A Simple Three-Step Plan To Cut Power Usage

Armed with some common sense and a few simple skills, resellers can not only help customers reduce their monthly utility costs, but earn themselves extra revenue in the process. Here's a simple plan.

Step 1 - The Walk Through

The best plans are often the easiest, and the principles for inspecting a data center's efficiency (and pocketing some easy cash) will be within your grasp by the time you finish reading this section.

Regardless of the size of a company's data center (or server closet), they all have one thing in common: all contain equipment that generates heat that must be removed. And because cool components last longer, run faster and operate more efficiently, it's easy to make people understand the benefits of an efficient data center.

Aside from the computer room air conditioners (CRAC), the primary factors effecting computer room cooling efficiency are the orientation of the computing equipment relative to their intake and exhaust.

So the first thing to determine is whether the servers, switches and other heat generators are all pushing their hot air in the same general direction. This hot-aisle-cold-aisle technique helps to isolate air of like temperatures, and ensures that air going into machines is as cool as possible and that chilled air is not being sucked out of the room before being drawn into any equipment.

Once the room is arranged into hot and cold aisles, efficiency is further improved by adding partitions between those aisles, just as it's best to raise the windows when a car's air conditioning is on. A variation of this technique is to isolate just the hot aisle, directing computer exhaust toward the CRAC unit's intake or an opening in a plenum ceiling.

Next: Steps Two And Three Step 2 - Get The PUE

Now that they've been adequately partitioned, customer data centers are saving as much as 15 percent on air conditioning costs and 67 percent on the cost of fan systems, according to a Lawrence Berkeley National Laboratory study.

Next is to calculate the data center's PUE or power usage efficiency. This is a necessary precursor to a full power audit, because it will provide a baseline for tracking power efficiency improvements. Simply take the total power usage (in watts) of the data center and its infrastructure (including lights, cooling and all IT equipment) and divide it by the IT load itself. IT load includes only components directly related to computing, such as only servers, storage, switches, routers and so on. Don't include UPSes, cooling units or anything that's not plugged into a computer.

Compare the organization's PUE with those of the Uptime Institute's national averages. The number should be somewhere between 1.8 and 3.0.

The perfect PUE would be 1.0, indicating that compute resources constitute the entire infrastructure. This is impossible in most cases, of course; a PUE of 2.0 or 2.5 is considered average for small and medium-sized data centers. With a PUE of 1.2 is Google, which runs the most efficient data centers in the world. The company publishes its power saving methodology. Google is helped by the ability to design and custom-manufacture all of its servers.

Step 3 - Full Power Audit

For the rest of us, measuring the power consumption of components on data center racks and shelves is the only option available. One quick route is to examine the age of the customer's uninterruptible power supplies. Today's UPSes are far more efficient than those manufactured even three years ago. And if they're more than five years old, they're probably ready for the scrap heap anyway.

The next step is to measure and record the power consumption of the remaining IT components. To simplify the job, we've obtained this Power Audit worksheet from Appropedia.com, which includes formulas to calculate usage. Once completed, the most inefficient components can be identified and targeted for power reduction based on their individual settings or replaced or upgraded to more energy efficient models.