Smooth data-center move sets template for disaster recovery
Alaska’s Enterprise Technology Services team used a unified computing system to quickly move existing services to a new data center, and validated its disaster recovery approach and failover capabilities in the process.
Alaska’s Enterprise Technology Services team relied on a unified computing system to quickly move existing services to a new data center in Anchorage, validating its disaster recovery approach and failover capabilities, according to the state’s top enterprise architect.
Disaster recovery is a big focus for Gov. Sean Parnell as well as ETS, said Corey Kos, enterprise architect with Enterprise Technology Services, a division of Alaska’s Department of Administration.
“I had been burned by tape-based backup before,” Kos said. So the goal was to deploy a system that didn’t use tape and was quick and easy to set up.
The system, Cisco Unified Computing System, serves as the platform for a private cloud by which ETS delivers enterprise services such as e-mail and e-mail archiving to state agencies.
Cisco UCS integrates x86 blade servers, access and storage networking, virtualization and management capabilities into a single system. The goal is to eliminate redundant devices and layers of management complexity.
Alaska ETS deployed FlexPod, an integrated combination of Cisco UCS computing, VMware virtualization and NetApp storage devices that are connected through Cisco Unified Fabric as a way to provide secure delivery of IT services to users.
Alaska has three primary data centers: in Anchorage, Fairbanks and Juneau, the state capital. The Anchorage and Juneau data centers are basically mirrors of each other. Fairbanks has a smaller data center that switches over to Anchorage in a crisis or outage.
In December 2011, ETS was able to validate how UCS can help quickly migrate existing services when the state relocated its Anchorage data center to a new site.
The ETS team moved computer services from the older Anchorage data center to the location in Juneau over the weekend. It took the team 18 to 20 hours on a Sunday to complete the task. On Monday the team made sure the environment was stable and on Tuesday they tore down the unified environment and moved to the new facility in Anchorage.
Instead of rushing, they took the next week to make sure everything worked properly. Then on Friday they reversed the workload from Juneau to the new Anchorage facility. Having learned from the previous migration, this time it only took the team three hours to fail over the services that had lived in Juneau for a week, Kos said.
“When the team called and said they were finished, I didn’t believe them,” Kos said. “I thought they were jerking my chain.”
However, the UCS system has high availability and fail over capabilities in the same environment, Kos said. ETS used two UCS systems; one in Anchorage and the other in Juneau. “We used a model where we could cookie-cut these systems,” he said. They also developed methodologies to account for the geographical differences between the two locations.
“We tried to keep the two environments identical,” he said. “That way I can bring up capacity.” If they need to add computing capacity to the Anchorage location, they can add capacity preemptively to Juneau as well. The team can do capacity analysis and determine if there is enough computing capacity to bring up critical services in the other city.
“If you really want to do data center failover, you need two identical systems,” Kos said.