Energy-efficient server consolidation for multi-threaded applications in the cloud

Cloud services have been actively used for transactional and batch workloads. Recently, multi-threaded high-performance computing (HPC) workloads have started to emerge on the cloud as well. Unlike most traditional data center loads, HPC workloads highly utilize the servers. The energy efficiency and performance of HPC loads, however, vary strongly as a function of the amount of allocated resources. This paper proposes an autonomous resource allocation technique for multi-threaded compute-intensive HPC workloads with the goal of creating tunable energy cost-performance tradeoffs for the cloud administrators and users. The proposed technique adjusts the available resources for the virtual machines (VMs) based on application energy efficiency while delivering the desired performance guarantees. Experiments on a real-life multi-core server show that the proposed technique improves the system throughput-per-watt by 17% on average compared to existing techniques.