All datacentre products claiming to be green are not created equal. That's because the metrics to determine the 'greenness' of equipment is usually vendor-driven, and measuring energy efficiency can be a chore for network architects.
"There is currently no widely agreed upon definition in the industry for green," says Scott Scherer, an analyst with In-Stat and co-author of a report, "Green networking equipment: Who leads and who lags."
Individual vendors create their own meaning of greenness in measuring how well their equipment stacks up, Scherer says. "Some vendors use 'green' interchangeably with the energy efficiency of their products, while others use more holistic definitions that include corporate social responsibility, manufacturing processes, materials and recycling programmes," he says.
Such myriad options leads to a range of units of measurement that network executives can sift through when assessing the environmental friendliness of their network components, as well as individual products they consider as they refresh and expand their infrastructure.
One simple measure is how much power a device uses vs. the number of ports it has, so the unit is power per port. This harkens back to the popular measure of cost-effectiveness, price per port. Because port speed and type (copper or fibre) vary, this measure has its shortcomings.
An alternative is the throughput of a device compared with how much power it uses, so the metric would be bits per second/Watt. In practice this would likely be measured in Mbps or Gbps per MWatt.
But device-by-device analysis may not give businesses the big picture they seek, so metrics such as power per user come into play, Sherer says. This takes into account the range of devices that make up the network infrastructure and all the power they consume in order to serve a given worker population.
Because businesses measure productivity based on costs and expenses required to produce products, some look at greenness in the same way, Sherer says. This becomes power per unit of work, a complicated calculation that depends heavily on the type of business a company is in.
Coming up with an objective measure for the unit of work can be challenging, he says. "This is a metric that is user-specific and depends on the type of work that the network actually performs," he says. "These are often very complex and require the equipment for actual testing."
Despite the difficulty, setting the right measure of greenness is all-important, says The Green Grid, a consortium focused on datacentre energy efficiency. "Using the wrong metric in this process will lead to either erratic or invalid results," Green Grid says.
Green Grid splits measurements between power-performance and power-usage effectiveness and says that both are important, the first for determining what gear to buy and the second to measure how efficiently the gear operates in a real-world deployment.
For instance, route-control service provider Internap is building a collocation centre near Boston with greenness in mind, and measures it based on differences between the gear in its legacy collocation centres and the new one, says Mike Frank, the company's vice president of datacentres.
To decrease overall power consumption, Internap pumps in cool outside air to drop datacentre temperatures to a lower level so the air conditioning doesn't have to work as much, reducing the electricity use by the centre, Frank says.
It also uses ultrasonic humidifiers to make the air moist rather than heating water until it is steam, reducing both the heat and the electricity used. Each hot humidifier uses 6 to 7 kilowatts per hour, and each of more than 30 air conditioning units in the centre requires one. The ultrasonic humidifiers run at 55 degrees Fahrenheit and use just 4 kilowatt hours for the whole site, he says.
The company faces a three- to six-month payoff on the humidifiers in energy savings. "There was a 250Kw drop after we unplugged the old humidifiers. That's a quarter of a megawatt that just disappeared," Frank says.
All electrical circuits feeding the facility are monitored to determine how much energy is being spent on each device, which is measured in CPU cycles per watt per dollar. "You can begin to make some pretty interesting decisions about the equipment you use," Frank says.
Ideally, businesses should evaluate equipment based on metrics they can measure before they buy and follow up to see that it performs optimally once it is deployed, Sherer says.
"A customer can calculate most of these metrics and their variations quickly by hand for quick comparisons and get a pretty good picture of where different vendors' equipment falls," he says. "However, to truly discern which vendor has the most efficient equipment, a combination of both equipment-level and system-level metrics should be used, as well as real equipment under varying loads of network traffic."