What's New in Windows HPC Server 2008

  • Article

Applies To: Windows HPC Server 2008

Several new features in Windows® HPC Server 2008 help improve the productivity and reduce the complexity of your high performance computing environment. The following sections summarize new and changed features to help you get up to speed quickly.

Product name changes in Windows HPC Server 2008

The names used to refer to this product and product components have changed in this version of the product. The changes are summarized as follows:

  • Windows Server® 2008 HPC Edition has replaced Windows Server® 2003, Compute Cluster Edition.

  • Windows® HPC Server 2008 has replaced Windows Compute Cluster Server 2003.

  • Microsoft® HPC Pack 2008 has replaced Microsoft Compute Cluster Pack 2003.

Administrative tools

The administrator console (HPC Cluster Manager) in Windows HPC Server 2008 is now based on the Microsoft System Center user interface. The new interface integrates all aspects of cluster management and simplifies the completion of essential tasks such as the configuration of cluster networking, setting of user permissions, and monitoring of jobs and the operational health of the cluster. The HPC Cluster Manager contains several new features that improve the experience of deploying and managing nodes:

  • Node templates that use Windows Deployment Services to significantly improve the deployment of compute nodes.

  • Customizable node groups that help you to view and manage collections of nodes.

  • A new heat map view feature that allows you to view cluster health at a glance.

  • New cluster and node diagnostics

  • Built-in extensible reporting

  • The ability to configure two head nodes for high availability so that jobs can continue running in the case of head node failures, and new jobs can be submitted within a short period of time after a failure.

  • Support for cluster updates, including the ability to specify when nodes are updated so that running jobs are not interrupted.

Compute Cluster Server 2003 Equivalent:

  • The Compute Cluster Administrator enables simpler completion of essential tasks such as configuring cluster networking topology, imaging and reimaging compute nodes, setting user permissions, and monitoring jobs.

  • Remote Installation Services (RIS) is used when deploying nodes using the automatic method.

For more information, see the HPC Cluster Manager Help (https://go.microsoft.com/fwlink/?LinkId=124146).

Networking

New wizard-driven configuration of network services such as Dynamic Host Configuration Protocol (DHCP), network address translation (NAT), Domain Name System (DNS), and firewall settings. Plus, support for NetworkDirect, a new Remote Direct Memory Access (RDMA) networking interface built for speed and stability.

Compute Cluster Server 2003 Equivalent: Wizard-driven configuration of network services such as Dynamic Host Configuration Protocol (DHCP), Internet Connection Sharing (ICS), Remote Installation Services (RIS), and firewall settings.

For more information, see the Windows HPC Server 2008 Design and Deployment Guide.

Cluster user tools

Windows HPC Server 2008 includes support for all Windows Compute Cluster Server 2003 command-line interface (CLI) commands. It also includes:

  • A new user interface for managing jobs (HPC Job Manager), with improved support for parametric commands.

  • PowerShell commands for scheduling and managing jobs. More than 130 command-line tools enable the automation of system administration tasks.

  • The ability to customize how and what jobs are displayed.

Compute Cluster Server 2003 Equivalent: Compute Cluster Job Manager snap-in made submission and tracking of user jobs simple.

For more information, see the Windows HPC Server 2008 Command Reference and Windows HPC Server 2008 PowerShell Reference.

Job Scheduling and Job Policies

  • Resource matching allows jobs to specify the types of resources on which they need to run.

  • Job templates for setting constraints and requirements on jobs, including the ability to route jobs to specific sets of resources. Job templates also simplify end-user job submission by enabling cluster administrators to specify default job values.

  • Multilevel resource allocation allows jobs to take better advantage of multi-core systems by requesting resources at a level appropriate for their performance characteristics.

  • Adaptive allocation of running jobs allows jobs with multiple tasks to shrink and grow as resources become available and work is completed.

  • Preemption allows high priority jobs to start sooner by taking resources away from lower priority jobs.

Compute Cluster Server 2003 Equivalent: "First-come, first-served" scheduling, backfilling of jobs, exclusive scheduling, and license scheduling.

For more information, see the Operations section (https://go.microsoft.com/fwlink/?LinkId=120726) of the Windows HPC Server 2008 Technical Library.

