Photo - Eddie Desouza, Head of Business Operations, APAC, Enlogic.
Intelligent PDU firm Enlogic said the commonly-used Power Usage Effectiveness (PUE) metric by itself does not present a true picture of data centre energy usage and that Malaysian DCMs need to review how they measure energy consumption.
Enlogic head of business operations, APAC, Eddie Desouza said data centre managers (DCMs) need to reconsider how the measure energy consumption especially as the Malaysian government will phase out fuel subsidies by 2016.
"PUE has become a commonly used metric for data centre managers to monitor energy efficiency, but when quoted in isolation it can often portray a misleading picture of a data centre's overall efficiency," said Desouza.
Companies should balance PUE with more up to date systems such as intelligent PDUs [power distribution units] and performance-related metrics to more accurately measure energy consumption, which would allow energy cost savings initiatives, he said.
"Achieving success in a competitive market place is about finding efficiencies and optimising operational costs," said Desouza. "For years data centre managers have been convinced that PUE is the magical metric that will give them this competitive edge - but it's not."
Invented in 2007 by the Green Grid, PUE is calculated against the ratio of the total amount of power used by a data centre facility (including lighting, heating, cooling) divided by the power delivered to computing equipment.
While PUE does provide an effectiveness metric of the facility services to the IT equipment; it does not provide any meaning in relation to IT equipment efficiency, and in fact can become quite misleading, Desouza said.
"Take an extreme example; imagine a data centre full of thousands of computer servers sitting idle, but consuming energy. Suppose such a data centre has a well-managed facility with a PUE of 1.4; would one believe this is an efficient date centre despite no computing work being done?" he said.
"In the more realistic situation, PUE can mislead users into making poor decisions," said Desouza. "For example, an existing data centre consolidates their equipment and shuts down idle servers in an effort to save energy. In this situation, PUE will often increase after the change since the reduced IT energy use does not typically reduce the total facility energy by the same percentage. In effect, the energy use is lower, but the PUE can be higher."
He added that while reducing total energy consumption is ultimately more important than reducing the PUE, many managers are setting their individual and facility performance goals based on improved PUE, not necessarily reduced energy which is harder to quantify. "If key decision makers base their decisions on PUE alone, it could mean paying considerably more than necessary for energy and carbon taxes, despite best intentions to improve carbon efficiency."
Desouza cited an example: Consider an existing data centre below with existing IT power load averaging 100kW. After consolidating equipment and turning off un-needed extra servers, the remaining equipment performs the same work with lower total energy usage. But, the PUE has increased and can improperly imply that the new condition is less efficient.
"The potential energy and financial savings that data centres could achieve are entirely overlooked when PUE is used as a single efficiency metric," he said. "The most energy efficient data centres are actually reducing overall energy consumption and costs by looking holistically at PUE in conjunction with IT performance per watt type metrics."
Sign up for CIO Asia eNewsletters.