Fusion-io Distances Itself from DAS AND SSD; Interview with Fusion-io CMO Rick White Part II

Fusion-io installs its ioDrive in server PCI Express slots and its software driver in the server OS so it is easy to understand why Fusion-io gets labeled as DAS. While it may certainly be used that way and started there, Fusion-io has a much different endgame in mind. As Fusion-io’s CMO Rick White explains in this second part of our interview series, Fusion-io creates an active data tier that seamlessly moves data between the new active data tier it creates and backend high capacity disk.
Jerome: So what are some of the ramifications of decoupling performance from storage capacity and how to do you differentiate yourself from other SSD vendors?
Rick: Today to have persistent, high speed storage for the enterprise, it comes from the SAN. But you have to buy the hottest storage. Generally, a lot more disk, a lot more RAM. You have to build these things up to get the performance.
Things like thin provisioning were introduced to work around the fact that you had to buy performance and capacity together. Fusion-io decoupled that.
Now performance for your applications may be purchased separately from your storage. Why is that important?
We are the only company doing anything with enterprise flash. This does not mean disk drives are going away but they only follow Moore’s law for capacity.
The problem with many of the SSD vendors, which we are not, is that they say they are the new disk drive. Look at how fast we are! Isn’t this awesome?! Whoa! Now you are coupling capacity and performance together in the disk. Fusion-io does not want to do that.
We want to introduce an active data tier which is for active data sets. The thing about active data sets is that after you finish with them – generally they cool off after 30 days – you only access them again 2% of the time. So for that data you need something like Data Domain.
This is why one of Data Domain’s founders was one of Fusion-io’s first angel investors. He got it. Data Domain is the archival tier which needs just really cheap storage. Fusion-io is the active data set.
The problem with an all-in-memory database or any other application is that you can only scale RAM so much. It is super expensive and it is not persistent so you have to keep it powered on all of the time. So Fusion-io becomes a really great memory tier in the server and that is when we realized we decoupled performance from capacity.
We decentralized performance from today’s storage mainframes. Now we need to migrate data which is why we developed direct cache and some of our storage targeting software.

Jerome: So is Fusion-io DAS?

Rick: Everyone wants to call us DAS as they come to trade shows and see our booth with thousands of videos running on the wall. But we have a little hand-out that says, “Fusion-io has set this one up as an iSCSI target. All of these servers are network booting off of our disk and all of these videos are streaming off of this disk over Ethernet and Infiniband.”
And yet they say, “That’s great! You are just DAS, right?” Clearly that is not the case but it is hard to convince people otherwise.
Rather we have our own software that migrates data as it cools using our caching software. So while we can migrate data off of raw metal or raw iron, we found out we could not do it
with VMs..This is why Fusion-io acquired ioTurbine.
For us it is about decentralizing the performance and then being able to migrate that data as it cools off to an archival tier of disk drives because you cannot beat cheap, high capacity disks and you should not try. RAM did not try to displace them and neither should flash. There may be a niche for placing high-RPM disks but even that is silly. Really, it just needs to be a new high-memory tier.
In Part I of this series, Rick discussed how server-based flash is poised to change the enterprise.
In Part III of this series, Rick explains the new Fusion-io Octal drive, what makes it different from Fusion-io’s earlier ioDrives and how Fusion-io is going to market with it.
In part IV of this interview series, Rick and I discuss why Fusion-io
is opening up its virtual storage library (VSL) APIs to developers.
In the final Part V of our interview series, Rick provides Fusion-io’s takes EMC’s Project Lightning (now known as VFCache) and the gap that persists between SSD providers and Fusion-io’s ioMemory.