With many data centers consuming as much electricity for cooling and other non-IT needs as for the IT servers and storage units themselves, power and cooling know-how is a hot item indeed (figuratively speaking.)
Add to this the size of many modern data centers, each with power requirements equivalent to that of a small town, and you can see that data center providers need to be on the ball for:
- Correctly handling and distributing large amounts of power
- Providing solid power backup solutions
- Correctly handling cooling requirements throughout the data center
- Making the above available at prices that won’t blow enterprise or customer budgets out of the water.
A PUE of One is Power Nirvana
It’s also technically impossible, but data center providers continue to get closer and closer.
Those that need as much power for non-IT needs as for IT needs to have a PUE of 2. The lower the power consumption for non-IT needs (cooling, lighting, coffee machines, and so on), the lower the PUE gets.
If IT needs were all that had to be satisfied, the PUE would become 1. PUE figures of as low as 1.12 have already been logged. While paying due attention to reliability and performance, providers with lower PUEs can then lower the bill to their users.
What Should You Be Looking For?
The only realistic way to assess the power and cooling know-how of a data center is to find out directly how the provider addresses issues like the following:
- Cooling capacities adapted for the equipment concerned. In colocation, for example, high-end server racks may need more cooling, whereas legacy machines require less. Running the whole data center at the maximum rate would be wasteful, whereas applying a minimum rate would soon cause problems. Separate racks and/or aisles need properly adapted cooling.
- Cabinets with suitable power ratings for different densities of servers. Too low a power rating may force the distribution of high-end server clusters over numerous cabinets, defeating the objective of high-density computing.
- Redundant power feeds for servers that are correctly configured to allow enough power in the event of circuit failure but without excessive charges for the provision and availability of the redundant circuit.
- Responsiveness to changes in user or customer power needs for sufficiently rapid increases in provisioned power with the option of a remote installation and configuration service to save users as journey out to the data center to perform their part of the upgrade.
- Backup power that fully supports the total cooling requirements. Not every backup system is dimensioned for this. Getting the hard facts is the only way to find out what will happen in the event of a mains power outage.
- Correct thermal isolation between racks, including those that are part empty. To prevent hot and cold air from mixing where they should not, the solution may be as simple as ensuring the use of blanking panels where rack space is empty.
The best data center providers are also savvy enough to help customers to help themselves. Two examples are advice about putting power-dense equipment lower in racks and using proper cable management to leave ventilators unobstructed.
Which technologies have you found useful for reducing power consumption in servers and storage units? Share a tip that has worked for you by leaving a note in the Comments section below.
If you're in the data center, mission-critical, or cloud services industries, or you sell to the data center industry, don't miss our weekly update newsletter -- Data Center Sales & Marketing Institute (DCSMI) Update Newsletter. Get notified about new reports, events, podcasts, and blog posts.