Back from another Indian trip, and interesting discussions around how organisations can measure their datacentre’s “effectiveness” when it comes to carbon footprint. One measure out there is power usage effectiveness (PUE), where the total amount of energy being used by a datacentre is divided by the total amount of power that is being used in producing useful IT function. For example, if there is 300KW of energy being used in running the datacentre in total, and the servers, storage, network kit and so on use only 100KW, then the PUE will be 3. This shows that for each unit of useful power, two units are being used in cooling, in UPS overheads and so on. For many datacentres, this figure can be as bad as five – the general aim is to get to somewhere around 2.3.

The problem with PUE is that the higher number you get, the worse things are – which is counter-intuitive. Therefore, datacentre infrastructure efficiency (DCiE – the lower case “i” is nominally to differentiate it from DCE, but really is beyond me) divides the useful power by the total power – and a higher number means that you are doing well. In the above example, the DCiE score would be 0.33.

However, both scoring systems have their built-in problems. For example, look at the impact that rationalisation, consolidation and virtualisation would have on the calculation. You are still left with the same datacentre (unless you have done some fairly serious retro-fits to downsize it), so your non-useful power usage will change little. However, as you are now making better use of your IT kit, the amount of useful power will actually drop – and so your PUE and DCiE figures will actually get worse, even though you have done things for the good.

Conversely, adding a load of computer and network kit that you don’t mind losing and running it at high temperature until it fails will increase the nominal amount of “useful” power being used – according to the calculations being used. So, by collapsing your overall datacentre compute, network and storage utilisation rates to single figures, your PUE/DCiE figures could improve considerably.

So – surely some mistake? A different measure was promoted by Shaheen Meeran of Schnabel India. This one is based on work done by CS Technology out of the US, and looks to combine the PUE of a datacentre with the efficiency of the power generation mode used to create the electricity used. Within a large geographic area such as the US, India or China, this does provide a means of seeing if moving from one area to another (in the way that Google built a new datacentre in Oregon right next to a hydro-electric dam) will improve carbon footprint.

But, this still does not solve the problems with PUE/DCiE itself. For this, we are working with some of those responsible for datacentres in Asia-Pacific to look at how the overall efficiency of the IT components of a datacentre can be taken in to effect. In this way, those with high utilisation rates can guarantee that their modified PUE/DCiE figures will reflect the steps that they have been taking – rather than suddenly finding that they have been saddled with a comparative figure that makes their efforts look as if the datacentre is worse after their work than it was before. Another, but more complex way is to bring in Quocirca’s Total Value Proposition (TVP) methodology into the mix, and look at how much value the applications running in the datacentre add to the business itself – we’ll be looking at this too, but feel that it is still some way down the line.

By Clive Longbottom