Growing workloads and rising
energy costs have made data centre energy efficiency a critical
issue for many organizations, especially for those with older
facilities that lack the power and thermal capacity needed to
support growing needs.
The CERN computing centre was
built over 35 years ago and, even after refurbishment, is only
capable of providing 2.5 MW of electrical power and 2.5 MW of
cooling capacity. Furthermore, CERN needs a very powerful
computing infrastructure to handle the 10 - 15 Petabytes of data
per year that are to be generated by its LHC experiments. Thus, in
order to provide the maximum computing resources within this
limit, CERN is using a variety of techniques and tools to
achieve the best possible power efficiency.
Andreas Hirstius,
Sverre Jarp and
Andrzej Nowak authored last February a technical report
giving an overview of CERN's approach to energy efficient
computing, entitled 'Strategies
for increasing data centre power efficiency'.
In this report, as part of the
different techniques and tools presented, the authors
highlighted how CERN benefitted from the move to multi-core
processors with a more efficient architecture. Indeed the
multi-core processors based on the Intel Core microarchitecture
enabled CERN to provide five times more compute power within the
same compute power per Watt than singlecore processors based on
the earlier Netburst microarchitecture. Through the openlab, to
maximize data centre performance per Watt, CERN and Intel also
partnered on dynamic frequency control, virtualization and
software optimization.
Intel recently published on its
website an
abstract of the report entitled 'Reducing Data Center Energy
Consumption, A summary of strategies used by CERN, the world's
largest physics laboratory' and a presentation on the
following webpage.
|