A Python Calculation Grid Based On Service-Oriented Parallel Computing

We have been seeing ever-growing demands on computing capability. However, due the limitation of heating and power consumption, the improving of performance for single chip has met the bottle neck. It's under this circumstance that the multi core system and x86 based PC cluster has entered the main stream. But, for the past decade, we haven't seen any revolution on software and programming side to solve the parallel programming complexity. Programmers are still paying a lot of attention to hardware and distributing computing details.Windows HPC Server 2008 has provided a service-oriented parallel computing model. This paper introduces the implementing techniques, high level architecture, and the main advantage of a Python based calculation grid. And we emphasize on the zero-deployment/simple API feature thanks to advanced language feature of Python. Also, the service oriented architecture make the Python calculation grid suitable for the modern internet infrastructure and enable the grid to use computing resource across platform.After introducing the product, the paper introduces a real world performance test on the Python calculation grid. Using the Amdahl law, we analyze the experiment data, and propose efficient ways to use the Python calculation grid.Finally, the paper take a forward look on the next step of the Python calculation grid, including improved manageability, smart analysis on cluster performance, and support on QoS, etc.