Last week’s acquisition of NexGen Storage by Fusion-io was greeted with quite a bit of fanfare by the storage industry. But as an individual who has covered Fusion-io for many years and talked one-on-one with their top executives on multiple occasions, its acquisition of NexGen signaled that Fusion-io wanted to do more than deliver an external storage array that had its technology built-in. Rather Fusion-io felt it was incumbent for it to take action and accelerate the coming data center transformation that it has talked and written about for years.
When I first learned of Fusion-io’s decision to acquire NexGen Storage, it caught me a bit by surprise. I had talked with its CEO David Flynn as well as with Rick White (its CTO, CMO and general person extraordinaire) on a number of occasions and their overriding opinion (paraphrased) was that disk storage, as it is currently used in production data center applications, was on the way out and would need to be replaced by a new faster tier of storage.
While many in the industry hold to this opinion, Fusion-io’s position differed (at least early on) in that it contended this new tier of storage would emerge not as a new tier of storage (such as Tier 0 or flash memory/SSD-based storage) but as a new memory tier. It contended (and still does to the best of my knowledge) that the processing overhead and time involved to convert data in memory to data storage and back again so that it is usable for processing would become unacceptable over time.
They contend (and rightfully I might add,) “Why not just store data in memory all of he time and eliminate this unnecessary transformation of data that needs to occur during processing?” This thought process and positioning explains why Fusion-io puts its ioDrive technology inside of servers, refers to it as ioMemory and ships with its own driver that eliminates all of this back and forth data transformation that is currently needed to store data on disk. It also explains why others such as EMC and Dell are actively working on solutions that compete with Fusion-io because of the huge acceleration in performance that this approach offers over even using flash memory based storage arrays.
However if Fusion-io’s approach is so much better than any flash memory array shipping today, then why buy NexGen? This raises the following questions that I wish to examine in this blog entry:
- One, while Fusion-io’s technology has been widely accepted, it probably has not experienced the broad market adoption and displaced traditional storage solutions as quickly as many thought it might despite the vast improvement in performance it offers over these arrays, to include flash-based arrays?. The question is, “Why has this not occurred?“
- Second, if Fusion-io’s technology is so much better than using storage arrays, why buy NexGen Storage? Granted, NexGen used Fusion-io cards in its storage systems. But the ugly truth is that NexGen does not eliminate the data transformation process that needs to occur from memory to storage back to memory again since NexGen Storage is connected to servers via traditional 1Gb and 10 Gb Ethernet connections. So while NexGen may be faster than other tradtional storage arrays that use SSDs because it uses Fusion-io cards, it does not offer the same level of performance as a native Fusion-io card internal to our server.
I’ll answer these two questions in the order that I posed them.
As to why Fusion-io’s technology has not taken off the way some thought it might, there are at least four easons for it.
- Too expensive. While Fusion-io ioDrives are high performing, that performance comes at a cost and the vast majority of application cannot justify paying that much for the boost in performance it offers.
- Cannot be shared. When a Fusion-io ioDrive is put in a server, it can only be used by the application(s) or virtual machine(s) on that physical server. This adds to the difficulty in cost-justifying the purchase of a Fusion-io card and explains why companies like NexGen that used Fusion-io in their systems emerged in the first place: to better share and optimize the available performance and capacity on Fusion-io’s ioDrives.
- Flash memory and SSD solutions are “good enough” for now. Going from 10K or 15K RPM spinning FC or SAS disk in a storage array to flash memory or SSD has resulted in 3-10x or greater improvements in application performance. While the performance is not as good as running Fusion-io inside the server, it is still a substantial improvement in performance at a cost that more organizations can absorb.
- Difficult to upgrade and/or maintain. Adding more storage capacity and performance inside a server or replacing a faulty ioDrive can be problematic since it may require taking the server offline to do so. Adding more capacity or making a repair on an external storage array is much easier since they are typically designed for 24×7 uptime.
It is for these reasons and others that Fusion-io has arguably not achieved the adoption rate that many thought it might. This also explains why hybrid and flash memory storage providers like NexGen have emerged since they do a better job of making flash available to a greater number of applications at a lower cost.
But if that is the case and where all storage will eventually end up, Fusion-io should simply look to partner with as many of these providers as possible and put its ioDrive technology inside of them. If anything, its acquisition of NexGen Storage could almost be seen as a detriment to its plans for growth. After all, why would other storage providers want to use Fusion-io’s technology if Fusion-io is building its own storage arrays? In that sense, Fusion-io would have been better off staying independent so it could become the Intel (so to speak) of flash memory based arrays.
This gets to the real heart of why I believe Fusion-io bought NexGen Storage. Yes, Fusion-io wants to capitalize on the current craze in flash memory based arrays and have its own product offering in this space. This it accomplishes with its acquisition of NexGen Storage.
By acquiring NexGen Storage, Fusion-io in the near term provides some answers to the valid criticisms mentioned above about its current suite of product offerings. It drives down costs, it makes its technology shareable by storage attached servers, it provides the “good enough” storage solution that SMEs need now and it delivers the solution in a manner that is easier for them to maintain and upgrade.
Bringing NexGen into its fold also addresses another key concern that SMEs have: Support. While anyone can take a server, install a couple of Fusion-io ioDrives in it, put a Windows or Linux OS on it and call it a “storage system,” by Fusion-io bringing NexGen in house, SMEs get the interoperability matrices and application support guarantees that they need to host their applications on such a storage solution.
Fusion-io having solved these near term tactical issues, it can leverage NexGen to achieve its more strategic goal: creating a new memory tier in the data center that will accelerate the data center transformation.
As Fusion-io has previously told DCIG, being a hardware company is really not core to its DNA: deep down Fusion-io wants to be a software company but needs hardware to deliver on the promise of its software.
So here is what I think is going on. NexGen Storage already uses Fusion-io internally and offers ioControl software on its arrays. Therefore it is not a stretch to believe and which, in fact, may be the most logical conclusion to draw,is that Fusion-io will look to extend its existing server-based ioMemory capabilities into arrays residing in the the storage network starting with its newly acquired NexGen arrays. While I would not look for such capabilities in the next few months or even the next year, I see it as almost inevitable.
While companies are currently enthralled with the 3-10x improvement that flash memory arrays provide, that thrill will soon be gone and they will again be looking for another 3-10x or greater jump in storage performance without a huge price increase, By Fusion-io extending its ioMemory capabilities into its NexGen arrays, there is no reason to believe Fusion-io cannot meet these forthcoming SME expectations . More importantly, Fusion-io is arguably better equipped to make this leap than competitors since Fusion-io’s DNA is to be a software company that operates at the memory as opposed to the storage level in the data center stack.
Everyone knows that a data center transformation is underway and it is occurring in ways which few of us fully grasp. However from my conversations with Fusion-io’s executives, I do get the sense that they do more than grasp the level of change that is about to occur. They see the role that they could potentially play in facilitating it. Fusion-io’s acquisition of NexGen signals that it is tired of playing a passive role in bringing about this data center transformation and is going to, in the very near future, start shipping products that hastens this occurrence..