Performance and monetary cost optimizations for HPC applications in the Cloud
Date of Issue2016-11-24
Benefit from cloud computing, high performance computing (HPC) tasks can be performed on virtual machines instead of a physical cluster. Because of the pay-as-you-go nature, performance and monetary cost optimizations are significant in not only improving productivity but also reducing ownership cost. This thesis focuses on both of them. For performance, virtualization hides network topology information and causes traffic interference. Many existing network optimizations that rely on topology information or stable network performance are no longer effective. So we develop novel performance optimization algorithms and further propose a novel network performance model to decouple the constant and volatility components from the dynamic network performance. For monetary cost, Amazon EC2 spot instances give us a chance to reduce costs for HPC applications. So we leverage both on-demand and spot instances to guarantee deadline constraint and develop a cost model guided approach to optimize the monetary cost.