To investigate further, I opened the in-flight EasyJet magazine. A full page was also dedicated to telling the bargain-fare consumer what a great job they did saving the environment by choosing EasyJet. A scan of the page is to the right, click to enlarge. In fact, EasyJet claims they emit "22% less CO2 per passenger km than the traditional airline." But note the tiny unreadable text at the bottom of the page, a footnote to this claim: "Based on the same aircraft flying the same route." It turns out that EasyJet only has 1 class of service, and the seats are crammed together. By doing so, they can fit more people onto the same plane used by other airlines that actually give you enough room to sit down. So my flight from Luton to Paris emitted the same carbon emissions (147kg CO2) as every other airline, but since EasyJet puts more people on the plane, each person's carbon contribution is lower (that means 22% more seats crammed into the same tiny space on EasyJet). Lame, I know, because the flight would have departed regardless of the number of people on the plane. Don't worry, you're not the only one confused. EasyJet was reprimanded by the British government for deceptive advertising. Next time, I should take the train -- which actually offsets the carbon impact of my journey as part of my ticket price.

At this point, you may be wondering what this has to do with a energy management or a datacenter. It turns out that EasyJet isn't the only company using statistics to their advantage -- the datacenter world is doing it too, and the biggest culprit is an often thrown around ratio called PUE (note that I call it a ratio and not an efficiency metric). As I'm sure you're aware, PUE is the ratio of the total energy used by a facility to the energy used by the IT equipment. The goal of PUE is to benchmark yourself -- to understand if you're making your facility's operations more efficient relative to it's past performance. But recently marketing departments have taken hold of PUE and are using it to promote a game of "my datacenter is better than yours", like an elementary school kid.

Why is it a problem? The first reason should be obvious: It was a ratio that was never intended for comparison among datacenters! There's no procedural requirements to PUE, so one company may promote their PUE taken during the coldest day of the year for example (not to mention that weather can cause +/- 10% variations in PUE measurements). Bill Kosik at HP writes about how we can improve PUE measurements, or at least standardize how those measurements are taken within your facility. But keep in mind that PUE isn't bad! In fact, it is a great way of getting a baseline and tracking the progress of efficiency projects, as long as you take the readings the same way, every time. And probably the most important thing about PUE is that it is raising awareness about the lack of metering today and what an impact it has on your business.

This isn't just a rant about PUE though. The carbon impact of our operations is significant, and if you're not thinking about CO2 and how to track it, you should. I'm not the only one that thinks so, Mike Manos from DRT recently wrote about the CRC in the UK and how it will impact a data center near you. I also spent time in London at a BCS event with Jim Smith, CTO at DRT, who told me that DRT installs meters on everything within their datacenters and customer datacenters. When asked "why?", Jim said, "When it comes to metering, I give my guys an open budget. You have to be prepared, know where you stand, be able to accurately bill your customers and track expenses, and be armed with the data when legislation and regulation comes. And it will come, you can be sure of that." While CRC doesn't directly target datacenters, it impacts companies that run datacenters -- which is just about every company in the world, from a small server closet to a federation of datacenters around the globe. Keep in mind, that 147kg of CO2 from the 1 hour EasyJet flight is equivalent to the same emissions from 1 hour of running a 150kW datacenter (about a 200 server datacenter)!