Partway into a big data center consolidation project, a large mobile data services provider has looked beyond simple...
server virtualization and turned to storage and network virtualization and, eventually, to cloud computing to get the job done.
Bellevue, Wash.-based Motricity now uses server virtualization technology from VMware Inc. to consolidate its servers and data centers, which had ballooned to 12 when it purchased a competitor in 2007. It now runs a more manageable six data centers across the U.S. running about 1,000 virtual machines on 100 ESX hosts, serving up things like text messages and ringtones for mobile telecommunications vendors.
Eventually, Brian Jeide, Motricity's director of architecture, hopes to have just one data center. "I have lofty goals," Jeide said.Turning to the cloud for disaster recovery
But server virtualization alone will get Motricity only so far. For disaster recovery purposes, Jeide insists on having at least two data centers: one primary, and one for failover.
To consolidate down to a single data center, Motricity needs to sign with a hosting provider that can take over in case of disaster. The problem is, Motricity's services are "keyed to an IP address or storage [device]," Jeide said. To meet his goal, "I need to be able to replicate services to another [failover] data center without involving the developer team," he said.
In other words, "If I want a single data center, I need to make my services portable," Jeide said.
Virtualized storage technologies from NetApp and Cisco's Nexus switch platform should help separate Motrocity's data services from the location of the underlying physical infrastructure, Jeide said. For standby compute resources, Motricity is looking at DR services from INX and, conceivably, the public cloud.
That's still a good two years away, though. In the meantime, Jeide is educating different Motricity business units about the true cost of running their services. He uses an IT cost modeling service from Apptio, including a new server virtualization template that was released this week. "I can show them, by utilizing our servers at 60% to 70%, we could get everything in to our largest, lowest-cost data center, and we could save tens of millions of dollars," Jeide said.
Getting started with cost modeling in Apptio was hard, Jeide said. "It took us a while to get organized; each data center had its own Excel spreadsheet." But once the initial template was built, getting the data modeled was easy, he said.
The work is paying off. "When you have data in a place that you can start to do analytics, it changes things," Jeide said.
Let us know what you think about the story; email: Alex Barrett, News Director.