Hardware Reference
In-Depth Information
ing
from the server manufacturer is always conservative; it's the maximum power a server can
draw. The first step then is to measure a single server under a variety of workloads to be de-
ployed in the WSC. (Networking is typically about 5% of power consumption, so it can be ig-
nored to start.)
To determine the number of servers for a WSC, the available power for IT could just be di-
vided by the measured server power; however, this would again be too conservative accord-
ing to
Fan, Weber, and Barroso [2007]
. They found that there is a significant gap between what
thousands of servers could theoretically do in the worst case and what they will do in prac-
tice, since no real workloads will keep thousands of servers all simultaneously at their peaks.
They found that they could safely oversubscribe the number of servers by as much as 40%
based on the power of a single server. They recommended that WSC architects should do that
to increase the average utilization of power within a WSC; however, they also suggested using
extensive monitoring software along with a safety mechanism that deschedules lower priority
tasks in case the workload shifts.
Breaking down power usage inside the IT equipment itself,
Barroso and Hölzle [2009]
re-
ported the following for a Google WSC deployed in 2007:
■ 33% of power for processors
■ 30% for DRAM
■ 10% for disks
■ 5% for networking
■ 22% for other reasons (inside the server)
Measuring Efficiency Of A WSC
A widely used, simple metric to evaluate the efficiency of a datacenter or a WSC is called
power
utilization effectiveness
(or
PUE
):
Thus, PUE must be greater than or equal to 1, and the bigger the PUE the less efficient the
WSC.
Greenberg et al. [2006] reported on the PUE of 19 datacenters and the portion of the over-
head that went into the cooling infrastructure.
Figure 6.11
shows what they found, sorted by
PUE from most to least efficient. The median PUE is 1.69, with the cooling infrastructure using
more than half as much power as the servers themselves—on average, 0.55 of the 1.69 is for
cooling. Note that these are average PUEs, which can vary daily depending on workload and
even external air temperature, as we shall see.
Search WWH ::
Custom Search