EDITOR’S QUESTION
WENDY TORELL, SENIOR
RESEARCH ANALYST, SCIENCE
CENTER, SECURE POWER DIVISION,
SCHNEIDER ELECTRIC
irst, it’s important to
understand what is
F
meant by ‘optimum
performance’. For most
data centres, availability
is a key performance
criterion, since the IT loads they support
are mission-critical and the cost of
downtime is high. However, there are
other performance metrics to consider as
well – for instance, how energy efficient
the data centre is, what its Power Usage
Effectiveness (PUE) rating is, how cost
effective it is in terms of CapEx, OpEx, or
total cost of ownership (TCO) and how
sustainable or ‘green’ the design is.
It’s important to begin the design planning
with discussions around all the key criteria
for the data centre’s performance to avoid
having to go back to the drawing board
and get the project implemented on time.
Key design attributes that can help
achieve optimal performance in these
various ways include:
• Efficient design – Operators should
select power systems (i.e. UPSs,
PDUs) that are designed for high
efficiency; and choose cooling
systems that allow for maximum
economiser cooling hours within their
geography. They should also pay
close attention to design choices that
impact air distribution (containment
systems, raised floors, etc.), as those
have a big impact on overall energy
efficiency of the data centre.
• Connected systems – Selecting
systems that utilise cloud-based,
mobile-friendly data centre
infrastructure management (DCIM)
software, which can be managed
remotely and is connected to a data
lake, allows the user to leverage
information beyond that of their
data centre for more predictive
maintenance vs. traditional calendarbased
maintenance. This helps to
maximise the life of components,
while ensuring availability since
components like batteries can be
replaced just before they fail.
• Redundant systems – If high
availability is necessary, operators
should design with redundancy
in mind; this ensures failures
of components or required
maintenance activities don’t bring
the critical load down.
• Scalable designs – Future loads are
often unknown, so having a design
that can easily scale is an important
attribute. Many UPSs are modular,
so as the load grows, additional
power modules can be added. Also
consider the current and future
density needs. As chip density
grows, can the cooling design
keep up? There is an increasing
trend towards liquid cooling as chip
densities climb above the limits of
air-cooled data centres.
• Standardisation – Using
standardised designs helps to ensure
predictable performance; unique
designs can mean unique problems.
Pre-integrated, pre-fabricated
systems also reduce data centre
deployment time, which may be an
important performance criterion.
Today, there are many resources freely
available to help decision-makers
understand the trade-offs of data centre
performance attributes. For instance, the
Schneider Electric Capital Cost tool or PUE
tool can help operators to see the drivers
of cost and efficiency, enabling the right
choices to be made given performance
priorities for the data centre(s).
30 Issue 19
www.intelligentdatacentres.com