Enough Data Center Management Ideas to Make Your Head Spin

Posted on April 18, 2013 by Gento

I’ve been back from the GreenGrid Forum 2013 for over a week now, but the insights I formed there are still making my head spin.

Early on, the GreenGrid was focused on keeping data centers cool. We all remember when walking into a data center was like stepping into a refrigerator. Now that is changing. Most IT equipment can now operate reliably at much higher temperatures. So this improved the PUE of data centers without having excessive cooling.

Now, in the age of Petabyte of data being processed, a key theme at the conference was resource efficiency among huge array of compute equipment. To achieve this, we are all now focusing on more granular management. No more refrigerator approach. Now data center managers are looking for efficiencies down at the rack and device level. Take servers for instance.

Consumers of power in a server:

1. Server processors
2. Memory
3. Discs or PCI slots
4. The motherboard
5. The fan
6. Networking interconnects

A GreenGrid study that was of great interest to me focused on the power required for different CPU utilization in data center servers from idle state to active state. When this state changes, the jump in energy draw is dramatic.

The main difference in power consumption between an idle server and an active one is based on the spinning of the disc when the server must move the read-write head. The differential in power utilization from idle to active is about 30%. You can see that in the chart above.

CPU utilization studies have shown that many servers sit idle for large periods of time. But even in that idle state, they spin. And the chart above shows that they are still consuming a large amount of power. The GreenGrid paper suggested segregating servers so that some development and build servers should be isolated and shut down completely when not in use. Backup servers are usually only needed at nighttime when the devices they’re backing up are idle. During the day, they could be shut down.

Another great example in the GreenGrid presentation was a server farm running 10 servers that are utilized only 50% each on average (we suspect your utilization will be even less). Five servers utilized at 100% would be exponentially more energy efficient. The question is, how to understand what each of your servers is doing on a granular level.

How to gather granular data and manage your data center power

Power IQ is our energy management component of Raritan’s DCIM solution, and it helps your data center become dramatically more energy efficient. Here is a partial list of what this pro-active software solution can do:

• User configurable dashboard
• Agentless graceful shutdown of Windows®, Linux® and Unix® device groups
• Psychrometric chart to calculate possible energy savings from simply increasing temperature
• Energy cost bill-back reports to drive behavior
• Decommission flags to forecast energy savings when eliminating servers
• Baseline consumption and changes of various energy saving initiatives
• Capture peak loads under compute stress to identify excess power capacity in the rack
• Power capacity forecast charts
• Power capacity utilization gauges
• Smart Rack View with Quick Charts

Looking through that list, you can see why Raritan is excited to be in the right place at the right time. We are leaders in intelligent power management for the data center thanks to our intelligent rack power distribution units and our powerful energy management software.

Click here to read the Green Grid White Paper

Click here to learn more about Raritan data center power management solutions