What would happen if your data center went dark? Most companies don't want to think about this possibility, let alone flip the switch and see what happens when the lights go out.
It might be worthwhile. According to Computer World, data centers in the United States used a collective 100 billion kilowatt hours (kWh) of power last year, which amounts to 2 percent of national use. The White House has developed a voluntary program to ramp up energy efficiency 20 percent by 2020, but so far only 19 data centers are on the list — 12 of them government run.Without any legislation to prescribe efficiency or mete out penalties, private industry has no compelling to lower energy use. But with increasing cabinet densities and a push for higher baseline operating temperatures, both power consumption and the chance of failure are on the rise. Going dark may not be an “option” for long.
According to Data Center Knowledge, social giant Facebook decided to shut down a data center and observe the results. The company's engineering team did some infrastructure prep work, selected a site and then turned off the power. Jay Parikh, global head of engineering for Facebook described it as “tens of megawatts of power that we basically turned off for an entire day to test how our systems were going to actually respond.” And the result? “It was actually pretty boring for us.”
There were issues, of course, but nothing was permanently damaged — and applications kept running. Not in theory, not on paper but in practice and for an entire day. Still, does this kind of risk really benefit the system at large, or is this just showmanship?
Managed data centers and colocation providers want your business. To get it, they'll provide high availability, low latency and — of course — uptime guarantees, all spelled out in a service level agreement (SLA). But how many actually put this to the test?
Some talk a big game about power outage plans, point to generators or other backups but balk at the mention of testing. They're like the Wizard of Oz: what's really behind the curtain? Others are willing to flip the switch and have developed reliable, repeatable procedures to move from grid supply to generator power. They'll be upfront about when these tests are run, for how long, and with what kind of results. It's a question of word versus deed: is the supposedly massive, never-seen backup a safer bet than observable, small-scale stability?
There are two kinds of power failures — those that happen by chance, and those that happen by choice. Every data center is subject to chance, but the best opt for failure by choice: ultimately, preparation trumps perception.