EMC CEO defends federated business model, debunks storage myths
- 04 September, 2014 06:35
There's EMC and then there's EMC.
Confused? Well, there's EMC, the vaunted storage and converged infrastructure company. But then there's EMC, the capstone of a corporate federation -- including VMware and big-data startup Pivotal -- that has positioned itself quite nicely to capitalize on the critical trends reshaping corporate IT. (Think mobile, cloud, big data, etc.) This summer, EMC's federation model came under fire from activist investors who claim the company would be better served by bringing assets like VMware in-house.
In this installment of the IDG Enterprise CEO Interview Series, EMC's Joe Tucci spoke with Chief Content Officer John Gallant about the federation model's value for IT leaders and how the partners -- including security provider RSA -- coordinate strategies to deliver innovation and choice. He also debunks what he views as nonsense being spewed by newcomers to the storage market and talks about how EMC is empowering customers to build private/hybrid clouds that stand up to the best the public cloud has to offer.
Computerworld: Let's start off by discussing your federation model. That's been in the business news this summer, but I want to talk about it from the perspective of the IT buyer. What is the benefit of that approach for IT and why is that better than, say, having all those capabilities under a single roof?
Tucci: It starts off with focus and how you build the best products. To build the best products you've got to have the best people. If you have a focused mission, one that can make a difference, something that really captures your attention, you'll get more directly rewarded for what you actually are doing than being part of a big company. You're likely to pull in better talent, which we have. Then I go back to where I started there, which is on what we're going to build. You say: Here is my mission. I'm going to build the best software, best solutions, and the best technology for this set of problems or opportunities.
Then it goes to giving customers choice. The principles of the federation are [that] it's highly strategically aligned with a little bit of overlap, because if you don't have overlap then you have seams and that's where all the competitors go. You align the missions of the companies in the federation and then we operationally align so we can bring all that technology together as if we were one company. But if a company says: Hey, I love this piece of VMware but I want to use another storage product or I want to use another converged infrastructure product or I want to use another security product, we say go ahead. Giving customers that choice is critical.
So it's all about, again, how do I get the best talent to focus on their mission? How do I align the missions with the different companies in the federation? Then how do I say to customers: We've got a whole stack for you. It's incredibly compelling but I believe in the right of choice. I think that's what the federation does for us.
Computerworld: One big challenge of that federation model is getting customers who deal with these individual companies to understand the whole federation strategy as well as how all the pieces come together. How do you deal with that challenge?
Tucci: The last piece of it is -- and this is the piece, in hindsight, maybe we should have done earlier or better -- is operational alignment. How can we go together to a customer? One of the things I should have done, which we're taking steps [to address], is for the larger accounts, have one global account manager that each of the federation works with to make sure that where the customer wants it we can appear as one. Where they want more point solutions, we can also [do that]. Any choice a customer makes, we have an opportunity someplace in the federation. It's kind of a fine-dining menu. On the left side you've got the prix fixe and on the right side you've got a la carte, and if you want it all blended and matched together with wine, delivered elegantly, we've got that. If you say: Well, today I only want the salad and a pasta entrée -- you can do that.
Computerworld: How do you and the leaders in this federation set direction? If you look at the model most people are familiar with, a big tech company buys other companies, blends them all under one roof and the CEO sets the direction. How do you do that in that federated model where you have multiple CEOs working together?
Tucci: That's the strategic alignment. That where you're saying: Okay, here's a set of missions you're going to work on. Here's a set of missions that the second CEO is going to work on. In our case we have kind of three-and-a-half companies in the federation. Basically we run it as if we had a security company, a storage/converged infrastructure company, a virtualization and end-user computing company and a big data company. Paul Maritz runs Pivotal; Pat Gelsinger runs VMware; David Goulden runs EMC Information Infrastructure; Art Coviello, who reports to David, runs RSA. Each of them is developing their strategy and what you're doing over the front is saying: How do these strategies fit together with a minimal amount of overlap? But I'm a big believer in the overlap theory, or else you're going to leave seams and, as I said before, that's exactly where your competitors will exploit those seams.
Computerworld: Today, the entire IT stack we've known for years is being changed, with mobility, cloud, big data, etc. Lots of massive shifts, massive opportunities. Today I think if you talk to any customer, it's deeply ingrained in their mind that EMC owns that storage layer of the stack. But what do you want people to think about EMC? How would you position EMC in their minds if you could start fresh?
Tucci: I think part of the problem is EMC is used twice. Once it's used as EMC, which is the federation, then it's used as EMC II, as the business that has the storage. For EMC, people think of EMC II. I'd like, when they think of EMC, they say: You've got tremendous offerings in cloud. Basically, as you said, what's driving the future is in no particular order; mobile, cloud, big data, social networking and then security is just the key enabler across everything. If you look at those drivers, we're basically focused on those metrics.
On end-user computing, we've given that to VMware. They just bought AirWatch and they're doing more on the mobility side. If you think of cloud, VMware is doing a software-defined data center, but EMC II is doing ViPR - a big piece of software-defined storage. You put that together and I think we have a better story than anybody. You think of big data, it's exactly what Pivotal is doing. They're doing it with the HDFS new-wave scale-out big data approach. RSA traditionally came from IPV -- identity protection and verification -- then security analytics, which assumes that bad guys have penetrated your system and stolen IDs and then you basically look at every packet for activity and either allow it or don't allow it or ask more questions based on what you're seeing against real profiles. Then, of course, are we adhering to policies and good governance against the policies that we've set, either as an enterprise or set for you by a government. That's what RSA is.
It's almost three missions apiece. VMware is working on end-user computing, they're working on software-defined data center, they're working on a hybrid cloud. EMC is working on storage, software-defined storage and converged infrastructure. Pivotal has a developer's platform called Cloud Foundry, developers' PaaS. On top of that they have big data capabilities on the back of HDFS and then they have Pivotal Labs, which helps you develop programs very rapidly to use these big data stores. RSA is IPV, security analytics and EGRC [enterprise governance, risk and compliance].
Now, within that, under vSphere, VMware has done a little bit in VSAM, but that's for a specific use cases in a vSphere environment. That's a little bit of the overlap and everybody, of course, writes about the overlap. But again, if you look at what we have and the way we're approaching it, [the overlap is] way over-hyped. You know, all the press about VMC and VMware, aren't they competing at the edges? But we have these swim lanes and I think it's very clear which technology you should use for which swim lane.
Computerworld: At that upper federation level, what would you like people to have seared into their brain about it?
Tucci: If I really believe that cloud, big data, social networking and mobility is driving my future as an IT executive, which company so much believes in that, has leading positions and leading technology and great people in each of those categories and they're providing me what I demand in this new world, choice? These are whole new worlds and I don't want to get locked in. I don't want to get locked in on what phone I use. I don't want to get locked in on what cloud I'm going to use. I could go on and on.
The [federation] was set up to do that. Then again, go back to the first premise, when you give people clarity of mission and they can get more direct reward for what they're doing. The VMware people get paid on what VMware is doing, not the big EMC. The same thing with EMC II, the same thing with RSA. You attract a better talent base to accomplish your missions.
Computerworld: Often, when you hear executives talk at EMC they will refer to EMC becoming more of a software company. Why is it important for people to know that about EMC?
Tucci: The margin is much more in software. Technology is much more software-based. If we have 15,000, round numbers, engineers in the company -- the developers who develop these great products - 500 do hardware, 14,500 do software. We try more and more to use as much COTS (commercial off-the-shelf) as we can. For instance, in our latest storage products, we've taken all the ASICs out. We're using as close to vanilla as we can. But don't get me wrong. There's a lot of value in the packaging, serviceability, how you design for reliability, so those 500 engineers do great things, but we're not using different components. Ergo, the value that a customer sees is very software-driven.
Computerworld: Some pundits talk about private and hybrid clouds as kind of a myth, or at least an interim myth, because corporate IT won't ever really be able to match the economics of public cloud. They say it would make sense for most customers to really start making a more aggressive move to the cloud as opposed to trying to put that off through private or hybrid. What is your view on that? Are these real workable, viable models for the long term?
Tucci: Well, I agree with what you say. Customers should have plans on how they're going to move 100% to the cloud. Obviously, that's a long journey. But I would define the cloud differently than you would. You would define a cloud as something that sits on the public.
Computerworld: By cloud I mean highly scalable up and down, the economics are great, prices going down as opposed to going up, very easy to manage, easy to deploy, self-service. Those seem to be capabilities that are difficult for companies to match internally compared to public cloud providers.
