Menu
Compare the best free open source Windows Clustering Software at SourceForge. Free, secure and fast Windows Clustering Software downloads from the largest Open Source applications and software.
This guide. describes how to build a small Windows HPC cluster that you can use to run your parallel software applications, cluster-enabled Microsoft Excel workbooks, or service-oriented architecture (SOA) based applications. This guide is intended for those of you out there who want to set up a small development or proof-of-concept cluster, or for those of you who do not have an IT department to set up a cluster for you, and so you need to do it yourself. After you complete the steps in this guide, you will have a cluster that is ready for you to install applications and start running jobs. Many parallel computing software applications provide documentation about how to install their software on a Windows HPC cluster. If you have all the required parts and permissions, and you already have the Windows Server 2008 R2 operating system installed on your computers, the steps in this guide should take about one to two hours to complete. In this guide:.
The plan. What you need.
Set up hardware and join a domain. Install Microsoft HPC Pack 2008 R2 to create the head node. Perform the initial configuration tasks on the head node. Pre-configure the compute nodes. Join the nodes to the cluster.This guide is the companion to the video. In this guide, we will walk through the steps to build an HPC cluster that consists of at least three nodes (computers with the server operating system installed). One node acts as a head node (to manage the cluster) and as a compute node (to run jobs).
Additional nodes act only as compute nodes. Note: The head node also acts as a WCF broker node to manage SOA service requests and responses.
The WCF broker functionality is necessary if you are using the to offload your UDF calculations or Microsoft Excel spreadsheets, or if your application uses the SOA programming model. The cluster nodes are connected to each other on an isolated, private network. This network is basically used to coordinate the workload on your cluster. The head node has an additional network connection to the enterprise network (a network with an Active Directory domain controller that manages logging in, security, and authentication in your work environment). After your cluster is set up, you can submit and monitor jobs by logging in to the head node directly, or by connecting to the cluster from another computer on the enterprise network (if you install the HPC Pack client utilities on that computer). Because we are building a small cluster, and to keep things simple, we will build a cluster with preconfigured compute nodes. That means that the computers that we add to the cluster already have the operating system installed, and we will manually install the HPC software on each node.
Note: If you are building a larger cluster, then read about. In a bare metal deployment, you take some extra steps to set up automated installation of the operating system and HPC software to all your compute nodes. Note: You can reopen the Initial configuration task console by opening a command prompt window and then typing “oobe”.
In the System Properties dialog box, in the Computer Name tab, click the Change button. In the Computer Name/Domain Changes dialog box, type a name for the head node, for example “HEADNODE”. Select Domain, type the name of your domain, and then click OK. When prompted, type the domain credentials that have permission to join the server to the domain.
A dialog box appears to confirm the domain join. When prompted, restart the computer to apply these changes. Connect the private network:. Plug the switch in to a power source. Connect all nodes to the switch. Log on to the head node with your domain credentials.
Insert the installation media for Microsoft HPC Pack 2008 R2 and run setup (setup.exe). Follow the steps in the installation wizard, and select the following options:.
In Select Installation Edition, select HPC Pack 2008 R2 Express or HPC Pack 2008 R2 Enterprise and HPC Pack 2008 R2 for Workstation. Note: Both editions work for creating a head node. Select the Enterprise edition if you want to use or explore the features for Excel offloading and for adding workstation computers to your cluster (this lets you use workstation computers that are on the enterprise network to run HPC jobs during non-working hours). Accept the license terms. In Select Installation Type, select Create a new HPC Cluster by creating a head node. In the next three screens, accept the default installation locations for the HPC databases and for HPC Pack. In Select Method for Updates, select Use Microsoft Update.
In Customer Experience Improvement Program, select Yes (this helps us improve the product in later versions). At the end of the wizard, click Finish. The wizard starts HPC Cluster Manager, which is the administration console for configuring and monitoring your cluster. In HPC Cluster Manager, in the Deployment To-do List, click Configure your network. In the Network Configuration Wizard:. In Network Topology Selection, the wizard detects the installed adapters and preselects a topology accordingly. Verify that the selection is as expected (in this guide, we built a Topology 1 cluster).
In Enterprise Network Adapter Selection, in the Network adapter drop-down list, select the network connection that you named “Enterprise”. In Private Network Adapter Selection, select the network connection that you named “Private”. In Private Network Configuration, accept the default settings. In Firewall Setup, accept the default settings.
Click Configure. In the to-do list, click Provide installation credentials. In the dialog box, you can type the same domain credentials that you used to join your head node to the domain. When you type your user name, include the domain name in the form DOMAIN User. Note: This procedure assumes that the nodes already have the Windows Server 2008 R2 operating system installed. Switch the keyboard, monitor and mouse to the compute node. Log on to the node as an administrator.
In the Initial configuration tasks console, select Enable remote desktop. Rename the node, for example NODE01. Join the node to the domain and then restart the computer when prompted.
Log on to the node with your domain credentials. Insert the installation media for Microsoft HPC Pack 2008 R2 and run setup (setup.exe). Follow the steps in the installation wizard, and select the following options:.
In Select Installation Edition, select HPC Pack 2008 R2 Express or HPC Pack 2008 R2 Enterprise and HPC Pack 2008 R2 for Workstation. Note: Both editions work for creating a compute node. Select the Enterprise edition if you want the compute nodes to be able to run Excel offloading jobs. Accept the license terms. In Select Installation Type, s elect Join an existing HPC Cluster by creating a new compute node. In Join Cluster, select or type the name that you gave to your head node (for example “HEADNODE”).
In the next screen, accept the default installation location for HPC Pack. In Select Method for Updates, select Use Microsoft Update. At the end of the wizard, click Finish. Switch the keyboard, monitor and mouse to the head node. In HPC Cluster Manager, click Node Management. Your preconfigured nodes appear in the node list.
At this point, the nodes have Node State marked as Unknown, and Node Health as Unapproved. In the node list, select the new nodes, right-click your selection, and then click Assign node template.
In the dialog box, select the node template that you created, “Preconfigured nodes”, and then click OK. To monitor the provisioning process:. Select a node in the node list. In the Details Pane, select the Provisioning tab. Provisioning details are displayed. When provisioning is complete, the nodes have Node State marked as Offline, and Node Health as OK. Select all the nodes, right-click your selection, and then click Bring Online.
Note: Bringing a node to the Online state tells the HPC Job Scheduler Service that you want this node to run jobs. Keep in mind that Online only reflects the intended usage of the node, and that the node must also be healthy and reachable. If you want to stop using a node for jobs, then bring the node to the Offline state. Now you’re ready to install your software applications or cluster-enabled workbooks! For information about managing the cluster and submitting jobs, see the following sections in the:.
Contents. General information Software Maintainer Category Development status ArchitectureOCS / Platforms supported Cost Paid support available, SysFera, Open Source All in one GridRPC, SPMD, Hierarchical and distributed architecture, CORBA HTC/HPC, Free Mavimax, Ltd. Job/Data Scheduler actively developed SOA Grid HTC/HPC/HA GPLv2 or Commercial Linux, FreeBSD, MacOS, Solaris, AIX Free / Cost Yes Monitoring actively developed, NT/XP/2000/2003/2008, FreeBSD, NetBSD, OpenBSD, DragonflyBSD, Mac OS X, Solaris, AIX, IRIX, Tru64, HPUX.