Energy efficiency in data centers has a broad impact including cost savings and environmental benefits. The project investigates techniques for improving the energy efficiency in high-end data server environments. Specific focus is on the disk and memory subsystem running multithreaded database applications on multiprocessor architectures. Multiprocessor phase characterization and prediction being investigated will also impact performance issues in multiprocessor software. The research extends the current focus beyond sequential applications on uniprocessor architectures, models the effects of DMA and network memory accesses, studies application cooperation strategies for energy savings, and employs more realistic evaluation methodologies.
