Software-Defined Data Centers Offer Flexibility
A few years ago, Choice Hotels International was feeling pretty good about its technology strategy. The Rockville, Md.-based company, which operates 6,300 franchises around the world under 11 brand names, claims to have been the first hotelier in the world to deploy a cloud-based property management system, and, by 2011, more than 80 percent of its servers had been virtualized.
But Choice was being held back by one critical component: its data center.
In this expanding company with seasonal fluctuations, the decades-old 1,500-square-foot facility had evolved into an over-crowded, power-starved nest of segregated silos hard-wired into an inflexible infrastructure. It was located in the same building at Choice's western regional headquarters in Phoenix, which also housed the infrastructure and application development teams.
"Users were definitely seeing that we were at a capacity limit," says Rick Dockall, senior director of infrastructure services. "Requests were met with a lack of resources."
Specifically, the company couldn't meet the dynamic computing needs of its booking engine and property management system. That's when Choice began to think hard about deploying a software-defined data center (SDDC), which would give the company the flexibility and responsiveness it craved.
"It's kind of a no-brainer to take your logical network and collapse it into software," Dockall points out.
Software-defined data centers were in their infancy two years ago, but today they're one of IT's hot trends. A growing number of companies that are striving to be more flexible and agile are finding that controlling data center resources through software is the way to go.
"We're making a transition into an era in which our control becomes much more fine-grained," says Eric Hanselman, chief analyst at 451 Research.
How quickly that transition is happening is unclear, as estimates of the SDDC market are "wild-ass guesses," Hanselman says. It also doesn't help that the definitions vary from one vendor or market researcher to the next. But there's no question that many IT shops in large companies are weighing the benefits of SDDC in preparation for eventual data center modernization.
Hanselman says the move toward SDDC is a reflection of the dramatic shift in the way information moves through an enterprise. The north-west data traffic predominant in database application user hierarchies has been giving way to the east-west nature of today's more modular environments—in which there's more interaction between users and data—and within applications.
"The nice linear flow has gotten hashed up, and data moves more circuitously," he says. "What SDDC promises is the ability to make it much simpler to scale the interconnection capabilities between these different compute instances."
Looking for Flexibility
That kind of flexibility is what Choice was hoping to get from its SDDC. But while it might have been a no-brainer, it was far from simple.
Dockall and two members of his team—Ross Davies, senior virtualization engineer, and Lou Young, director of shared infrastructure services—started by literally drawing up designs on a napkin and then evangelizing the idea to various stakeholders in the business, eventually building the consensus needed to get executive buy-in.
Then Dockall and his team started figuring out how the new infrastructure could be shoehorned into the existing facility. It didn't take long for them to realize that the money was better spent migrating to a co-location facility instead of trying to squeeze the new environment into the existing space.
"Buying generators doesn't help us get more guests in our hotels," Young point out.
Having rethought what the compute, storage and network components would look like, as well as where racks would be located and how they'd be powered, the team spent the next few months setting up the new facility. Then it migrated more than 1,000 servers, of which 850 were virtual, easing physical migration costs.
To do this, they temporarily treated the new data center as an extension of the old one, load-balancing between the two as the migration took place.
There were some change management challenges, as a lot of IT staff members were anxious about data center operations being 12 miles away and largely unstaffed. However, Dockall says the effort—which involved 140 people and came in on time and under budget—has been a huge success that's delivering numerous benefits. Most prominent among them is the ability to respond to fast-changing compute requirements.
For instance, whereas it took two or three months to connect a new distributor into its environment, Choice can now do that in hours once approval has been granted. Plus, now that it's not hampered by an inflexible architecture, the company is able to be as fluid as its franchisees' seasonal needs.
"As we go through the seasons, we don't want to have to stand up new hardware for four months a year," Dockall says. "We want to be able to extend out to the cloud and then pull back."
A Deliberate Choice
Whereas Choice's move was necessitated by a data center that had fallen behind the rest of its IT environment, investment management firm BNY Mellon has been steadily moving toward SDDC by design.
Starting with its move from UNIX technology to commodity hardware in 2010, followed by its development of infrastructure- and platform-as-a-service environments in early 2012, every move BNY Mellon has made has pointed toward simplifying, standardizing, virtualizing, automating and cloud-enabling its infrastructure.
"Software-defined data center architecture can only be achieved with these five things," says Swamy Kocherlakota, head of global infrastructure, architecture and engineering for the New York-based firm, which manages $1.5 trillion in assets in 100 markets around the globe. "That's been our journey over the past four years, and we have made serious progress. We believe we're ready for this notion of a software-defined data center."
BNY Mellon had many goals for its SDDC. It wanted to address its difficulties with technology refreshes, which couldn't be completed fast enough in 52 weekends a year. It also wanted to shrink its 196,000-square-foot data center footprint and get away from the server-hugging mentality of many users.
In addition, the financial firm wanted to give its 13,000-person IT staff the ability to spin up development environments on demand to support business initiatives. More than anything, however, it wanted to get the most out of its already ambitious private cloud environment, which supports more than 3,500 instances.
"I'm a service provider much like an Amazon," says Kocherlakota. "We knew this was how we had to build services, and this is the platform that will allow our business to elevate. SDDC enables business to be more agile and create services in a more nimble and agile way."
Today, everything in BNY Mellon's data centers is managed through an orchestration engine, one of the few proprietary components of its SDDC.
"The only thing that's manual is the developer saying 'I need a stack' and requesting a development instance," Kocherlakota says. "That's done within two hours."
One pending enhancement to its SDDC is a business activity monitoring system that BNY Mellon has been building in-house over the past several months to help with trouble-shooting in an increasingly complex environment. The system is essentially an API link into a Hadoop big data cluster that Kocherlakota says will enable the company to "follow the breadcrumbs" that applications leave behind. By sometime in mid-2014, "every app in the cloud will be using the business activity engine," he adds.
Meanwhile, BNY Mellon's steady march toward an SDDC has paid off by reducing bureaucracy, making developers more productive and cutting the costs of innovation.
"In the past, to pursue a business idea, you had to seek funding, and there was a lot of paperwork," Kocherlakota recalls, noting that a lot of ideas died during the process. "No longer do we have to worry about shelving a project. We're better at bringing vision to reality."
And BNY Mellon isn't done. It still has to commoditize more legacy infrastructure components, train its infrastructure teams further on SDDC and continue shrinking its data center footprint. It won't be satisfied until the environment is just about running itself.
"Everything is auto-configured," says Kocherlakota. "That is really the end game."