High Performance Computing at Microsoft?

  • Article

So, you have the latest multi-core, multi GHz, multi-processor, multi-everything laptop. That is a high-performance computer, right? Well, almost:

  • High Performance Computing is a branch of computer science that studies systems designs and programming techniques to extract the best computational performance out of computer systems. It is usually associated with parallel computing either on specialized machines or on clusters of commodity machines.

Compute Cluster Server 2003 is our HPC offering. We’ve had it for about a year, we’re now at service pack 1, and we’re planning version 2.

It is made of two parts:

  • Windows Server 2003 Compute Cluster Edition: An x64 operating system that supports up to 4 CPUs and 32 GB RAM per instance. It is an artificially limited version of 2003 Standard, sold at a significantly reduced price for computational workloads only.
  • Compute Cluster Pack: The tools to build a HPC cluster, i.e. the job scheduler, management tools, API, message-passing (MPI) stack.

For developers, we offer:

  • Message-Passing Interface (MPI) libraries for parallel application development in the Compute Cluster SDK
  • A remote parallel debugger in Visual Studio 2005 Pro
  • Support for the OpenMP multi-threading specification in our C, C++ compilers with Visual Studio 2005

What is the strong point of Compute Cluster Server?

In one word: integration. We integrate with the Microsoft infrastructure tools for deployment & operations, with the development tools and with the end-user desktop environment.  Our platform is where the value is.

Getting Started

Books:

  • [A. Grama, A. Gupta, G. Karypis, V. Kumar], “Introduction to Parallel Computing”, 2nd edition, Addison Wesley, 2003
  • [I. Foster], “Designing and Building Parallel Programs”, Addison Wesley, https://www-unix.mcs.anl.gov/dbpp/

Specifications:

Sites: