The Windows HPC Server 2008-based Cluster in a Linux Environment
< Go Back
This paper details the process of integrating a Windows HPC Server 2008–based cluster into a Linux environment and executing high-performance computing (HPC) applications in both environments through the use of the Sun Grid Engine (SGE) workload management system. To help you replicate this process, the steps start from bare metal (i.e., empty systems), and notes are provided as to why each step is required.
The paper describes the following major steps of the process:
- Windows-based cluster installation
- Installation of development tools for Windows-based applications (optional)
- Review of a typical Linux cluster installation
- Installation of the necessary Linux–to-Windows interface for the cluster management software on the Linux cluster
- Review of the configuration of a Linux workstation to ensure the user can easily access the Windows environment
- Configuration of the Linux job management software (SGE) to allow the submission of batch jobs to either the Linux cluster or to the Windows-based cluster
The final target application is the HP LINPACK benchmark from the University of Tennessee, which is used by multiple cluster owners as a first trial of a new cluster. HP LINPACK helps verify that the cluster is fully functional and gives an indication of the ultimate performance of the cluster.
You should be able to replicate this integrated system with a minimum of four systems (two for each cluster). This can be accomplished with four real computers or with four virtual machines using a virtualization package like Microsoft Hyper-V Server.
Several lengthy appendices detail how to install and configure various components, programs, and packages related to integrating a Windows HPC Server 2008–based cluster into a Linux environment.
