When I visited the data center for 1&1 Internet, a huge hosting and cloud services provider located in Lenexa, Kan., I was struck by the size of the emergency generating system. The Caterpillar diesels that power the generators are the size of locomotive engines.
The four generator sets power the 40,000 servers in an N+1 configuration, and they provide the power to operate the cooling and other building services necessary to keep the data center alive when the power goes out. N+1 refers to a widely used practice in IT to allow for at least one extra system or a set of systems that will serve as a backup in case of failure.
It's all on a grand scale. The rows of server racks seem to go on forever, networking equipment rooms line each side of the building. The main hallway of the building seems to stretch into infinity. In short, 1&1's data center is one big place.
But not every company has or needs a data center the size of the Lenexa, Kan., facility owned by 1&1. In fact, even some fairly large companies have data centers with less than 50,000 square feet or so. They don't need, and probably can't afford, a massive emergency power system like the one that backs up 1&1. The problem for these companies with smaller needs is that diesel generators frequently don't scale very well, especially when they need to be run in parallel so that they can use an N+1 configuration.
"In the past, paralleling was done with switchgear," said Curt Gibson, Western Power Solutions manager for Generac, a manufacturer of emergency power systems for a wide range of uses, including data center power. Gibson explained that the switchgear for running big generators in parallel was custom-built for each application. The problem has always been that as the size of the required power unit decreased, the cost did not decrease proportionally.
Gibson said that as power solutions come down in size, companies start feeling the pain of those custom parallel solutions around the time they get to a 1,500-kilowatt power backup system. Michael Kirchner, technical support manager for Generac, said this is the point where customers usually start looking at single-engine designs for their emergency generators, simply because a parallel solution is too expensive.
Unfortunately, a single emergency generator is also a single point of failure for your data center. If that one engine quits during an emergency, you're out of business. But now there's a new option that lets smaller businesses have parallel generating capacity, while keeping the cost competitive with single-engine solutions.
The new technology is built-in parallel generating support, in which the electronics necessary to sync the generators and provide N+1 or even N+2 failover is built into the device at the factory. Generac is now incorporating this technology into modular generator designs that allow what Gibson calls "pods" of generators that can consist of several modular units grouped together to power all or part of a data center.
According to Gibson, the latest practice is to power a data center with several pods, each connected to a part of the data center but isolated from each other to provide additional redundancy.
This article was originally published on 11-22-2011