Ultra low energy cloud computing using adaptive load prediction

The explosion of cloud computing networks throughout the world has lead to a need to reduce the sizeable energy footprint of cloud systems. We discuss a research investigation leading to ultra-low power cloud computing systems. Our methods apply to system such as those used in data centers and web hosting companies. Our analysis indicates massive power reductions up to 80% when optimal dynamical allocation of data center components occurs. We base this upon the application of adaptive load prediction and smart task distribution systems that can be built from current commercial off the shelf (COTS) components integrated with our new concepts. We show that adaptive prediction algorithm in ultra-low power cloud models, coupled with optimal task allocation, leads to design methods for cloud computer architectures optimized around low latency, lower power, and energy dissipation proportional to workloads.