Page 212 -

10: Green IT Case Studies for Universities and a Large Company 177

As the IT organization should also set a good example for its own cus-
tomers by demonstrating implementation of its strategy within its own
resources, the IT organization should virtualize the systems management
servers in the subject data center. We created a project to virtualize those
servers and submitted the project in the Fall plan for 2008 implementation.
We received funding for this project, and have begun implementing the
project creating standard deliverables for how to build and maintain the
internal infrastructure virtual images.

Step 4. Manage and Measure
The next step was to implement power management software. Previously,
a data center space request tool was implemented that allowed the accounts
to indicate when they had new data center requirements. This tool was
enhanced to not only handle space requests but also to determine the power
and cooling requirements. The management system was enhanced to report
on power (utility, UPS, and generator), space, and cooling.
Prior to the physical resources being constrained, the name-plate power
for the equipment was used for planning. The name-plate power was a con- ptg
servative estimate of the amount of power needed to run a fully configured
system. Unfortunately, many of the servers we ordered were not fully config-
ured, causing us to oversize the power and cooling requirements. Equipment
manufacturers recognized this problem and have provided power calculators
that allowed us to more accurately size the power requirements because the
calculators enabled us to specify equipment that was not fully configured.
This improvement still represented an overstatement of our power usage,
as many servers do not use maximum power at all times. For example, power
supplies do not use all their power when the servers are not running at maxi-
mum utilization. To improve the accuracy of the power consumption, we
worked with our real estate organization to provide us with monthly meas-
urements of the actual amount of electricity consumed in the data center.
The IT team started basing demand on actual data versus estimated data.
There was still the concern on what percentage of name-plate power we
should use to size new requests. Analysis was performed on the power usage
of the existing equipment. The analysis team determined that using an over-
all 60 percent of name-plate power would provide more accurate sizing than
the 80 percent they had previously used for new demand requests.
As equipment manufacturers implement the energy savings features we
are familiar with in notebook computers (along with new server-based energy