The technology giant’s endeavors in virtualization led to a full-scale transformation of how IT functions within the company. EMC CIO Sanjay Mirchandani explains the strategy—including the ups and downs—behind the company’s private cloud journey.


WORKFORCE NEWS: How did this whole private cloud transformation come about? 

When we started out about four years ago on this journey, it was never called the “private cloud.” It was really a journey in virtualization—down the path of virtualization, down the path of information management, and trying to get useful life out of our data centers. If you think back to 2006-2007, we were growing and acquiring companies. We had to integrate them. We were growing organically. We were spending on R&D. We were growing our footprint. All of these challenges of growth and integration, and growing data centers is not something we take lightly—we had to think it through. We had a plan we were trying to meet, which had a timeframe of this year. I wasn’t the CIO at the time, but at the time, we weren’t fully utilized. In fact, utilization was reasonably low, not unlike most companies of our size and nature.

So the team did what, in retrospect, were some of the smartest moves—basic blocking and tackling to get cooling and power in order, and then they started doing aggressive virtualization on the server side. And they were doing information lifecycle management on the storage side.

What we saw was a transformation of how things we run, deployed, governed, etc. Subsequently, the industry—including EMC and VMware—saw that what we were beginning on with 100% or close to 100% virtual infrastructure had the tenets of what was being done in cloud computing—particularly public cloud computing. You see the Amazon cloud services, the Google cloud services—you wanted the same tenets inside your data center, and hence the term “private cloud.” We didn’t coin the phrase, but we took a step back and said, “Everything we’ve been doing is really getting us toward the private cloud.” It was by circumstance that the industry evolved and the technology evolved. We jumped on the bandwagon in our second “crank,” in 2008, 2009 and this year, as we continued on this journey.

What were the phases of the project?

1. Low-hanging fruit; dipping your toes in virtualization. We called this “IT production.”
2. Taking on more mission-critical and business-critical applications where we started sharing infrastructure across applications and doing more of a shared model across storage-compute networks, etc. That phase we termed “business production.” In most companies, when you end that phase, we think we’re at about 60% to 65% virtualized on the OS side.
3. What we’re all trying to get to, as fast as possible, is IT as a service: self-service, provisioning, metering, billing, true costs back to the business, etc. We’re at different points in that phase in trying to get to all aspects of the private cloud.

Talk about how you worked with the business to communicate what you wanted to do and get their input on timelines, returns, etc.

The technology really took rapid shape over the last 24 to 30 months. Some aspects were actually easier because we’re in the business of helping customers build out cloud infrastructure. So I use that internally in conversations where we’re looking at demand over the last two years. We’re trying to rationalize spend, rationalize what projects we’re working on, and we’ve got to share more across the infrastructure. We’ve got to get smarter to get better utilization. So I’m putting in more private cloud infrastructure. That’s where we really kicked into the business production phase, where IT had to bring different applications to different groups, and share and consolidate infrastructure across the business—be it middleware, be it tools, etc.

The conversations we started having with the business were around, “Let us give this to you cheaper, because we want to do it faster. And, oh yeah, you’ll be sharing it with different groups, but don’t worry—we’ll manage the peaks and valleys and how it all gets used.” And we knew we wouldn’t charge them more upfront; we had to feel it out, but instinctively they knew it would be cheaper, but we wanted to see how we could expand capacity, utilization and so forth.

So the arrangement was about not raising the costs to the business for 2009 and 2010, and trying to do it with less. We also shared the fact that our focus on managing demand overall was, what classification of applications could be considered enterprise-wide? What was specific to compliance and things that we have to do? And the rest was business unit-specific spend. And, by the way, even with that, we’ll share within business units and share infrastructure across business units. Simultaneously we standardized, we took stuff out of the non-standard mix—that was IT for IT, just taking care of things behind the scenes.

So that’s how we did it: We used demand, we used budget pressures, and we used the fact that they wanted it quicker. A little bit of a carrot, a little bit of help, I guess. We worked on the timelines overtly with the business in 2009 and now in 2010.

What about the obstacles? What key roadblocks did you hit through the various phases? What lessons did you learn that would help other IT organizations as they initiate similar transformations?

We reflect on that every day. Not to sound cliché, but this thing is a journey. It took us years to get as far as we have in the data center. We’re not going to achieve everything the cloud promises overnight.

It’s a journey on multiple fronts. The bits have to come together in the right way. And our people have to adopt them in the right way. I’d say the former is almost easier than the latter. With standards and innovation and R&D, the bits will come out fast and furious; we have to take a step back and figure out how we get our people to transform and absorb and understand the opportunities that can take us to the next level. So the point I’d stress is that it’s a journey: it’s a journey in terms of the technology and in terms of the people and the organization. A lot of people—including me—will think more about the bits than about the cloud and how you get this stuff absorbed. In the last six to eight months it’s been apparent to me that as well intentioned as the roadmaps are, we have to make sure we don’t leave anyone behind.

This is not a project you want to embark on without a solid level of corporate executive sponsorship. Whatever your line of business, whatever your industry, you have to make sure that the folks who run the line of business understand that it is a shared asset you’re building to give the business agility, lower costs and transformational capability, but they’re going to have to share, and they’re going to have to better understand peaks and valleys in their business. In the old days, we built for peak loads, and you overcompensated and hoped that everything would be okay on Christmas day, or Monday morning, or whenever you peak load was. In this case, there is no recourse—you have to provision a plan for peak loads collectively across the enterprise, so people need to talk more, and you need the executive sponsorship to around it, and you need to stay the course. It’s a bigger conversation with the business.

The other thing I say a little tongue-in-cheek is after a long time, you’ll see state-of-the-art infrastructure actually leading applications, as opposed to applications dragging infrastructure. IT guys care about it, but the business didn’t. Now, with the state-of-the-art shared asset that is your private cloud, you have agility—it becomes a corporate asset. You want functionality—the functionality has been dropped into the cloud. The business will say: “I want this in the cloud.” After a significant time, this is where applications have been dropped into the cloud, as opposed to the way it was done traditionally. That changes the conversation with the business even more.

As you said, it’s a journey. What’s left to do?

There’s a lot left to do. We should have a very significant piece of our x86 architecture virtualized by Q1 of next year. I’m pushing for all of it, but realistically we’re driving toward 85% to 90%. In addition, we’ve started the journey down storage virtualization. You have to look at different technologies, like V-Plex. We’re looking at 2011 as being a pivotal year for us where the storage layer gets virtualization done on it. The third piece we’re dipping our toes in is virtual desktops and next-generation client experience. We’re not calling “desktop experience”—we’re calling it “client experience,” because we have to think about provisioning the user experience, not the device. Different users will have different experiences based on different devices, but we have to provision that experience for them. We’re doing some thinking on that. We’re looking at the next client experience as being a largely virtualized one.

Management tools need to be rolled in and evolved. Applications need to be looked at as they come up for renewal. We have to see how they can be put into the cloud. We’re looking at different applications and rewriting some applications. There are layers of things being looked at, but immediate deadlines are to get as much of the x86 architecture virtualized, continue down parallel storage, look at virtual desktops, look at new tools, and simultaneously build applications into the cloud, and use those experiences to help our people write the new frameworks. As much as it sounds sequential, you do a lot of these bits in parallel.

Also see the second half of our interview with EMC CIO Sanjay Mirchandani. 

Back to Workforce News