Installation Instructions for a Windows HPC Server 2008–Based Cluster on HP Proliant DL145 G2 Hardware
< Go Back
This document details the installation of Windows HPC Server 2008 on a set of HP Proliant DL145 G2 servers. In this configuration, the needed Active Directory server will be incorporated into the head node itself. This is representative of the first installation of a Windows-based cluster in an otherwise all Linux or UNIX-based shop.
The installation of a cluster based on Windows HPC Server 2008 has more steps than a similar installation using the Linux-Rocks+ package. However, the Windows-based installation is much more flexible, and it is possible to go through the entire cluster selecting defaults to simplify the process. Once the cluster is constructed, the Windows HPC Server 2008 provides some simple tests to determine how well the cluster can perform. Some of these tests can be used as diagnostics and tuning aids, which is a very useful addition to any cluster toolkit. The installation instructions described in this document detail each step of the process and include explanations, warnings, and screenshots.
The major steps of the installation process are as follows: check the hardware and firmware in each system. If possible, follow the homogeneity guidelines and check everything . Obtain the power consumption for your hardware configuration from the manufacturer or from actual measurements. Add a safety factor, and then configure your main circuits accordingly. Check the hardware arrangement, BIOS version (upgrade if the system is not running the newest version), and configuration and time settings. Clean the disks of any existing data. Configure the network. Install the keyboard, video, and mouse (KVM) switch. Install the head node. Follow the on-screen instructions to make the necessary settings. Again using the on-screen instructions, deploy, boot, and provision the compute nodes, checking to see that they appear in the correct order on the head node screen. Deploy the software to the compute nodes. Bring the cluster online, and run a simple diagnostic to make sure that everything is functioning correctly. This document provides in-depth instructions, explanations, and screenshots for each step.
