"It's the Platform, Stupid."
Through microscopic advances at the chip level, efficient computing can achieve large-scale results.
In May, I was privileged to attend
the 2009 Santa Barbara Summit on Energy Efficiency. Including a variety of
voices from private industry to government agencies, the summit brought together
the views of and information from stakeholders and industry experts in a forum
designed to “focus thought leadership in the energy sector,” while at the same
time promoting the development of new technologies for energy efficiency.
The topics ranged from lighting to
storage (with a special emphasis on the nexus between politics, technology, and
the economy), but what I found most compelling was the first day’s segment on
energy-efficient computing and data centers. Although data centers only account
for 2% of all US energy consumption, that small portion translates into a total
consumption cost of $4.5 billion. Additionally, according to a 2006 report by
the EPA, data centers use approximately 61 billion kWh per year.
As we continue to transform
ourselves from a paper-based to digital society, data centers—which are the
backbone of our information infrastructure—will continue to grow in size and
importance, and with that growth comes a responsibility to develop and maintain
energy-efficient protocols whenever possible. Hailing from some of the big names
in the computing industry—including Intel, Google, and Microsoft—the summit’s
opening day speakers discussed the variety of innovations being pursued and the
opportunities being made available, in the realm of energy-efficient computing
and data centers. Because their comments and suggestions have implications
beyond just data centers, I think it’s worthwhile to summarize what they
said.
Delivering the summit’s opening
keynote, Justin Rattner (Vice President and Chief Technology Officer at Intel)
laid out the two major challenges faced by the computing world: the need to
reduce power demand by one order of magnitude while maintaining chip performance
and, alternately, the need to increase chip performance by one magnitude while
keeping power demand constant. According to Rattner, the newest Netbooks to hit
the market already contain processors that fulfill the first portion of this
challenge, and it’s anticipated that the new processors that are currently in
the works are poised to fulfill the second part of the challenge within the next
few years.
These anticipated “multi-core”
processors have the potential to significantly impact energy usage at data
centers, because they are designed to provide automatic or manual core control.
This core control allows these new processors to respond to power demands in
real time, thereby creating the possibility of active, intelligent power
management at the processor level. For example, high-voltage DC power
distribution can reduce the amount of AC/DC conversion and result in a 7–8%
increase in efficiency.
Luiz Barroso, Distinguished
Engineer for Google Inc., elaborated on the efficiency opportunities available
in large data centers. Barroso discussed how energy must be proportionate to
components (so that less power is needed during “sleep” mode for example).
Because the energy footprint of data centers is “no longer negligible,” Barroso
emphasized that data centers must be proactive in adopting efficiency practices.
For Google, that means refocusing efficiency research with special attention
paid to data centers’ special needs. After six months of implementation, Google
Inc. recently released the results of its efficiency efforts: a ratio of Power
Use Efficiency (PUE)—the amount of energy entering a given building divided by
the amount of energy needed for IT equipment—reduced to 1.16, almost half of the
industry standard 2.0 PUE.
Wrapping up the computing portion
of the summit’s program, Feng Zhao (Principal Researcher for Microsoft Research)
went to the heart of the matter: how efficiency at the milliwatt level results
in megawatt savings. Discussing the development of a modular sensor platform
(which resembles a Lego-like structure), Zhao explained that accurate
forecasting can result in bigger and better savings. A “Joule Meter” profiling
application, for example, can anticipate performance events and can result in
applications that are more energy-efficient.
As I stated earlier, the
successful implementation of energy efficiency protocols at data centers can
provide a solid foundation for broader application of these protocols beyond the
computing environment. Rattner even pointed out during his keynote that the
success of hardware control could be applied to the big energy consumers,
specifically by applying these computing efficiency tenets to buildings and the
grid, while simultaneously empowering the energy consumer.
Advertisement
In order to survive under a new
national energy policy that calls for accountability and quantifiable results,
data center operators will no doubt begin to focus even more intently on energy
efficiency. And because technological innovation is often the engine that powers
change, energy-efficient computing has the power to reach beyond the confines of
a cooled server room to influence and inspire other energy-intensive industries
users how onsite power can mitigate the cost of acquiring energy from a
centralized supply (like the grid).
And
the benefits wouldn’t be restricted to a reduction in energy demand. Reiterated
often by many of the summit’s speakers and participants was the notion that
reducing demand through new technologies can stimulate the economy, while
alternately decreasing our dependence on foreign oil and helping resolve some of
the issues associated with climate change and carbon emissions.
