Introduction
Saving power in data centers is a high priority for
data center operators and consumers of data center services, as well
as utilities, government regulators, and electricity consumers
because of the significant and growing economic and environmental
costs associated with the increasing demand and energy consumption.
According to the Lawrence Berkeley National
Laboratory 2008 report, in 2005 the direct electricity used by
information technology equipment in data centers represented about
0.5% of total world electricity consumption. When electricity for
cooling and power distribution is included, that figure is about 1%.
To date data centers consume about 2 percent of the world's
electricity.
According to the latest available report from the
U.S. Environmental Protection Agency (EPA 2007) this sector consumed
about 61 billion kilowatt-hours (kWh) in 2006 (1.5 percent of total
U.S. electricity consumption) for a total electricity cost of about
$4.5 billion. Federal servers and data centers alone account for
approximately 6 billion kWh (10 percent) of this electricity use, for
total electricity cost to U.S. taxpayers of approximately $450
million annually. Data centers use 150-750 W/ft2, corresponding to
2.5% of current US electricity consumption
Objectives
The project aim is to demonstrate the feasibility of
deploying a network of Performance Optimized Datacenters (PODs),
geographically distributed to exploit the availability of renewable
energy for their operation. Such a distributed system has the
potential to significantly enhance the energy efficiency,
reliability, security, and overall performance of the data center by
several means, including optimizing the utilization of the available
renewable power for computing by intelligently redistributing
computational load depending on the availability of renewable energy
and minimizing losses associated with power transmission by placing
the PODs near the power source. This concept provides data center
operators the means to avoid performing expensive utility upgrades as
the availability of wind power through New York State and other
states grows, keeping the infrastructure and Transmission &
Distribution (T&D) costs low, thus making it possible to use the wind
power that is currently stranded, i.e. not-delivered to the grid due
to the T&D constraints.
During this demonstration project, Clarkson
University in partnership with Advanced Micro Devices (AMD) will
demonstrate the feasibility of deploying a network of Performance
Optimized Datacenters (PODs), geographically distributed to exploit
the availability of renewable energy for its operation. Such a
distributed system has the potential to significantly enhance the
energy efficiency, reliability, security, and overall performance of
the data center by several means:
- Optimizing the utilization of the available renewable power for computing by intelligently redistributing computational load depending on the availability of renewable energy;
- Minimizing losses associated with power transmission by placing the PODs near the power source;
This concept provides data center operators the
means to avoid performing expensive utility upgrades as the
availability of wind power through New York State and other states
grows, keeping the infrastructure and Transmission & Distribution
(T&D) costs low, thus making it possible to use the wind power
that is currently stranded, i.e. not-delivered to the grid due to the
T&D constraints.
Benefits
Optimizing compute load shifting in this manner has
the potential to generate significant efficiencies for the
utilization of renewable energy sources and data center operations
and potentially create a new economic model for both renewable energy
and data centers. Under this new model, the high costs associated
with transmission infrastructure for renewable energy is replaced by
relatively low cost and highly efficient data transmission
infrastructure.
The main technological benefit of the proposed
project is the development of a distributed POD network that can be
powered from renewable energy, primarily wind, potentially consuming
a significant amount of constrained power that is not otherwise
beneficially utilized