Online resource kit >

HomeAt a GlanceDeveloperIT Pro

Parallel Extensions to the .NET Framework (PFx)

< Go Back

Windows HPC Server 2008 provides a powerful framework for developing and executing high-performance computing (HPC) applications. This tutorial presents an HPC scenario in which a sequential C++/C# developer is redesigning an application to take advantage of parallel processing. The discussion and exercises involve a number of high-performance, parallel solutions using a range of technologies, including Open Multiprocessing (OpenMP), Parallel Extensions to the .NET Framework (PFx), Windows Communication Foundation (WCF), and Windows HPC Server 2008. The complete tutorial includes lab exercises, program solutions, and miscellaneous support files.

The tutorial begins with a statement of the problem domain and a discussion of the problem that will be demonstrated with the Mandelbrot application (an example of a simple mathematical definition leading to complex behavior). The discussion and lab exercises then step through a sequential version of the application, and then use shared-memory parallel program and multi-threading in C# with the Task Parallel Library (TPL) to improve performance. The next sections involve a client-server version of the Mandelbrot application, first running the sequential client-server version, and then the sequential VC++ or C# version.

The tutorial then covers working with Windows HPC Server 2008, and submitting the Mandelbrot application to the cluster. See how to enable parallel, cluster-based execution of the application and experiment with the various parallelization strategies, such as the parametric sweep and multithreaded techniques (thread class, OpenMP, and TPL).

Finally, the tutorial examines the application through another perspective: service-oriented architecture (SOA)-based parallelization using the Windows Communication Foundation (WCF) and shows how to parallelize the C# version of the Mandelbrot application using the WCF-based capabilities of Windows HPC Server 2008.

View