IBM and Some College Students Aim to Simplify Data Center Disaster Planning
The picture at right is of the flooded ground floor of Verizon headquarters at 140 West St. in lower Manhattan after Hurricane Sandy blew through the Northeast last year.
For a few weeks last year, communicating was a real challenge in the storm zone. Power was out. Networks were stretched to the limit, in part because of a surge in demand, but also because a lot of the underlying equipment required to run them was offline or damaged.
When you’re in charge of managing digital infrastructure, planning for natural disasters is kind of a big deal. Businesses that rely on key systems in order to continue operating have to think long and hard about all the variables. Usually it involves building up a separate site to fall back on when the primary one can’t operate normally.
Then there’s that moment when the switchover, technically known as a reprovisioning, actually happens. Ideally it’s the sort of thing that requires days to carry out. The problem is the people doing the work usually get a few hours.
So IBM said today that it had come up with a way to make that switchover happen a lot faster. Working with some students at Marist College in Poughkeepsie, N.Y., Big Blue said, it demonstrated a way to use software-defined networking to accelerate the process of reprovisioning so that it takes only minutes.
When networks supporting voice and data communications go down, people lose the ability to communicate when they need it most. During the aftermath of Sandy in New York and New Jersey, a fair portion of the region’s Internet capacity failed, along with the power in all of lower Manhattan. A lot of techs lost a lot of sleep keeping certain systems running, and mitigating the impact from those that did fail.
For those people, the best part of this invention is that it allows everything to be done remotely.
With software-defined networking, adding to or changing the composition of networks inside the data center is handled by tweaks to software rules and not by adding or subtracting more hardware. Companies like Big Switch Networks and Nicira, a unit of VMWare, have sought to build a business around SDN that might challenge the established networking players like Cisco Systems and Juniper Networks. Hewlett-Packard and Dell are also seeking to go big on SDN.
Marist College is the site of a research lab devoted to developing SDN technology. Big Switch is also a partner.
A blog post by Zachary Meath, one of the Marist students who worked on the project, goes into detail. They knew, he writes, that they could create what he called “network agility” using Openflow, the open source software-defined networking protocol.
They then wrote an application that gives the network the ability to make its own changes on the fly based on the ebb and flow of demands on its bandwidth. “Sometimes, for example, an administrator may not be able to predict how much bandwidth is needed, but the network will know when to precisely add or remove bandwidth,” Meath writes.
They also created an app called Avior that gives a network manager the ability to carry out a reprovisioning from any Web browser, anywhere.
There are other uses for this in addition to disaster planning. If you’re expecting a surge in demand on a particular resource, you can use it to quickly shift some of the capacity out to the public cloud. This can be used to make sure there’s enough bandwidth to handle it.