Considering the fact, that only 30-40% of the energy consumed by a data center is used by the actual computational equipment and considering that another 30%-40% of the energy consumed by the IT equipment is converted to heat, only 21%-24% of the energy eaten up in data center is actually converted into computing power. Looking at these numbers from the other side means for each Euro spent on “computational energy” 4,76 Euro are spent on “overhead”. This ration can be improved by optimizing air conditioning, getting rid of heat hot spots or generally using energy efficient and modern equipment. Still it seems unlikely that this will help to get anywhere below 3 Euros of “overhead” for each Euro spent on “computational energy”. These numbers are taken out of the keynote presentation given by Steven Sams at IBM PULSE 2008 (Also check out the “Raised Floor Blog”, where Steven Sams is one of the authors)

On the other hand this means that reducing the energy needed for computational equipment will in absolute numbers decrease the excess energy consumption by a factor of more than 4. So improving the facility is a good start, but reducing the energy actually needed by computational equipment is the real price. The way to reduce energy needed is a direct result from capacity management. Generally speaking this means – in the best case – turning off as many components as possible – or if that is not possible, at least cutting their energy usage by slowing the CPUs or putting virtual instances into suspend mode until their service is really required. Does this sound easy? Well it does, but how does it work? Virtualization certainly is the key technology, but what good would virtual machines be, if their resources could not be allocated automatically depending o their actual use or – if you want to be cautious – by their predicted load and therefore by their predicted usage. A specialized set of rules is put behind process and operational automation, to perform the scale-down and scale-up of the virtual machine resources. This automation can even decide to turn off hardware, that is currently not needed or at least to slow the CPUs of hardware that cannot be turned off, but is in little use.

Modern “or very green systems” come along with special agent to detect energy consumption and usage deriving possible executions. But how about all the legacy applications – the applications that are running on more than 95% of all the components, using up energy in our data centers today? An automation engine that actually acts like an operator (someone who could manually cut down on power use) could examine the equipment in the IT landscape it is acting upon and execute general rules to reduce energy usage. By combining both technologies – the more effective combination of modern hardware and specialized software for new applications and a general automation engine for all the legacy applications – the power of virtualized components can actually be converted to green power. This is not just a fabulous business case, but it also is a good thing for the environment and hence for all of us.