Tucci: I disagree. The reason it's tough to match internally is because if you go to, say, a big bank, what do they have? They have mainframes, probably two or three flavors of UNIX, etc. And you're right. That's not homogenized, and to help get some standardization they've bought all kinds of management tools. Not unusual for a big Fortune 500 company to have 100 different tools they use to manage their infrastructure. It's hard. But say I'm going to standardize; every cloud that I know of is standardized on x86 technology, all the successful big clouds. I mean Amazon, Google, Facebook, Twitter, on and on.
Secondly, you want to abstract. You want to pool your resources. I've got pools of memory, pools of storage of different kinds with different capabilities. I've got pools of network capability and then basically, based on what the application actually needs at that exact moment, I will pull those resources out of those pools, assign them to that application and then as soon as the application lightens up or is done, there's a whole different load. You're throwing the cores, the memory, everything back into the pools and something else is using them.
If you do that it's very, very efficient, incredibly inexpensive, and it's kind of seductive. Some of our guys call it the California Hotel. It's easy to check in but it's very difficult to check out. You use more and more and more and when you get the bill you go: Holy cow.
I believe that if you're of some size -- I won't even define "some" -- but if you're of some size, you could build a very compatible, very cost-effective, maybe more cost-effective [private cloud]. And then when you hit a peak you don't want to go buy more infrastructure, you can federate some workloads into a compatible public cloud.
That's hybrid cloud computing. I think a big reason to do that for a long time would be security. There is nobody that will be able to protect your data like you will. You get a denial-of-service attack, you can shut down. It's yours. It's under all your control. You can do what you need to do.
Economically we can prove it to you. You can stand up a cloud that is very cost-effective [compared] to anything you can buy publicly. But what you don't want to do is what customers are doing who grew up in the client-server environment or the mainframe environment, where you're overprovisioning, buying for the peaks. That's what hybrid cloud computing is.
We believe there's a huge opportunity in private clouds, huge opportunity in public cloud, but what customers really want is hybrid cloud. You design your base for lower than the midpoint of your workloads and when you hit peaks you'll federate into a public cloud.
Computerworld: It's awkward to ask a CEO to second-guess his customers, but what do you think CIOs should be doing differently, faster or better?
Tucci: I don't second-guess them because they've got a huge problem. They too understand the benefits of cloud computing, but they've got applications that run their business that were written years ago and they've got thousands of them per CIO. In some cases tens of thousands of them. It's just not so easy to snap your fingers and say: You're on a mainframe today; you're going to run on an x86 cloud tomorrow. It's just not that easy. You're on an Oracle database. I want you on a Hadoop-based information system. It's not that easy.
Their CEOs and their board [members] are reading all these cloud things and saying: Why don't we do that, why is your core infrastructure like it is? I understand the pressures they're under every day but I think what we can help CIOs do, and what they should do, is plan out their journey and explain their journey. Say it's going to be four years. In four years I can be here, or in five years I can be here, or three years I can be here. Timing will change. In 10 years I could be here.
But if you don't plan a journey and you don't start the journey, you never end up at a destination. You need a destination, just like you're taking a trip. If not, you just walk around in a circle. So you've got to plan your destination. Here's where I want to be. Then you've got to back up some. Here are the steps on how I'm going to get there. And they've got to explain the economics all the way along. And then I can do some green-fielding and anything new I do is going to be here.
Computerworld: From an EMC perspective, how concerned are you about the move to the cloud? Are you concerned about the white-box phenomenon of cloud providers who would not set up infrastructure the way traditional IT shops set up infrastructure?
Tucci: That's what Amazon does. That's what Google does. But what we forget is they have a tremendous amount of talent and money; they pay for that talent that set that up for them. Because, let's face it, there hasn't been a period of years where a technology component didn't have a severe latent defect. When that shows up, you've got to replace 5,000 [of them], you've got to keep the genealogy of what you bought, you've got to keep setting it up. If you don't set it up exactly right... Well, it's not free to do that.
So yes, you can just go buy the white box, but time and time again customers start out that way and say: Can you do that for me? Because that was a lot harder than I thought it was, and it's a whole set of resources I've got to acquire [that don't have] anything to do with my business. We need people that are going to help run our business, not people who are not going to add a lot of value.
So yes, you could build up that talent yourself, which is exactly what the big cloud guys have done. Or you can come to a company like EMC and we'll do it for you. Our fastest growing and biggest business is to cloud providers. So why don't they do it? Well, I just told you why. They have so much to do that if I want to hire 100 people, do I want 100 people worried about the infrastructure, or do I want 100 people developing my next app, or 100 people working with one of my business units to use software-defined something or other, a software-powered business model that's going to be highly disruptive?
