Comments on: Data Center Efficiency – is it time to measure it? http://www.bladewatch.com/2008/12/04/data-center-efficiency-is-it-time-to-measure-it/ Tracking servers and enterprise technology. Fri, 03 Feb 2012 11:11:22 +0000 hourly 1 http://wordpress.org/?v=3.3.1 By: Jacque Swartz http://www.bladewatch.com/2008/12/04/data-center-efficiency-is-it-time-to-measure-it/#comment-44314 Jacque Swartz Thu, 04 Dec 2008 16:59:40 +0000 http://www.bladewatch.com/?p=4375#comment-44314 This is certainly a laudable goal. Calculating MPG is very straightforward, dividing two easily quantifiable values together. Defining the analogous variables for the data center is difficult. You would have to agree on the level of the measure; the entire IT environment at a company, a single data center, a single server, a set of servers supporting a specific application? <br><br>One variable would be kWH, replacing Gallons in the MPG calculation. Defining the equivalent for miles is vastly more complex. The challenge lies in the definition of a unit of work: What do you measure per kW? Is it a transaction, compute cycle, or availability? How do you compare a virtualized environment to one that is not virtualized? The configuration permutations are infinite. And this is only about the servers. Add in the complexity of the infrastructure systems, all of them, and I struggle to imagine how two data centers can be effectively compared.<br><br>Using the PUE (Power Usage Effectiveness) ratio from The Green Grid does establish one benchmark, but it does not evaluate the efficiency of the IT equipment load. Improving the efficiency of the IT equipment can actually increase PUE.<br><br>There is the potential for a relatively simple measure though; Revenue/IT spend. This would require establishing standardized reporting to ensure all companies included the same expenses in the IT spend and businesses would have to be willing to publish this data so that comparisons could be made. Any CFOs out there care to weigh in? This is certainly a laudable goal. Calculating MPG is very straightforward, dividing two easily quantifiable values together. Defining the analogous variables for the data center is difficult. You would have to agree on the level of the measure; the entire IT environment at a company, a single data center, a single server, a set of servers supporting a specific application?

One variable would be kWH, replacing Gallons in the MPG calculation. Defining the equivalent for miles is vastly more complex. The challenge lies in the definition of a unit of work: What do you measure per kW? Is it a transaction, compute cycle, or availability? How do you compare a virtualized environment to one that is not virtualized? The configuration permutations are infinite. And this is only about the servers. Add in the complexity of the infrastructure systems, all of them, and I struggle to imagine how two data centers can be effectively compared.

Using the PUE (Power Usage Effectiveness) ratio from The Green Grid does establish one benchmark, but it does not evaluate the efficiency of the IT equipment load. Improving the efficiency of the IT equipment can actually increase PUE.

There is the potential for a relatively simple measure though; Revenue/IT spend. This would require establishing standardized reporting to ensure all companies included the same expenses in the IT spend and businesses would have to be willing to publish this data so that comparisons could be made. Any CFOs out there care to weigh in?

]]>