In 2015, tech consulting outfit Accenture set upon an ambitious cloud computing plan: to take the bulk of its business...
and IT operations into the public cloud in a few short years -- from 10% cloud-supported to 90%.
Preposterously optimistic? Hogwash. Three years in, the global company is within spitting distance of its goal, said Merim Becirovic, managing director for global infrastructure. A clear, uncompromising vision of cloud as the future of IT -- plus sheer will and tenacity -- got it there.
"We laid out the program early; we set out a strategy from the get-go on how we wanted to approach it; we had the right buy-in from the business," Becirovic said. He spoke earlier this month in an interview about the global company's cloud computing plan.
Accenture's breakneck speed was possible, Becirovic said, because it had put much of its infrastructure and applications onto virtual machines in the years before its cloud move. And that "put us in a nice position to be able to move aggressively to the cloud," he said. Once the company started moving to the cloud and closing data centers, it didn't look back. It plans to complete its move in December.
"I'm not going to tell you it was perfect -- it's new, and people were trying to understand how it works. But we built momentum, and we learned along the way. Today, it's a finely tuned machine that continues to run."
Becirovic discussed Accenture's cloud computing plan, the knowledge it has gained and how it is transferring that knowledge to its clients. Here are excerpts of that conversation.
Tell me about Accenture's three-year cloud computing plan. How did it start?
Merim Becirovic: Right now, we are 86% in the public cloud -- and our goal is to get about 90% in the public cloud. One of the things for us is, the business we're in, we're telling our clients to go to the cloud, so we as an IT shop supporting Accenture felt it was our duty to do that first.
We went in very fast. When we started the program, we told folks, 'Hey, look, we're going to the cloud. We're going to take the business to the cloud.' Our mantra was cloud-first, cloud-only moving forward. And if you think you need an exception, well, let's us talk about it. So that's how we started the program. Then, we organized around moving all of our existing workloads out of data centers and knocking them off one by one and making it into the public cloud. For our custom applications, we have a two-provider [Amazon Web Services and Microsoft Azure] cloud approach. We're never going to be on one cloud. And then obviously as we consume platforms, those platforms exist in their own providers' environments.
Platforms -- so you have a lot of SaaS applications?
Becirovic: We do a lot of SaaS; we do a lot of cloud-native. We run the whole gamut. As we've been in the cloud, we have continued to say we want to consume platforms and not build the infrastructure ourselves. If we can consume it as a platform, we will. One of the other hats I wear is, I own all of our enterprise architecture. Any new capability that we're going to develop that's custom that can't be supported by a platform, it's going to be cloud-native. So it's going to be not built on [infrastructure as a service], but literally consuming services, microservices, by our cloud providers.
For example, we run a tool called Performance Achievement, which is our global tool that supports all of our performance feedback. So, working with our HR organization -- we have been out there now two years -- this is the custom tool we built to support our business that is completely running on a cloud-native infrastructure.
Why microservices instead of IaaS?
Becirovic: I don't want more IaaS proliferation, in the simplest terms. Every server spun up requires patching and OS upgrades, scanning, monitoring, etc. Ideally, as I build more applications based on microservices, I am getting out of the business of server maintenance and the like and have the provider take care of those services. I can then keep teams focused on creating new capabilities. Our current IaaS footprint is our legacy .NET applications we have built. As we reimagine those services, we will eventually refactor those applications in a microservice-based architecture if we cannot put them on a SaaS platform.
What enabled you to move so quickly, and how were you able to sustain that pace?
Becirovic: We were early proponents of all the virtualization capabilities -- we were early proponents of pushing all of our applications onto virtual machines. And we standardized heavily on virtual machines, mainly to get all the cost savings that were going to come our way by not having all the physical servers taking up all the footprint in data centers. By having everything virtualized, it put us in a nice position to be able to move aggressively to the cloud. Because then, especially when we were going to lift and shift workloads, it was no different. We didn't change our processes to run those workloads in the cloud. We extended our processes of how we ran things into the cloud -- except, today, I'm just not provisioning capital to go buy new servers or replace aging servers or storage, etc.
Now the other part of this was, we owned the governance and the compliance centrally. So when Andrew Wilson, our CIO, said, 'We're going to go to the cloud,' it was easier to do that because we got the business buy-in and support. So there's a collective unity at the top that says, 'Cloud is the future; we're going to do this.' I'm not going to tell you it was perfect -- it's new, and people were trying to understand how it works. But we built momentum, and we learned along the way. Today, it's a finely tuned machine that continues to run.
What were some other motivations behind Accenture's move to the cloud?
Becirovic: It was really speed and agility, to tell you the truth. If we want to do new things and build new things, how do we do that in a way that we're not restrained by either location or capital constructs that may imply the waiting game? There's no waiting game in the cloud in many ways.
And our teams love it. They're not waiting for the provisioning of an environment or anything like that. They're able to spin up workloads now -- much more efficiently and quickly -- to do a lot of the work that they are on the hook for. But they can do it faster now; they can do it cheaper now. And they can use it when they need it. It really becomes this consumption model, so it eases the burden on trying to figure it out or plan for a perfect capacity.
I grew up as an enterprise architect, and one of the things every enterprise architect will tell you is they build systems that can perform and support scale at peak capacity. And you performance-test everything, and then you design your infrastructure to make sure that you can support that maximum scale, plus some amount of headroom.
When you think about it, how often do you actually need all that capacity? When you're in your own data center, you're building for some small amount of time you need that capacity to be there. And the rest of the year, it's nowhere near that capacity. In the cloud, you don't have that limitation. You can build it for that top capacity but then turn it down for the times that you don't need it. And that's exactly what we do.
When I talk to our clients about how you actually save money in the cloud, it's being very aggressive on the sizing, on the scheduling of environments. How often is it up? How long does it need to be up? Is it the right size? Does it have the right compute footprint? Should it be bigger? Should it be smaller? It's that analytical capability and applying that week in, week out to how you run that infrastructure that we were able to save money.
To learn how cloud computing at Accenture is strengthening IT-business alignment, go to part two of this interview.