EMC Corp. will probably spend $10 billion over its "natural course of business" during the next five years, and software will continue to command the bulk of the investment, according to Jeremy Burton, the company's president of products and marketing.
Burton said only about 300 of the 13,000 employees in EMC's product organizations currently work on hardware. He said the main difference between past and present initiatives is the decreased emphasis on specific software tested to work together with a specific set of hardware.
"With the new stuff, you can say, 'You can run it on pretty much any commodity drive,'" he said. "But, it's not a massive departure from what we've probably been doing since 2008."
That was the approximate timeframe when EMC moved to industry-standard components, such as Intel chips and Seagate hard drives, according to Burton. EMC also started development on its Atmos object storage, which Burton said was superseded by the company's software-defined ViPR storage projects for petabyte- and exabyte-scale storage.
EMC signed its first petabyte-scale customer when Burton joined the company in 2010. Two years later, the number of petabyte-scale customers had soared to 1,000, and EMC now has one customer with exabytes of data, he said.
"In five years, I'll bet we've got a thousand," Burton predicted.
In an interview with TechTarget, Burton expanded on EMC's strategy to cope with the explosive growth of data through an expansive product portfolio, which includes the ViPR software-defined storage, upcoming DSSD server-side flash product (due later this year), XtremIO all-flash arrays, traditional storage arrays and Pivotal Software spin-off which focuses on building distributed applications and analyzing big data.
Which of your storage products would exabyte customers use?
Jeremy Burton: The trend that plays to most [exabyte customers would be] software-defined storage on commodity drives. What you find is that for the last 20 years, databases and things like that run on block storage arrays. A lot of the new apps that are built, they're not going to use block storage devices. They're going to use object stores ... It really is the ViPR projects that have given us the entrée into Web-scale systems … We have a Hadoop infrastructure that sits underneath analytic systems … Is everything going to be commodity? Well, a lot will be. The volume of data will be on commodity drives. But when you want to do real-time processing and high-performance big data analytics, the gating factor becomes how quickly you can get the data from the physical medium it's stored on into the memory and processed … And DSSD is one example of a very, very relevant flash hardware company that's got big use cases for in-memory and big data analytics. But once the data has been processed, it's going to get dumped out to the commodity storage, to the object store. So, these things are going to work together.
Who is your biggest competition in this new world of storage? Who do you run up against in deals?
Burton: In the new world, it's still getting going. But there are a number of the object store guys that are out there. You've got DDN, Scality. Ceph you see a lot, which is open source. Now it's mainly used in dev and test. With a lot of these new distributed systems, it's still in the early days. The hope would be, when we have a bunch of things coming up, to make access to our next-gen software much more frictionless. You're going to see a lot of interesting things this year that EMC five years ago would never have done.
How seriously do you take drive makers moving into the storage arena, such as Western Digital/HGST, which just acquired Amplidata, SanDisk, and Seagate?
Burton: They're trying to move up the stack. I think the challenge that these companies always have is, what's the route to market? What those companies do have, by virtue of selling drives, is access to the hyperscale guys. But then the question is: Are the hyperscale guys going to use the software that comes with the drive? Probably not. They're probably going to build it themselves. So, they're not going to sell it to the hyperscale guys. Who are they going to sell it to? They're going to go sell it to Morgan Stanley? All right, so now you need a sales rep. That's not their business model, because if you need a sales rep, then you need someone in support. And then you've got this expensive enterprise infrastructure that's got to surround the sale.
So, does it technically make sense that these guys would want to move up the stack? Yes, because I think every tech company over time tries to move up the stack. But, in the same way as we didn't fold Pivotal into EMC -- because it's a different gig, it's a different buying center -- by just rolling a set of storage software into a drive company, it's not clear to me that will be successful. If you look at the way those businesses are set up, they're set up for OEM.
IBM recently announced a $1 billion investment over five years in its Spectrum storage software, which can run on commodity hardware, and two years ago, made a $1 billion investment in flash. How will EMC's investments compare?
Burton: We're going to spend more on XtremIO each year than they will on their entire software-defined storage strategy, just to put it in perspective. They tend to do these billion-dollar announcements, but over five years, that's not much.
To what degree are you selling to traditional storage buyers?
Burton: It's still 95% infrastructure IT guys. But, in the last six or nine months, we'll talk to business guys, sometimes the CIO, and they'll bring a new guy in who's going to do software development. I distinctly remember the first time, probably a year ago, and I was saying, 'Whoa, that's weird. You didn't bring the IT team with you. Who's this guy?' And they said, 'Oh, he's our head of research and development.' And this happens every couple of weeks [now].
What kind of companies come in with their research and development departments?
Burton: It's all over. It's automotive, retail, financial services, pharmaceutical. It's across the board.
Are they developing their own custom applications?
Burton: Yes, custom software. But, bear in mind, these guys come in to talk to Pivotal. Because of the setup we have, it may be the EMC rep that said, 'Hey, you're going to do software? Come talk to Pivotal.' And so Pivotal then will say, 'Hey, what are you trying to do?' And they'll convince them that Pivotal has got the chops to teach them how to build software and give them a dev platform -- which is Cloud Foundry -- to do it on. And our interest is to say, 'By the way, when you build that application, you're going to architect it differently, and you're going to need a different infrastructure. You should hear what we've got by way of software-defined storage. And you're going to want a Web-scale infrastructure to go with that app."
Pivotal is not really a storage story, is it?
Burton: It's not, but I'll tell you the connection point. The guy who built Symmetrix, Brian Gallagher, he's been around forever. We moved Brian out of that role about six months ago. Brian is now working on the very bottom end of Pivotal, Cloud Foundry, [where there] is this layer called BOSH, which connects to the infrastructure. I have a team of people working for Brian that now work on the open source project, that BOSH layer, to best connect Cloud Foundry to the underlying infrastructure. BOSH is basically a cloud abstraction layer. BOSH allows you to go deploy into a VMware EMC private cloud, into [Amazon Web Services] AWS, Azure, OpenStack, whatever.
Is it a translation to the object storage?
Burton: Yeah, and Brian's trying to make that layer smarter, so that if you've got something like a DSSD down in here, it knows the way these things treat storage. Now it might be advantageous to know, oh, that's super-fast storage. Can we add some attributes up here to let this thing know that, hey, if you throw it at this, it's going to go more quickly?
Read part two of the interview with Burton to get his take on the role the DSSD flash appliance will play in EMC's storage lineup.
EMC previews ViPR storage at EMC World 2013
EMC World 2014 conference coverage: ViPR, federation, cloud
First product to form out of DSSD announced by EMC