You think of a company like Uber. They're driving all the cab drivers and all the limo drivers nuts. What have they done? The power is in their software. So if you're a company like Uber, is the real power in developing infrastructure or is the real power in the software?
Computerworld: I want to talk about how you navigate the competitive landscape. You've got some big traditional storage competitors and at the same time you've got a slew of upstarts that are attacking those seams that you mentioned or new market opportunities. How do you balance the commitment to the technology paths that you're laying out for customers with the kind of rapid innovation that you're seeing on that startup front?
Tucci: First of all, startups have no magic. ViPR was done by an incredible group of engineers, the leader of which was in Seattle, and he did a tremendous job, built a tremendous team. We funded him to go disrupt, purely go disrupt. That's an example of one of the hottest software products, software-defined storage products, out there that we developed ourselves with a very novel team saying: Just go do that.
Computerworld: So a skunkworks-type project.
Tucci: It's not a skunkworks; it was very well-defined, very well-funded. Don't worry about any of our existing [products]. Go do that the way it should be done. Don't worry about carrying any legacy. You fit in the legacy later, not as the design point. You design it to be the best software-defined storage and then later figure out how I make my legacy work with that.
The second thing is you buy things like XtremIO [Ed. Note: A flash array company.] or ScaleIO [Ed. Note: A software-only storage-area network company]. Right now a lot of companies make a lot of noise on the flash side but there's no company that has grown even close to what XtremIO is growing at. There's a case of two huge winners, one we developed internally called ViPR and one we bought. By the way, we bought it before it had a dollar of revenue so we funded it heavily and actually delayed it because we wanted them to add some even additional features which they hadn't planned. We kept the original development team and it's been a home-run success. ScaleIO is another one.
There are three kinds of innovation. There's incremental innovation, disruptive innovation that you do internally and disruptive innovation that you get from outside. What you don't want to do is buy incremental, not disruptive, from the outside. By the way, if you do three rounds of incremental innovation you've probably taken a massive step and it probably is disruptive.
Computerworld: Companies are always buying other companies for that external innovation, but it seems like big companies fail around that dynamic internal innovation. So how do you spark that? How do you generate that?
Tucci: You just build it in your culture. Basically it's the old adage. If you don't allow people to fail, you're not going to get innovation. What percentage of startups is successful? I don't know the answer to that but it's small. Very small. You've got to allow failure. You've got to let the engineers dream.
When you have an existing product with a big base you have a lot of requirements from that base so they're having more input, whereas these guys don't care about a base because they don't have one. They get a clean sheet of paper and sometimes, most times, that clean sheet of paper is wrong, but when they get it right it's pretty amazing. I can tell you pointblank that I didn't have many customers coming to me saying: Hey, you know what I need? I need somebody to come along and virtualize my microprocessor.
You've been around a long time. Did you ever hear anybody asking you for that?
Tucci: I can tell you I've been around a long time and I didn't hear anybody asking me for that. I actually found out about VMware from a customer. They said it was the coolest thing they had ever seen. Come look at it, I need you to support it. And I actually did more than that, as you noticed.
Tucci: It was pointed out to me from a customer, but even that customer would tell you that they didn't dream back here, you know what I need? I need that microprocessor virtualized. If you listen to your customers and 75% of your customers say I need this feature, you put that feature in. That's incremental innovation.
But it's important to do all three, and I think a lot of companies don't know how to do two. I think all big companies know how to do one, incremental innovation. If you have a big enough war chest you can always buy a company, [number three].
The question is identifying a company. Importing the company and keeping the talent is a different question. But very few companies do two well, where you can really say: Hey, you guys, go disrupt. Use your brain. I'm not going to tell you what to do. Here's a bunch of money. Go disrupt. But very few companies do that well.
Computerworld: Well, I'd suggest a couple of reasons. Big companies are afraid of cannibalization or they like the idea of failure much more than the reality of failure when teams set out to do things.
Tucci: I think you're right. Plus it's a hard case to prove. I mean you go to a board and you get some say ROI guys, and there's no case. It's not: Here's a huge business case, go do it. You're seeing something that might be good and you're saying: You're kind of special. Here, go out and go do it. I'll fund you.
Computerworld: I want to go back to that flash memory market. You said, and these are your quotes: The competitors "make all kinds of stupid statements" about flash, and that a lot of this discussion is "bunk".
Tucci: Sounds like me.
