Everyone is in favor of driving more efficiency in the data center, but what happens when that desire starts to interfere with availability?
And even more importantly, how will you know when that line has been crossed?
I had a chance to sit in on a few sessions at the Uptime Institute Symposium this week, and one of the more interesting dealt with the renewed focus on more efficient power and cooling. If the consensus of the group is any indication, it seems that the future is going to see IT scaled to extreme degrees as enterprises seek to cover their data needs in ways that require minimal power and cooling infrastructure.
“Scale matters,” said Antony Wanger, president of I/O Data Centers. “It drives efficiency by using large-scale systems.”
Organizations that are large enough to consolidate disparate resources into massive centers will do so, while smaller groups will migrate to the cloud. Either way, the trend toward massive, centralized IT facilities will be a key factor in reducing overall power requirements.
That’s the end-game anyway. But what about the transition process, which could last more than a decade? How will IT executives know they are gaining a more efficient operation, but not so efficient that it fails to meet service levels? In a word, measurement.
“If you don’t measure, you can’t manage,” said Dr. Bob Sullivan, senior consultant at the Uptime Institute. “If you can’t manage, you lose control, chaos rules and someone else makes your decisions for you.”
And one of the most effective measurement tools is the familiar power meter, said Joerg Desler, vice president of engineering and production at A/C systems company Stulz-ATS.
“I’ve seen many approaches over the years to measure data center power, and the simplest way is to look at the power meter,” he said.
But as we’ve seen with the debate over the Power Usage Effectiveness (PUE) model, simply calculating the best way to use available energy is a challenge in itself. If you cut your power draw in half, are you more efficient if productivity drops to a quarter of what it once was?
“This goes directly to how you use a system — how you load it,” Wanger said. “You can have a hybrid car running idle all night long, but that’s not as efficient as going 60 miles per hour on the highway.”
And that’s the conundrum when it comes to energy efficiency: We can lower power consumption all we want, but the work still needs to get done. In the end, it will find the resources it needs somewhere, but who’s to say it will be in the most efficient manner?