Singing in the Rain: How the Cloud Can Pour Down Money
Today's consensus is no longer around an "if" for cloud computing, but the "when" and what types of cloud models are best suited for any particular company. The present challenge, then, is about the proper transitions to leveraging cloud for improved IT and for far better business results.
At VMworld, as part of the main keynote address, one company and its design and implementation of a private cloud rose above the rest. Revlon and its CIO were in the spotlight for such impressive returns on their cloud. In just two years, Revlon has benefited by nearly US$70 million from savings due to cost avoidance and reductions.
Here to tell us about how such savings emerged and to describe one of the most efficient enterprise private cloud implementations in the world is David Giambruno, senior vice president and CIO at Revlon. The discussion is moderated by Dana Gardner, principal analyst at Interarbor Solutions.
Listen to the podcast (20:19 minutes).
Here are some excerpts:
Dana Gardner: Is there a reason for doing private cloud holistically, completely, rather than piecemeal? What's the benefit for doing it that way?
David Giambruno: From a technology standpoint, we look at ourselves as doing oneness. We pick one way and we get very good at. We own that technology, so we can command it. It's really about the density of our skill sets and our capability around that order to execute for the business.
When you look at that, it drives a degree of simplicity of execution, because at the end of the day, what we're really focused on is delivering IT capability back to the business faster, cheaper, better. That's essentially what our cloud was planned to do and has delivered.
Gardner: And this is no small undertaking. It's over 530 applications, 15,000 automated moves a month. Give us a sense of what you've done with this singular approach to full competency at this particular data center and your approach to private cloud?
Giambruno: It's not a data center. It's the globe. That's important for everyone to understand. Revlon's cloud covers all of Revlon's presences globally. It's not just a single data center. We have a core data center and then we have little data centers around the world that everything replicates between and things move between.
We've built this entire ecosystem to deliver our applications. We started this about five years ago with this whole idea of oneness. We re-ITed the planet. We have one DNS and DHCP structure. We have one global directory. We have one SAN globally. We have one desktop image. We have one server image. That simplicity allowed us to use the cloud as a competitive advantage.
We've saved or avoided $70 million in last two years. If you go by a simple timeline, the first 18 months was laying that oneness foundation. We did that in 18 months. The second 18 months was the virtualization of the servers, the network, and the storage systems globally.
At the end of the 18 months, we were done. That was the first three years. We've been running this way for the last two years. We're not in the "I think" mode. We're looking now at how we continually extend the capability of our cloud.
Gardner: Our listeners might be familiar with Revlon, your brand, but tell us more about the scope of your operations and the extent to which IT is supporting your business.
Giambruno: Revlon is a global cosmetics, hair color, beauty tools, fragrance, skincare, anti-perspirant/deodorants, and beauty care company. The vision of Revlon is to deliver glamor, excitement, and innovation through high-quality products at affordable prices.
We are arguably one of the strongest consumer franchises in the world. Our brand is incredibly powerful. We've got offices around the world. Our global headquarters are in New York. Our flagship manufacturing facility is in Oxford, North Carolina, and our consumers are women around the world. Our products are sold in more than 100 countries.
So we are big, as far as our reach and our capability. Essentially, our cloud delivers roughly 95 percent of all Revlon IT services around the world. We've got a couple of systems that aren't in there yet. They will be shortly, but for all intents and purposes, we operate everything off of our cloud.
Gardner: Let's go back to how you got to this point and how you're able to enjoy such significant savings. You have a comprehensive virtualized approach of servers, network, applications, and services. Why is that important?
Giambruno: Again, it's that density of skill sets. Through this whole implementation, we only used about 10 percent professional services. We didn't spend any additional money, other than our normal capital refresh. The thing that we did was change the way we're spending our money.
We took that leap to do things differently, because at the end of the day -- I always say this just to keep my and my team's frame of reference -- we make cosmetics and personal care products. We have lots of brands, but it's the idea of simplicity.
We're not a revenue-generating piece of Revlon. How we can add value back to the business is by doing things faster, better, cheaper? If we're not spending that money, we're avoiding spending money, or giving money back, that means it can go into new product development. It can go into R&D. It can go into marketing. All activities focused on driving profitable growth.
Getting technology to facilitate the business and do things faster and more effectively is really important. To me, it's the most material thing we've done - if you look at your projects. We've increased the number of projects we complete every year by 300 percent. When you talk about the business alignment, getting what they want done faster, cheaper, better, to me, that's it.
Gardner: And you're talking about spanning the cycle from full development to implementation. What's the role that the cloud has played in terms of increasing the ease in which you move from development to operations?
Giambruno: I have a couple of buckets. We have reliability. Currently, our cloud has been operating at north of six nines uptime, which has allowed me to take resources out of operations, put them into projects and working with the business.
That's resulted in speed. If you want a server, if there is a demand for new application or testing something, our cycle time for getting a server up is anywhere from 15-20 minutes and there isn't the associated cost. For us, a server is just a file. If you want one, great, here you go.
And we manage capacity on the top line. So we essentially move that infrastructure barrier and cost. We've disconnected it. One of the greatest things that we monitor is our ratio of physical to logical servers. When we started this, when we first went live three-and-a-half or four years ago, our server ratio was 1:7. We are now 1:34. That's essentially a 500 percent increase in capacity without the cost.
