As the functioning of the data center becomes increasing tied to building infrastructure, it has become necessary that you learn enough about basic power and air conditioning to manage those critical parts of your data center. You don't need to become an engineer, but the principles are the same as those you use to manage your network design and traffic; a poor configuration or an overload can bring down your power network, or your data network.
Of critical concern to many data center managers is an unexpected power outage. Here are three most common reasons for power or grid failures and how to lower the risk of them happening to your organization.
A real power grid failure
The only answer to this one is a good generator plant, regularly tested, with a sure way of getting fuel delivery, and a well-designed UPS to bridge the time until the generator starts. Even if you have two power feeds to your building, and even if they're from two different power sub-stations, and run through two different street routes, those sub-stations are still part of the same regional and national power grid. If the grid fails, so does your power. No way around it, except to make your own.
Poorly or under-designed building transformers and/or primary switchgear
This is definitely a facilities problem. A bad design is very difficult to change without shutting down the whole building, and it takes a very competent electrical engineer to analyze. What can be fixed, however, is having too many single points of failure between the incoming power and your data center. Your feeds should go through as few switches, fuses and panels as possible on its way to your UPS. Your generator should also bypass as much of the building electrical gear as possible. If it doesn't, that can probably be changed as well, but you'll need to pressure facilities to get these things done. It's expensive work. If it wasn't, it probably would have been done better to begin with. Good maintenance is another important item that's facilities' responsibility. That's what those yearly power shutdowns are for. Dirt gets cleaned out of big switchgear to avoid arc-overs, and contacts get re-tightened. Don't complain when this needs to be done. It can save your neck. Hopefully, you'll have that generator available.
Overload
We figure circuit breakers and fuses will avoid this problem, but it's not always so. In a data center we have lots of circuits to our cabinets, but we don't come anywhere near loading most of them. If your UPS is fully redundant, you're supposed to be limiting the load on each one to less than 50% of capacity. Unless you're managing your power, you can build up load to where one of two things can happen:
1. You have a UPS shutdown and the load on the redundant UPS jumps to over 100%. Bingo – you're off the air. Facilities can help you check this out, but it's your responsibility to manage what you plug into where.
2. All your new hardware, plus everyone else's in the building, overloads main building fuses or circuit breakers. Or the load heats up main transformers that are getting a little old and haven't been checked in a number of years. Fuses can be replaced fairly quickly, and the source of the overload traced. Blown-up transformers are another matter.
It pays to discuss the entire building power status with Facilities. A good time to do it is when before you add more load. Get them on your side. Show that you care. But understand the basics as well.
It's not really that hard. You can read, attend seminars, and ask questions of good consultants. We'll be talking about this very topic at the TechTarget conference
in June.
McFarlane is the president of the Interport Division of Shen Milsom Wilke, Inc. He has spent more than 30 years in communications consulting, with experience in every segment of the industry including, cable design, fiber optics, data communications, telephony, financial trading systems, data centers, satellite earth stations, video and multi-media, and acoustics.
