For the last 20 years, power and space have been treated as endless commodities. We procure, rack and stack, run, and decom, growing the data centre infrastructure until the facilities department pulls the ‘out of power’ alarm, or the data centre manager forecasts space at a premium, and reactive measures take hold.
Enterprise data centres (say of 500 assets or more) usually don’t maintain an accurate inventory of what’s running inside and how much energy/power is being consumed. Scattered inventory of applications to host relationships, inaccurate inventory of equipment, and no earthly idea of power consumption is more the rule than the exception.
According to Connection Research, only 3.5% of CIO’s have responsibility and budget for the data centre power bill, 66% have NO visibility or responsibility, and the remainder have some visibility and some accountability (which isn’t saying much since this is a self-assessment scoring, usually biasing results to the favor of the assessed!).
Many ITIL advocates envisage cloud public/private architecture as the 3rd incarnation of ITIL, but seriously if basic inventory and asset management didn’t manifest in the last 10 years of ITIL adoption, I’m not sure it’s going to happen now when the pace of data centre virtualisation/consolidation is accelerating. If people power couldn’t keep pace with physical inventory there’s not a hope in the hell of asset management it will keep up with virtualized infrastructure….
If you look at a non-innovative data centre, you might be looking at over 1000 watts per server, while the present day opportunity to virtualise and consolidate could collapse an entire data centre into a score of high-density racks of x86 infrastructure and an astounding 45-60 kW/rack power draw nearly demanding water cooling, or high-velocity air-cooling techniques only found in containers or psueudo-containers.
All of a sudden, space (and it’s cooling design) and power are crucial considerations as data centre footprints have an opportunity to shrink in the face of steady growth. As this happens, application to virtual asset inventory will become more crucial, and the physical inventory itself may become less relevant as non-virtualised infrastructures diminish. Logical management, metering, and measurement of the virtual infrastructure (still in it’s infancy) articulating power draw per application, and virtual compute/storage run-time consumption will then be the true measure between private and public infrastructure. But then, the graph above underscores the primitive nature of our ways in the midst of accelerating changes to the distributed computing landscape.