That makes a material difference in the business not having to pay for things. The speed at which we can nail up applications and the accuracy at which we can do it has made a material difference in our ability to deliver projects to the business.
Gardner: In addition to improving this cycle for development flexibility and resources, you've also devoted significant improvements to disaster recovery (DR). Tell me a little bit about why the private cloud has helped you in DR.
Giambruno: One of the things that we've learned very quickly was rate of change. When you're on a cloud, every time someone hits a keystroke on a keyboard, that's a change in your cloud. Our rate of change is anywhere between 20-30 terabytes a week.
We made a conscious decision as we don't tier anything in DR; we literally copy everything. There are two pieces of things. I'm most impressed with what my team has done.
One is if you take that rate of change and attach it to storage growth, you're roughly at $27 million a year. Through a series of technologies that we employ, we turn that $27 million into $400,000 of storage that we actually have to pay for. So, our shareholders get that benefit, because I don't think anybody else's shareholders would have that interest in place.
The second thing is that it does allow us to copy everything. Roughly a month ago, we lost our factory in Venezuela to a fire. Fortunately, no one was hurt, but from the time someone made a phone call, two hours and 40 minutes later -- and roughly two hours of the time of finding people, because it was a Sunday afternoon -- we moved the country of Venezuela up into our DR side, had everything running, and we're giving the users virtual desktops so they could keep working. That's the power.
Gardner: Peace of mind and trust.
Giambruno: And it's not fake. We've done it. Globally, we are minus-15 minutes replication in their stuff. That's a little longer or little shorter depending where it is and time of the day, but it goes back to the simplicity. We just copy everything so we don't have to worry about it.
Gardner: All right. Let's see metrics-wise what this gets for you in data reduction. What sort of volumes have you been able to improve?
Giambruno: We've run about 96 percent data reduction for everything from compression and de-duplication. As we've gone through this, we've also learned that with different storage protocols, block versus CIFS, you get better compression. Running at NFS you pick up 15 percent utilization over block.
Everybody has different business cases for why they need either, but we keep finding ways to squeak more out, because, again, the less money we can use, the better for the business. The more efficient and effective we are, the better for the business, and the less they have to spend on this.
Conversely, we keep leveraging those capabilities in extending our cloud. So we can sling a Windows 7 desktop to an iPad, or we're enabling our cloud so people can use resources wherever they are, regardless of the device. That just makes their lives easier and their ability to do business better, so we can support people growing the company.
Gardner: It's really impressive to me, David that the more value that you derive from you architecture and approach, the more it contributes to other things. For example, what you've described is great for DR, but you're also reducing your racks, restructuring your server licensing, and also getting to improve asset utilization. So it's sort of a snowball, but in a virtuous way.
Giambruno: It's interesting because in our DR site we run our test and dev. So the asset is never cold. We're actually using the virtual servers while they are not being used for DR to run all our tests and dev. It just contributes to the uptime. The data is already there.
We reuse assets all the time, and as we go forward, we have plans to go active-active. So now end-of-life servers that are coming out for maintenance, we just throw them in DR, because they can just stay there forever. It doesn't matter to us if one dies a year. So what? It's really that ability to keep using those assets to extend capabilities.
Gardner: How about the stack? Can you describe some of your products and what they've done for you? Are you venturing to some new areas around either management or governance to try to continue to tweak this to get more bang for the buck?
Giambruno: The bang for the buck for us is that we're working really hard on essentially creating an internal marketplace, like the Apple marketplace or the Android marketplace.
We've got desktop virtualization, but we see huge value to the business in creating this internal marketplace. We know a user. We know their devices. We know the applications they're supposed to be using. Depending on the device that they connect, we can format the application they are using and its view to that device to deliver them in context.
To some degree, it's like going from a LAN to a trusted WAN, where we know the device that's registered to you. We know you as a user so we can deliver very securely your information, and that information never leaves my data center. So you are only ever viewing the information they are working with.
Gardner: You're also now creating an application marketplace. How does that benefit from your cloud infrastructure?
Giambruno: Our cloud can send them anything. The applications are already running on the cloud. Essentially, when that device comes out, our cloud will understand context. We'll be able to deliver that application in the context of the person. You've got a highly secure environment. We're not moving data anywhere. We've got control of the device. We understand who the person is, and so we can deliver in context what they are supposed to have access to, regardless of where they are.
If you have an iPad or anything like that, you have an icon on the front. You'll have a Revlon marketplace. You open that up, and there will be a list of applications that you have access to that are already authorized for you to have access to, and we will start sending you those applications.
Gardner: What's your advice for folks getting started?
Giambruno: I tend to live by "isms" to make very clear pictures, because I had to move own organization through them. Two things: trust or verify, which maps into the second, which is just try. They are very symbiotic.
As you look at this, just try, and as you go through that, trust and verify that you're delivering the capabilities that the business needs and that you know they need. As you go along that path, you can build trust and confidence in yourself and your capabilities, your team can build trust and confidence, and you can show that to your business units.
That's like that snowball that you get rolling. Once everybody realizes that it can be done, it's more of a human experience thing than it is the technology. The technology works, and we've been doing this for a couple of years. I couldn't imagine operating any other way any longer until the next big geometry train comes, but that's probably another 10 or 15 years.