As NetApp's chief technology officer, Jay Kidd is responsible for the storage vendor's overall technology direction. We spoke with Kidd about what he sees as the top data storage trends for 2013, particularly around clustered storage, virtualization, flash and the cloud.
What are the major data storage trends you see for 2013?
Jay Kidd: I think the IT world can usually support one revolution at a time, and for the last few years virtualization has been the revolution. Now flash and cloud are building on top of virtualization for a perfect storm of revolutions.
Virtualization is no longer a disrupter, but it was three or four years ago, and it's still finding its way through the system as everybody exploits it more aggressively. Flash is a disrupter on a lot of levels, in terms of changing architectures in servers and storage.
We're also seeing cloud-scale IT and 'IT as a Service' emerging as a viable alternative for CIOs to pursue. That is causing a mindset shift for IT pros, even inside companies who now have to think like service providers to effectively compete with external server providers. This is a disrupter on the people side as well as the operations side.
Clustered storage is another disrupter. Virtualization lets people consolidate 10 applications onto one server. You then spread that across hundreds of servers attached to a consolidated storage environment. We've got customers running two petabytes of storage within a clustered environment running 10,000 virtual machines, and clearly they need to be able to start small, to scale large, but then deliver a level of uptime and availability that previously has only been considered out of the high-end monolithic mainframe-class storage. Now people are expecting it, and building it out of more modular storage technologies.
Our clustered OnTap is one of [the] key drivers of that trend as we push clustered storage into enterprise workloads. We're seeing some good uptake of that technology.
It took a while obviously to get your clustered technology acquired from Spinnaker in 2003 into OnTap. Do you consider OnTap fully clustered now?
Kidd: I don't think we'll ever be finished, because we have too many ideas to put into this, but it's something people are putting to productive use today.
There are always things we want to add and improvements we want to make. We consider it a platform that's at the early stages of its life. We have a rich innovation agenda laid out for Cluster Mode going forward. But it also brings forward a lot of things that people like OnTap for.
OnTap's been around 20 years, and we've built a lot of innovation into that. Things like storage efficiency and what we've done with snapshots and integrated data protection and application integration with Oracle and SAP and Microsoft is all available with clustered OnTap.
Is EMC's Isilon the main competition for clustered OnTap?
Kidd: I'm glad you asked that because there's a misperception that clustered OnTap will only compete with other clustered storage systems. Our main competition for clustered OnTap is EMC's VMAX, and a little bit of high-end VNX, and some of the other storage we occasionally see from IBM and HP. We're going up against traditional enterprise storage systems that are not clustered. Clustered OnTap brings the story of 'start small and scale incrementally' and nondisruptive operations that other enterprise storage systems just can't touch. We'll also compete with Isilon in some of the scientific or media-oriented vertical applications, but the primary focus for clustered OnTap is in the enterprise.
How prevalent do you see flash storage becoming in the enterprise?
Kidd: You think of flash as the Ferrari in your garage. There are certain use cases when you want to take it out, and certain applications where it's the perfect tool, like a Sunday-morning drive in the mountains. But you don't go to the grocery store with it, you don't go to the dump with it, you don't take your kids to school with it, it's not an every-application technology. Not yet. The performance is phenomenal, but the cost is still too high for enterprises to consider putting all of their data on flash yet. Someday, maybe, but we're a long way off from that.
But I do think anybody who's been deploying storage in a latency-sensitive high-IOPS environment will take those workloads and incorporate server-side flash or all-flash arrays as a platform for those workloads.
There have been startups over the past few years that take a different approach to storage for virtual environments, either through integrated stacks or arrays designed for VM-aware storage. Do you think virtualization requires new storage architecture?
Kidd: Virtualization tends to work pretty well, even on traditional architectures with compute and storage in separate devices. If you're a small business and you have eight to 10 applications running your business, it's viable to do that with a single multicore server or a pair of them with internal storage or a small SAN array. As soon as you get to a level of scale, where you've got virtualized and nonvirtualized applications in more than one location, external storage tends to make more sense, mainly because the consumption of compute has become relatively cheap. You can run a lot of applications.
Read more about NetApp's plans for flash
NetApp planning an all-flash array in 2013
That said, there is a customer appetite for a building-block data center architecture with some compute, some network and some storage. We found with our Flexpod reference architecture is that people love the idea of a converged infrastructure, but they love the flex part of it -- 'I like this but I want a little more compute to go with that storage or more storage to go with that compute.' I think converged stacks are important from an ease-of-use and deployment standpoint, but people want to tweak the size of the storage or compute within those stacks.
I think any of [the] single-vendor converged stacks will ultimately fall short in one or more of the technology components. Startups building converged stacks are basically integrators: They're taking off-the-shelf Intel, taking VMware and taking storage, and putting it together in what looks like an interesting package. But they won't ever have the 'best of breed' for compute and storage; they'll always be a generation behind. I think that's even true of big stack vendors; they tend not to be the leaders on the storage side or the network side.
What other new technologies do you see moving into the enterprise?
Kidd: This whole notion of enterprise Dropbox type of functionality is interesting. They meet a need for convenience. Particularly with the rise of tablets, people want to sync between their devices.
But CIOs are worried about the security aspects. It doesn't matter what data you put in their home directory, when you drop it into one of these services, there's a great worry about who can see it, who can hack it, what risks may be there to exposing this data. There are stories about things getting exposed. But enterprise CIOs are looking for a way to give people a service, to allow them to sync between their devices while still protecting the data.
I think in 2013 you'll see the rise of enterprise Dropbox alternatives, and users generally don't care what service they use. They want it to be easy; and they want to be able to tie in their phone, their tablet, their home machine and their work machine into a shared directory so they can move things back and forth. If enterprises provide that for their users, end users will be very comfortable using the one their company provides. That will make it easier for enterprises to make user home-directory data more accessible to mobile devices without the users having to do all this bridging.
So, will NetApp follow EMC's Syncplicity acquisition and offer its own file sharing service?
Kidd: You'll see a combination of things. We have no intention to be a public service provider for this type of service. NetApp's strategy is to provide equipment for this type of infrastructure, but we see the type of capability and tight integration of this as a good feature.
We announced a relationship with [the] Citrix ShareFile program. We want to tell customers using NetApp storage gives them an easy solution for providing enterprise Dropbox capabilities to their users. We'll solve that problem.
Where do you see object storage playing? NetApp acquired Bycast in 2010. Do you see its StorageGrid product continuing as a standalone system, or will object be integrated into OnTap?
Kidd: It will be both. We sell StorageGrid as a standalone solution, and we've paired it with E Series storage. We're also integrating object into OnTap, based on Bycast technology.
I think you'll see a lot of hybrid object file systems emerge in [the] market. In the end game, I think object will become the model people work with for large-scale storage. I think you'll see this gradual evolution to more object-oriented stuff. A lot of people are looking for that type of storage system, particularly with mobile devices.