IBM and Marist College in Poughkeepsie, NY, are testing a system they codeveloped to support near-instant network reconfiguration to allow dynamic movement of VMs over long distances between data centers in emergencies. The methodology, which requires a software-defined network, allows network administrators to pre-program network reconfiguration and move VMs and associated data out of harms way in an emergency and activate the plan with the touch of a few software buttons on an Android or iOS smartphone or tablet. The system can also be used for non-emergency situations such as automated load balancing among multiple data centers, without interruption or even stutter in service to end-users.

A year ago, before Superstorm Sandy, “if a big bank had called their carrier and said they wanted to move core VMs 300 miles, out of the way of the storm, they would have been told that was impossible,” said IBM Distinguished Engineer Casimer DeCusatis. “I know of a client who was sandbagging around its New Jersey data center during the storm. If the water had been a couple of feet higher they would have had water on the floor of their data center.”

If that had happened, they would have lost their core operational and customer-facing transactional systems. Ironically some of the data centers that were nearly flooded in Sandy were the same ones moved from Manhattan after 9/11.

The system is being tested under real conditions over a 125 KM optical metropolitan network built with technology from partner Adva Optical Networking, connecting three working data centers running a variety of IBM servers and storage boxes ranging from PureSystems to a System z mainframe. This was built as part of a larger research program centered at Marist College and including the City University of New York (CUNY), State University of New York (SUNY) system, and Columbia University. The network and the Marist SDN lab are central parts of the New York State Center for Cloud Computing and Analytics.

Marist undergrads were part of the team that wrote much of the open source software on which the automated reprovisioning system runs, including the user interface, Avior, that lets a network administrator program and control the software-defined network. “To have them be able to get direction from IBM and other companies and turn out something that has practical applications in the real world is really significant,” DeCusatis said. “It’s not just something academic that they read about in a textbook or look at online. It becomes real to them, and they are going to be perfectly positioned for opportunities at IBM and other companies in the IT industry when they graduate.”

Why this is important

Today, without SDN, it can take weeks and the efforts of dozens of highly trained technicians to reconfigure a network to move large applications between data centers. In an emergency you do not have weeks — at most in a Superstorm scenario you may have two or three days. In an earthquake or major terrorist event you may only have minutes to react. And in an emergency the network admin and CIO are going to need to get their families out of danger, not spend hours reconfiguring network switches.

“We envision the network admin heading for his car, and as he walks through the parking lot he pulls out his smartphone, taps a few buttons, and the network automatically reconfigures and moves the VMs to another data center, out of danger,” DeCusatis said.

He emphasized that the entire, preprogrammed, transfer happens without any interruption in service. In the test, he said, they are moving a recorded video webcast around the metropolitan network while people are watching the video. “There is no delay, no pixelation. The viewers have no indication that anything is happening.”

Today they are doing the transfers across 125 KM, mainly because of the limitations of the metro optical networking equipment they are using. Next year they will begin testing across 200-300 KM distances. From there they plan to expand to links in the thousands of kilometer range crossing international boundaries. “I don’t believe anybody’s attempted that before, so we’re not completely sure what’s going to happen,” he said. “But we think we know what the key pain points are going to be and that we’ll be able to design around them.”

Photos copyright IBM Corp. 2013