The main competitors that SwiftStack’s OpenStack object storage sees in enterprise deals are EMC’s Isilon scale-out...
NAS and Red Hat’s Ceph object and block storage software, according to CEO Joe Arnold.
In Part 2 of his interview with SearchStorage.com, SwiftStack’s Arnold discussed the company’s main rivals and provided a glimpse into the future direction of his company’s OpenStack Swift-based object storage. He also discusses how EMC’s recent acquisition of OpenStack cloud infrastructure startup Cloudscaling could have an impact on OpenStack adoption.
SwiftStack, a San Francisco-based software startup, sells a commercially supported version of open source OpenStack object storage, commonly called Swift. SwiftStack adds a proprietary management system and a file system gateway to extend the product’s functionality beyond what the Swift open source code offers.
You can find the first part of the interview here.
Do you view your competition to include companies such as EMC’s Cloudscaling, with its OpenStack-based infrastructure as a service, and Rackspace, with its OpenStack-based private and public cloud products?
Joe Arnold: We universally regard them as complementary, not competition. This is because we look to them to provide an OpenStack environment, and we frequently get pulled into partnership discussions so that we can provide a great object storage solution based around OpenStack Swift that can sit alongside what they are doing for the rest of OpenStack.
Cloudscaling and Rackspace don’t offer the same type of product that SwiftStack has built around Swift. We built a product that includes OpenStack Swift versus just supporting the OpenStack components of Swift. So, not only are we the leading contributor to Swift, we’ve also built a product that makes it easy to consume OpenStack Swift. And most of our collaborators in OpenStack Swift provide storage as a service as their primary motivator around OpenStack Swift, rather than provide a competitive private storage cloud product offering.
If you look at the [OpenStack] contributors like Rackspace, they have Rackspace Cloud Files. If you look at HP, their contributions are around their cloud storage product with Helion. HP is interesting because internally for their private cloud, they use SwiftStack. But then from a community standpoint they are contributors to Swift, and the team that does this is the team that works on their public storage cloud product. And IBM, with SoftLayer, has also done some work with Swift, and they also have a public storage cloud.
Who do you view as SwiftStack’s main competition?
Arnold: It’s still the traditional stuff. We’re still going head to head with [EMC] Isilon deployments on a regular basis. And when we get out into people who are building large-scale private clouds, we’re coming up against [Red Hat] Ceph. Those are the two that we see the most.
Why didn’t you mention object storage vendors such as Amplidata and Cleversafe?
Arnold: We just don’t see them as much and get pulled in with 'head to head' in accounts.
What are the main differentiators between SwiftStack, Isilon and Ceph?
Arnold: With Ceph, we see them mostly supporting virtual machines in an OpenStack environment. And that’s great, because we don’t do that. They’ve put an object storage gateway on top of Ceph. But the underlying infrastructure is still based on that same infrastructure. The main difference between us and Ceph is that we’re entirely focused on being great for object. By not being able to do virtual machines, for example, it means we can do a better job of scaling to multiple data centers. It means we can handle more concurrency. That gives us an edge when people are doing object storage workloads.
With Isilon, it’s a very similar story. People have been using filers for storing unstructured data, and they should be using an object storage system, because the way you can grow and scale those environments is much simpler. Instead of having multiple storage environments set up once the scale starts getting up, you can have one system to manage. That’s a lot easier for operators.
Red Hat acquired Ceph. Do you have similar aspirations to be acquired?
Arnold: We’re out to build a big storage company around the use case that we’re going for. We’re not trying to find our exit. We’re really thinking through what partners we’re going to use, what use cases we’re going after, how we’re going to communicate that out to the segments we’re going after, and build up the sales and marketing folks around that.
What’s in store for SwiftStack during the next year?
Arnold: There’s going to be two things. We’re going to have a very full suite of functionality now that we have the combination of storage policies and [in the coming months] erasure codes. That’s going to really give a lot of flexibility. That’s near-term.
Looking a little bit further ahead, we’re seeing a lot of change in terms of how devices are being manufactured. That’s going to increase the capacity in those devices and change how the data actually gets written to them. It’s going to be much harder to do random reads into those devices. That means it’s going to be even harder to produce large-scale block type storage, meaning more applications are going to convert to the file storage that we are especially tuned for, which are unstructured object-based interfaces. That’s going to be the way to get the most performance out of the system from top to bottom down to the physical media.
We’ve already seen this happen with flash, where when you write into those environments, it writes into another location. We’re starting to see alternative APIs into the actual storage medium. Seagate is launching its Kinetic drives, and that means it’s going to be object interfaces all the way from the point of entry into the storage environment all the way down to the spinning media. We’re tuning ourselves up so that we’re going to be in the best position for not only this wave of devices but supporting this onslaught of data that’s coming our way.
You mentioned erasure codes, which are due in the upcoming OpenStack release, known as Kilo, in the spring. When can customers expect to see erasure code support in the SwiftStack product?
Arnold: Definitely by the Kilo release, and we’ll have customers with it up and running, similar to how we had storage policies in our product ahead of when [OpenStack] Juno got released [on Oct. 16].
OpenStack Summit shows momentum for Swift, Cinder storage
Storage plays key role in OpenStack cloud