Businesses will be spending as much on powering
andcooling datacentresas they do on
servers unless they improve monitoring of equipment in datacentres,
Gartner has warned.
"There are no current metrics to compare how much more energy
efficient one datacentre is with another, but managers need to get
a stake in the ground about how much energy consumption is rising
in line with the processing power they are adding," said analyst
Rakesh Kumar.
Kumar said that datacentres should monitor the number of
blade servers they are adding to datacentres to make sure gains
in processing power are not being offset by increases in energy
costs.
He advised companies to create energy maps to understand the
where power was travelling to, how much each piece of equipment was
using, and its energy costs.
Reducing costs through virtualisation and consolidating the
number of datacentres would only work if they had these
measurements first.
Even if a company decides to outsource the problem of power and
cooling, they will be able to benchmark if the supplier is doing a
better job.
On the cooling front, Kumar said that firms should look at
thermal modelling software for data centres, so that cooling
systems could target specific hot areas at a specified temperatures
and times, rather than provide more expensive blanket coverage.
The average cost of running a UK enterprise data centre is
estimated at £5 million and powering and cooling costs currently
account for 4%-7%. This percentage will double by 2012 as
businesses will require more processing power per square foot from
their datacentre to cope with growth.
Intellect top tips for cutting energy bills
1. Consolidate equipment where possible to accept an higher
workload on like systems. Look at selective replacement with modern
multi-core processors which allow greater throughput at lower power
consumption and heat generation.
2. Adopt Virtualisation to increase the optimum loading of
equipment for like and unlike systems. Typical middleware allows
for dynamic relocation of processes allowing systems "nesting" for
out of hours power down.
3. Improve system management capabilities to enable remote power
down and wake up on Lan-type services to reduce power consumption
outside peak hours.
4. Consider alternative cooling methods. What about using
outside air most of the year? Can you run the datacentre at higher
temperatures? Consider smart cooling strategies to allow workloads
to be matched to optimal cooling loads.