Interoperability

The HPC Basic Profile web service provides the following features using open standards and specifications:

  • Job Submission

  • Job Monitoring

  • File Staging

Compute Cluster Server 2003 Equivalent: None

For more information, see the HPC Server Basic Profile Web Service Operations Guide (https://go.microsoft.com/fwlink/?LinkId=122648).

Message Passing Interface (MPI)

  • Programming interface that is identical to MPICH2

  • Improved cluster efficiency to shorten solution time

  • Integrated with NetworkDirect, a new, high-speed Remote Direct Memory Access (RDMA) networking interface for Windows

  • New implementation of shared memory communications that provides better support for nodes with many cores.

Compute Cluster Server 2003 Equivalent: Programming interface that is identical to MPICH2.

For more information, see the Windows HPC Server 2008 Software and Driver Development Kit (SDK). This SDK is available for download at the Windows HPC Server 2008 site on Microsoft Connect (https://go.microsoft.com/fwlink/?LinkID=119579).

Developer tools

The following changes have been added in this version.

  • Service Oriented Application (SOA) programming platform

  • Scalable HPC Pack 2008 API with eventing support

  • Microsoft-Message Passing Interface (MS-MPI) events are now integrated with Event Tracing for Windows. This enhancement helps you to troubleshoot and fine tune the performance of your applications without having to use special traces or debug builds. For more information about the new MS-MPI tracing capability, see the Windows HPC Server 2008 Software and Driver Development Kit (SDK). This SDK is available for download at the Windows HPC Server 2008 site on Microsoft Connect (https://go.microsoft.com/fwlink/?LinkID=119579).

  • Customizable default job template

In addition, the new version includes support for all COM APIs in Windows Compute Cluster Server 2003.

The SDK now contains ten scenario-based, developer-focused tutorials covering the following topics:

  • Newcomers to HPC

  • Classic HPC programming using C++

  • Classic HPC programming using Fortran

  • HPC programming using C#/MPI.net

  • HPC programming using F#/MPI.net

  • HPC Programming using Python/MPI.net

  • Porting applications from Unix/Linux to Windows HPC

  • Debugging HPC programs using Allinea DDT

  • Profiling/Tracing using Event Tracing for Windows

  • Using Math libraries on Windows HPC

Compute Cluster Server 2003 Equivalent: .NET (C#) and COM application programming interface (API) for the CCS 2003 Job Scheduler and Microsoft Message Passing Interface (MS-MPI).

For more information, see the Windows HPC Server 2008 Software and Driver Development Kit (SDK) (https://go.microsoft.com/fwlink/?LinkId=119579).

Compatibility with previous versions

The following list describes compatibility between Windows HPC Server 2008 and Windows Compute Cluster Server 2003:

  • All of the command-line tools in Windows Compute Cluster Server 2003 (such as Cluscfg for configuration, and Clusrun for running commands remotely on one or more nodes) are still included.

  • Windows HPC Server 2008 provides API-level compatibility for applications integrated with Windows Compute Cluster Server 2003. These applications might, however, require changes to run on Windows Server® 2008. If you encounter problems running your application on Windows Server 2008, you should consult your software vendor.

  • Windows HPC Server 2008 supports job submission from Windows Compute Cluster Server 2003 clients, including the command-line tools, the Compute Cluster Job Manager, and the COM APIs.

  • The Windows HPC Server 2008 client tools, including the cluster administration console (HPC Cluster Manager), the job scheduling console (HPC Job Manager), the command-line tools, and the APIs cannot be used to manage or submit jobs to a Windows Compute Cluster Server 2003 cluster.

  • Clusters that have Windows Compute Cluster Server 2003 nodes and Windows HPC Server 2008 nodes are not supported.

Additional considerations

  • A side-by-side installation of Windows HPC Server 2008 and Windows Compute Cluster Server 2003 on the same computer is not supported. This includes the Windows HPC Server 2008 client utilities.

  • The upgrade of a Windows Compute Cluster Server 2003 head node to a Windows HPC Server 2008 head node is not supported.

  • The installation of HPC Pack 2008 adds the following server roles to the head node:

    • DHCP

    • Windows Deployment Services

    • File Services

    • Network Policy and Access Services, which enables Routing and Remote Access so that NAT services can be provided to the cluster nodes.