Page 286 -

257
C
C h a p t e r 1 3 : h a p t e r 1 3 : G G r e e n i n g Yo u r I n f o r m a t i o n S y s t e m s r e e n i n g Yo u r I n f o r m a t i o n S y s t e m s 257

The PUE can range from 1.0 to infinity. In the best case, a PUE reaching 1.0 would show
100 percent efficiency. Currently, most datacenters are in the 3.0 and higher range, but
properly designed, a PUE of 1.6 is attainable.

Datacenter Density
Datacenter density is a measurement of your CPU cycles over square footage. The higher the
CPU cycles, the better. Datacenters with more rack density and higher space utilization score
higher. And as you virtualize and remove old equipment, you’ll see this number get larger.

Storage Utilization
This metric compares how much storage you are using, compared with how much you have
available to use. This can be a hard one to measure, because storage is such a moving target.
For instance, do you count virtual memory for live applications? How about backup data?
Duplicate data? It’s really up to you to decide what you will include in your calculations,
but it’s important to be consistent. That is, if you opt to use duplicate data in your initial
computations, it’s important to always include that data for an apples-to-apples comparison.

Storage Density
This is similar to datacenter density, but rather than measure CPU cycles per square foot,
you’re comparing the amount of storage to your datacenter’s area. This helps you determine
whether you are making the best use of your facilities.

CPU Utilization
This measurement examines the percentage of utilization of your servers’ CPUs. This can be
measured with a number of commercially available performance measurement tools. For
instance, you can track CPU utilization through System Monitor on Windows Server, as
shown in Figure 13-1.
If your server is running just one task, as the one in Figure 13-1 is, utilization is very
low, and not very efficient. As you add virtualized tasks to your server, utilization will
increase.
Optimally, you want server utilization to be somewhere over 50 percent, perhaps as
high as 80 percent, with the actual target depending on the applications, how predictable
demand is, and other factors. This space capacity gives you some headroom in case services
experience a spike in demand.

SWaP
When considering purchasing new servers, there’s more to consider than just overall PART V
performance. Sun Microsystems has developed a metric that calculates such issues as floor
space and power used, along with performance. The metric, known as Space, Watts, and
Performance (SWaP), is expressed as:
SWAP = PERFORMANCE/SPACE X POWER CONSUMPTION
On its own, the metric doesn’t mean a lot. For instance, let’s consider a server with the
following characteristics:

• 400 operations
• 4 rack units
• 300 watts