Computerworld: Well, in light of that, lay out for customers how you see that market developing and what they should be thinking about when it comes to flash.
Tucci: Basically, forever and a day, there's going to be what I call a persistent tier of storage, which you don't need high performance from. Look at the cost curves of this technology, spinning disk, and look at the cost curves of flash. We're shipping in volume now 6TB drives, that's before you dedupe it. Incredibly low cost. Not the fastest access time, but it's still direct access.
So I'm saying there's going to be a need for that for a long, long time, this persistent tier of storage. People talk about consumer flash and deduping, but you can dedupe regular disk and most deduping happens in memory anyhow. So there's going to be a persistent tier, then there's going to be probably a flash tier and in-memory. [SAP's] HANA talks a lot about in-memory and tons of talk about flash and you can do that on one array too. What I'm saying is a lot of these flash guys, flash or zip, they forget about in-memory, they forget about persistence. For some applications, like a rendering of your desktop, maybe 90% of that is duplicated and because of the duplication properties, maybe that is a use case. But they'll talk about a corner use case as if it was prevalent. But in the database market that's not true. In the Hadoop market that's not true. These things are going to persist for a long, long, long time.
You want to have the ability to have it here and here, and here is going to be some flash, mostly disk, here is going to be all flash and here is going to be in-memory. And now you work those tiers. That's what customers need and that's what customers want.
I'm saying when you listen to some of these flash startup guys it's like one size fits all. It's like I'm only going to build size-8 shoes and if you've got a size 10 foot I'm going to stretch it. If you've got a size 5 shoe I'm going to put a lot of cotton in the front. It's ridiculous. That's all I said.
Computerworld: As I listen to you it sounds like it's less about the price of that kind of memory driving its expansion than the applicability to all of the applications.
Tucci: What they'll do is they'll take disk and put it in a constant and then they'll say: Here's the price curve for flash. Ergo, I'm going to be very competitive. But if you talk to the disk guys, and we talk to both -- I think we sell more of both than anybody else -- they're coming down too. You can't take a static view. I think for the next five years, probably longer -- five years I can guarantee -- you're going to see disk technology, you're going to see flash technology, you're going to see memory technologies. Customers are going to want them all. Therefore, the one that can do the best job of moving data between those tiers is going to be the winner.
Computerworld: You've talked about ViPR, but what's the reality of software-defined storage today? How far along are we? Virtualization at the compute layer is pretty well understood.
Tucci: First of all, it's much more difficult in storage. To load everything you're going to put in compute is pretty quick. If you have a 2-petabyte information base and something happens and you have to reload that, you're in trouble. That's a lot of time. So you've got to do things differently. It's just hard. So if this was a baseball game we're in the early innings. And everybody defines everything they do [differently]. I can say everything we do is software-defined storage and I would be correct because I just told you we're taking out the ASICs, we're trying to use as much common hardware as we can. Ergo, that's software-defined. But real software-defined is how do I work with any environment?
Tucci: So I could have a VMware environment. I could have an OpenStack environment. I could have a Hyper-V environment. I could have a bare-metal environment, I could use containers, whatever I want. But no matter what you have, you could still use software-defined storage technology. So I think that's one of the factors that I would use for my definition.
Behind that there's billions and billions of dollars going into array technology and some of it is good. You don't want to destroy that. You don't want to put something in the middle that makes everything behind it look like a white box, because then you're destroying IP that might be valuable. You want something that talks down and says: What capabilities can you do? And if you do it well, I'll just get out of the way. I'll just pass it right to you and you do it.
Computerworld: It's not intelligence talking to dumb. It's intelligence talking to intelligence.
Tucci: That's my point. You want to talk intelligently in a policy-based way to a lot of different environments, not just one environment. That gives you choice. Down here I want to talk to intelligent and non-intelligent devices because you want to use the IP and somebody does something really, really snappy in storage, you're saying: Well, I'll use it. I'll just pass the I/O straight through and give it to them. That is not easy, but that's the design construct that we used.
Computerworld: Look ahead to 2015. What should IT executives expect from EMC?
Tucci: Hopefully they'll see a lot more innovation. You're going to see us continue to focus on the software-defined data center. You're going to see us continue to focus on converged infrastructures, you're going to see us continue to focus on storage, we're going to continue to focus on putting less differentiation in hardware and more differentiation in software and, hopefully, from my point of view, it's all in the user experience, where customers can say we give them the best value, the best service, the best support.