This paper presents a performance analysis of cloud computing services for many-task scientific computing. The authors analyze the performance of four commercial cloud computing services, including Amazon EC2, GoGrid, ElasticHosts, and Mosso, for scientific computing workloads. They find that current clouds need an order of magnitude in performance improvement to be useful to the scientific community. They also compare the performance characteristics and cost models of clouds with other scientific computing platforms, such as grids and parallel production infrastructures, using trace-based simulation. The results indicate that clouds are not yet suitable for many-task scientific computing, and that significant improvements are needed to address the discrepancy between offer and demand. The paper also discusses the challenges of using clouds for scientific computing, including performance, and highlights the importance of understanding the characteristics of scientific computing workloads. The authors conclude that clouds are not yet a viable alternative for scientific computing, but that they have the potential to become one in the future with significant performance improvements.This paper presents a performance analysis of cloud computing services for many-task scientific computing. The authors analyze the performance of four commercial cloud computing services, including Amazon EC2, GoGrid, ElasticHosts, and Mosso, for scientific computing workloads. They find that current clouds need an order of magnitude in performance improvement to be useful to the scientific community. They also compare the performance characteristics and cost models of clouds with other scientific computing platforms, such as grids and parallel production infrastructures, using trace-based simulation. The results indicate that clouds are not yet suitable for many-task scientific computing, and that significant improvements are needed to address the discrepancy between offer and demand. The paper also discusses the challenges of using clouds for scientific computing, including performance, and highlights the importance of understanding the characteristics of scientific computing workloads. The authors conclude that clouds are not yet a viable alternative for scientific computing, but that they have the potential to become one in the future with significant performance improvements.