Author's Bio: Elizabeth Cutright is the Editor of Distributed Energy Magazine
July-August 2009
"It's the Platform, Stupid."
Through microscopic advances at the chip level, efficient computing can achieve large-scale results.
In May, I was privileged to attend
the 2009 Santa Barbara Summit on Energy Efficiency. Including a variety of
voices from private industry to government agencies, the summit brought together
the views of and information from stakeholders and industry experts in a forum
designed to “focus thought leadership in the energy sector,” while at the same
time promoting the development of new technologies for energy efficiency.
The topics ranged from lighting to
storage (with a special emphasis on the nexus between politics, technology, and
the economy), but what I found most compelling was the first day’s segment on
energy-efficient computing and data centers. Although data centers only account
for 2% of all US energy consumption, that small portion translates into a total
consumption cost of $4.5 billion. Additionally, according to a 2006 report by
the EPA, data centers use approximately 61 billion kWh per year.
As we continue to transform
ourselves from a paper-based to digital society, data centers—which are the
backbone of our information infrastructure—will continue to grow in size and
importance, and with that growth comes a responsibility to develop and maintain
energy-efficient protocols whenever possible. Hailing from some of the big names
in the computing industry—including Intel, Google, and Microsoft—the summit’s
opening day speakers discussed the variety of innovations being pursued and the
opportunities being made available, in the realm of energy-efficient computing
and data centers. Because their comments and suggestions have implications
beyond just data centers, I think it’s worthwhile to summarize what they
said.
Delivering the summit’s opening
keynote, Justin Rattner (Vice President and Chief Technology Officer at Intel)
laid out the two major challenges faced by the computing world: the need to
reduce power demand by one order of magnitude while maintaining chip performance
and, alternately, the need to increase chip performance by one magnitude while
keeping power demand constant. According to Rattner, the newest Netbooks to hit
the market already contain processors that fulfill the first portion of this
challenge, and it’s anticipated that the new processors that are currently in
the works are poised to fulfill the second part of the challenge within the next
few years.
These anticipated “multi-core”
processors have the potential to significantly impact energy usage at data
centers, because they are designed to provide automatic or manual core control.
This core control allows these new processors to respond to power demands in
real time, thereby creating the possibility of active, intelligent power
management at the processor level. For example, high-voltage DC power
distribution can reduce the amount of AC/DC conversion and result in a 7–8%
increase in efficiency.
Luiz Barroso, Distinguished
Engineer for Google Inc., elaborated on the efficiency opportunities available
in large data centers. Barroso discussed how energy must be proportionate to
components (so that less power is needed during “sleep” mode for example).
Because the energy footprint of data centers is “no longer negligible,” Barroso
emphasized that data centers must be proactive in adopting efficiency practices.
For Google, that means refocusing efficiency research with special attention
paid to data centers’ special needs. After six months of implementation, Google
Inc. recently released the results of its efficiency efforts: a ratio of Power
Use Efficiency (PUE)—the amount of energy entering a given building divided by
the amount of energy needed for IT equipment—reduced to 1.16, almost half of the
industry standard 2.0 PUE.
Wrapping up the computing portion
of the summit’s program, Feng Zhao (Principal Researcher for Microsoft Research)
went to the heart of the matter: how efficiency at the milliwatt level results
in megawatt savings. Discussing the development of a modular sensor platform
(which resembles a Lego-like structure), Zhao explained that accurate
forecasting can result in bigger and better savings. A “Joule Meter” profiling
application, for example, can anticipate performance events and can result in
applications that are more energy-efficient.
As I stated earlier, the
successful implementation of energy efficiency protocols at data centers can
provide a solid foundation for broader application of these protocols beyond the
computing environment. Rattner even pointed out during his keynote that the
success of hardware control could be applied to the big energy consumers,
specifically by applying these computing efficiency tenets to buildings and the
grid, while simultaneously empowering the energy consumer.
In order to survive under a new
national energy policy that calls for accountability and quantifiable results,
data center operators will no doubt begin to focus even more intently on energy
efficiency. And because technological innovation is often the engine that powers
change, energy-efficient computing has the power to reach beyond the confines of
a cooled server room to influence and inspire other energy-intensive industries
users how onsite power can mitigate the cost of acquiring energy from a
centralized supply (like the grid).
And
the benefits wouldn’t be restricted to a reduction in energy demand. Reiterated
often by many of the summit’s speakers and participants was the notion that
reducing demand through new technologies can stimulate the economy, while
alternately decreasing our dependence on foreign oil and helping resolve some of
the issues associated with climate change and carbon emissions.