In the movies, zombies walk around with blank stares looking for brains to eat. In the real world, Zombie servers have that same “blank stare”, but they eat energy. Of course, as Zondits has noted in previous posts about the flaws of PUE, the energy to power these zombie servers is in the denominator of PUE with all the other IT load so it does not hinder achievement of a low PUE. In fact, unplugging them would make the denominator smaller so in theory could even increase PUE. A facility can have 100’s of zombie servers and still have a great PUE. But I digress, the point of the post below is that PUE seems to have hit some sort of plateau, with the best of the best facilities changing little for several years now. It can’t be that there is nothing left to do – even beyond the zombies we all know there are always improvements to be made. In this case the measure is flawed and provides false comfort! Might this plateau drive the industry to consider a better metric for performance – or worse, will they pat themselves on the back and think their job is finished?!
Survey: Industry Average Data Center PUE Stays Nearly Flat Over Four Years
Data Knowledge Center, June 2, 2014
While the majority of data center operators who participated in a recent industry survey measure Power Usage Effectiveness, their average efficiency ratio has not changed by much over the past four years.
This was one of the conclusions of this year’s survey by the Uptime Institute, which released the results at its annual Symposium in Santa Clara, California, in May. Starting with 1.89 in 2011, average PUE among the companies surveyed went down to 1.8 in 2012 and further down to 1.67 the following year. This year, however, it was up to 1.7.