Toward green data centers as an interruptible load for grid stabilization in Singapore

For sustainability and environmental friendliness, renewable energy (RE) and distributed generation (DG), for example, photovoltaic, are being integrated in electrical systems in many countries. RE and DG, however, can be unstable for the power grid. As the power grid integrates an increasing amount of RE and DG, we present technical solutions along with an economic incentive model to enable data centers to serve as a novel "interruptible" load (i.e. a power load that can be scaled down temporally) to stabilize the power grid. We propose a novel real-time power analytics framework called embedded software as sensors, where software hooks are embedded into a range of data center subsystems, from chip to system to application level, to monitor ICT activities and power usage in a fine grained, real-time manner. Data from these virtual sensors are then mined to construct energy consumption models, which in turn are used to develop optimal algorithms for energy aware operation of computing, power distribution, and cooling systems in the data center. This holistic monitoring and optimization framework can reduce the overall power consumption of the data center, and furthermore enables time shifting of workloads in the data center in response to power fluctuations arising from the integration of RE and DG to the power grid.