Sramana Mitra: Help us with the competitive landscape. Whom do you compete directly or indirectly? What does that landscape look like?
Momchil Michailov: Software-defined storage has been a hot and trendy area. In flash storage and software-defined storage, there was a total of $1.2 billion of venture investment last year. For the pure software stack, we have been positioned against companies like ScaleIO, which was recently acquired by EMC. We are certainly compared a lot to a company called Nexenta and a number of other players out there.
Software-defined storage allows you to converge compute and storage. When you converge compute and storage, you can eliminate a lot of the paths. You can improve performance. That’s been an architecture that has been successfully pushed and has gained a lot of traction for a company called Nutanix. We are strictly a software play. We don’t necessarily compete with Nutanix. You can certainly take our platform and layer it with Cisco UCS servers, with Whiptail Flash inside them and deliver a similar architecture to Nutanix on an open platform. There’re the pure software guys like ScaleIO and Nexenta and then there are the more converged players like Nutanix.
Sramana Mitra: What are you seeing in the market with your approach?
Momchil Michailov: Our market is a little bit different. Our biggest difference between Nexenta and ScaleIO is two-fold. The first piece is that we provide scale-out to 2,048 nodes. Nexenta does not provide scale-out. They’re focused on vertical scaling versus scale-out. ScaleIO has scale-out to a few dozen nodes. The first layer is that storage scale-out. The second layer that we provide is the shared file system on top of that distributed storage. Once you distribute the storage and provide all of the advanced storage features, the second thing is what we call the data management. That is where we layer our cluster shared file system.
Nexenta has a file system based on the Z file system (ZFS). It’s an open source project at this point and there’s a lot of storage solutions around ZFS. They are all scaled vertically. Vertical scaling makes a lot of sense for certain workloads. At some point however, that becomes very much similar to proprietary storage appliance where you need scale-out and not just capacity to scaling. That is where we take over. In our case, you can scale up to 2,048 nodes horizontally. You can scale up to 18 million exabyte in file system sizes.
The benefit is that all of the servers, whether physical, virtual, or cloud, concurrently access the same data. In Nutanix’s approach, you can migrate from one machine to another machine. However, when you get to larger data sets, at 200 terabyte data base, that becomes a cumbersome task. Our focus is on creating an agile storage layer. On top of that agile storage layer, we layer a clustered shared file system, which is workload diagnostic. At that point, our customers access that storage infrastructure and their data sets through any physical, virtual, and cloud machine concurrently. They have both infrastructure and workload scale-out. That’s the major difference between any other solution out there and ours.
Sramana Mitra: Let’s talk about where you see the market going. What are the emerging trends in this space? More broadly, what are the other developments in your segment?
Momchil Michailov: It’s obviously an area that has attracted a lot of interest, entrepreneurs, and funding. It’s a multi-billion dollar market. We see two separate trends. One trend is the entire reworking of the data center. It’s similar to Nutanix where you rip and replace a whole new architecture that is then tightly coupled with a particular hypervisor. That is certainly a very interesting and compelling argument. We’ve seen reluctance from certain customers for that complete overhaul. While there’re a lot of benefits to disruption of doing it that way, it’s also fairly significant.
We are also seeing a lot of innovation in moving traditional storage capability into software stacks. How do you take a very sophisticated storage array – something like EMC Symmetrix – and move that in a software stack? The notion there is that your compute today is so powerful and capable that there is no need for you to off load that capability to another device. We’re mixing each one of these trends. We do believe that there’s going to be a convergence. Certainly with flash technology coming into the storage world, the ability to run these workloads that much faster and have that flash performance and capabilities for less energy consumption and less cooling needed is a significant benefit, aside from just a pure performance piece. We have taken some of that convergence play. We allow customers to take industry-standard servers and put the entire storage infrastructure inside them. But we have also kept a lot of the traditional storage features and we’ve exported these in a software layer.
We don’t believe that customers would need to rewrite their applications to be able to benefit from this new architecture. That hyperscale architecture is currently mostly deployed in web tool workloads. But enterprise customers don’t run web tool workloads for the most part. We focus on the hyperscale architecture and bringing that converged infrastructure into play. We also enable customers to keep their traditional business continuity and management data and storage management operations. Snapshots and SLAs are now guaranteed through software instead of hardware. We believe that allows customers to mix the best of both worlds – the notion of pure conversion of infrastructure and eliminating the dependency on hardware devices and instead providing that in a software layer.