March 25, 2013, 6:17 AM — There's a maxim in the data center business that holds that you can't manage what you can't measure, and eBay has come up with a new metric designed to make it easier to measure -- and therefore manage -- data center performance.
The online auction service looks at the cost of IT operations in dollars, kilowatt hours and carbon emissions, and then ties those costs back to a single performance metric -- the number of transactions its customers conduct on its website.
The result is a "miles per gallon" type of metric that eBay can strive to improve over time, said Dean Nelson, head of eBay's Global Foundation Services unit, which manages the company's data centers worldwide.
"EBay is a single system; it's the sum of a million parts, and we needed a way to measure and convey the efficiency of this system," he said at a recent conference hosted by the Green Grid Forum.
EBay published the methodology for calculating the metric, called Digital Service Efficiency, in hopes that other companies will adopt it -- much as the industry rallied around Power Usage Effectiveness, or PUE, as a measure of data center efficiency.
In the process of sharing its methodology, eBay took the unusual step of releasing a wealth of information about its data center operations. For example, the company revealed that it had 52,075 physical servers at the end of 2012 and generated 740 metric tons of carbon per million users, or 1.6 tons per server.
EBay set a target of reducing its cost per transaction and carbon emissions per transaction by 10% this year, and it wants to increase its transactions per kilowatt hour by the same amount, Nelson said.
It may not be easy for other companies to replicate eBay's methodology because other organizations' business models may be more complex than a straightforward count of the number of transactions that customers make.
This version of this story was originally published in Computerworld's print edition. It was adapted from an article that appeared earlier on Computerworld.com.
Read more about data center in Computerworld's Data Center Topic Center.