few days ago, and was a huge one. Fusion-IO, probably the most known vendor and one of the leader in this market, is going to be acquired by Sandisk. This is big: Fusion-IO has already proceeded into its evolution journey, was already quoted into the stock exchange, and had a market value higher than a billion dollar. And nonetheless, some of the giants in the broader market of NAND memories (the ones used to create all these devices) like Sandisk is going to buy Fusion-IO, for a price around 1.1 billion dollars.
Is this a sign of a mature market that has already reached its peak?
In my opinion, yes. No new companies has entered this market in the last few years, many of the others have failed, some were sold (Violin Memory has dismissed his Flash PCIe solutions for example) or have been acquired (Virident and sTec were both acquired by HGST few months ago ….). I’m not an analyst, maybe I’m totally wrong, but all these news summed together for me are a clear sign of the reached peak. Also, it’s interesting to look at who’s surviving: not the Flash PCIe vendors, but mostly NAND memories producers: Intel, Samsung, Sandisk… The ones who produce the memories, not the ones assembling and packaging them. The supply chain is shrinking, and those vendors can now produce the memories and sell them directly, after packaging them into final products (like SSDs and Flash PCIe cards).
What’s next?
What’s next in reality is already happening. When a technology is mature and has reached its peak, it goes on for a long time, it’s not disappearing. But, innovation starts to come from further different technologies. I’m sure some of my friends working in this market are going to yell at me, and they are eager to explain me in great details all the tech behind their new product, and how it’s still evolving. But let’s be honest for a moment: a Flash PCIe device is a Flash PCIe device. New models will be probably faster, bigger, cheaper, but it’s the same technology, refined and improved.
What I’m seeing today are new technologies solving the same problem that Flash PCIe has solved (the need for incredibly fast local storage with ultra low latency), but in a different ways. And as I said, they are already here: mainly, DIMM Flash memories. Those are again flash memories, but this time directly connected into the DIMM slots of the motherboards. Even faster, with lower latency, because they are even nearer to the CPU.
Will they erode the Flash PCIe market? partly yes. Just like has happened in the past with other technologies, people who are fine with the performances of Flash PCIe devices will stick to them, others in search of higher performances will probably move to those Flash DIMMs. Flash PCIe devices will become more common, cheaper, like today SSDs are the main storage units inside notebooks and desktop computers while years ago they were an expensive alternative to spinning hard disks.
What I’m more interested into, however, is yet another technology, that is coming from software and not from hardware. Think about this: we already have an incredibly fast storage layer in any computer or server: the main memory. Nothing is fast as this, not even Flash DIMMs. I know what are you thinking. The huge difference between the two is the volatile nature of memory: once we power down a system, memory content is lost and no data survive inside it. Sure volatile memory cannot be a primary storage tier, but as a fast tier to be used as cache, it’s already superseeding any NAND device. Solution like PernixData FVP (announced memory support in the next release) or Atlantis USX (I talked about it in a previous post) are promising a software solution to safely use memory as a proper tier, at least for caching, by creating redundant and protected copies of data stored in it. But this is a topic for another post 😉
For now, let’s keep an eye on the Flash PCIe market: will see one year from now if I was right or wrong.