Frequently Asked Questions (FAQ) | Windows HPC

Microsoft Windows Compute Cluster Server 2003 is an integrated, standards-based compute clustering platform built on top of Windows Server 2003 x64 Edition.

Windows Compute Cluster Server 2003 is comprised of two CDs. The first CD is Windows Server 2003 Compute Cluster Edition (CCE), and the second CD is the Microsoft Compute Cluster Pack (CCP)—a combination of interfaces, utilities, and management infrastructure that makes up Windows Compute Cluster Server 2003.

Microsoft's goal is to empower end users by allowing them to easily harness distributed computing resources to solve complex problems. Windows Compute Cluster Server 2003 is the most recent step in that effort, as Microsoft continues to focus on making distributed computing resources easier to use, more efficient to deploy, and less expensive to operate.

Windows Compute Cluster Server 2003 is based on Windows Server 2003 Standard x64 Edition and can only be installed on 64-bit hardware.

Windows Compute Cluster Server 2003 Service Pack 1 (SP1) uses Windows Server 2003 SP2 as the foundation operating system. In addition to general bug fixes, the compute node deployment wizard of Windows Compute Cluster Server 2003 SP1 is based on Windows Deployment Services, which makes adding a compute node as easy as plugging it into the network and turning it on.

Purchasing information for Windows Compute Cluster Sever 2003 can be found on the How to Buy page and Pricing and Licensing page for this product. You can also download the trial version by following the instructions on the Windows Compute Cluster Server 2003 Trial Software page.

Minimum hardware requirements can be found on the Windows Compute Cluster Server 2003 system requirements page.

No. "High availability" describes systems designed for fault tolerance. Windows Compute Cluster Server 2003 is designed to support scientific and technical high-performance computing applications that take advantage of parallel processing.

Windows Compute Cluster Server 2003 hardware requirements are identical to the requirements for 64-bit versions of Windows Server 2003.

Windows Server 2003 Compute Cluster Edition (CCE) and Windows Server 2003 64-bit Standard Edition have the same code base and therefore, you do not need to re-test the HPC applications on CCE if you’ve already tested them on Windows Server 2003 64-bit Standard Edition. Patches and service packs are also the same across the 64-bit versions of Windows Server 2003 Standard, Enterprise and Compute Cluster editions. Please note that Windows Server 2003 Compute Cluster Edition has a different code base than Windows Server 2003 R2.

Windows Server CCE and Windows Server 2003 64-bit Standard Edition have different licensing terms. You may not use Windows Server 2003 Compute Cluster Edition (CCE) as a general purpose server, database server, e-mail server, print server or file server. In order to allow Windows Compute Cluster Server 2003 to be offered at a lower price, its server roles are restricted to computational use only. For example, if users want to install Microsoft SQL™ Server 2005 on a cluster node, they will need to purchase and install a full version of Windows Server 2003 64-bit Standard Edition or Windows Server 2003 64-bit Enterprise Edition on that cluster node. To maintain licensing compliance, Windows CCE takes advantage of a feature in Windows Server Standard to protect these applications from being executed. Please see the Windows Compute Cluster Server 2003 Pricing and Licensing page for more information.

Windows Compute Cluster Server is licensed for use with HPC applications. HPC applications solve complex computational problems using several servers as a group, also called a cluster, to solve a single computational problem or a single set of closely related computational problems. Applications that run on a single server are not considered HPC applications. Applications that are distributed across multiple servers may not be considered HPC applications, unless they are working on a set of closely related computational problems.

You may not use Windows Server 2003 Compute Cluster Edition (CCE) as a general purpose server, database server, e-mail server, print server or file server. In order to allow Windows Compute Cluster Server 2003 to be offered at a lower price, its server roles are restricted to computational use only. For example, if users want to install Microsoft SQL™ Server 2005 on a cluster node, they will need to purchase and install a full version of Windows Server 2003 64-bit Standard Edition or Windows Server 2003 64-bit Enterprise Edition on that cluster node. To maintain licensing compliance, Windows CCE takes advantage of a feature in Windows Server Standard to protect these applications from being executed. Please see the Windows Compute Cluster Server 2003 Pricing and Licensing page for more information.

Although Windows Compute Cluster Server 2003 requires 64-bit computers for use as compute nodes, 32-bit applications running on 64-bit compute nodes is a fully supported configuration. The 32-bit application will not be able to use the larger memory-addressing available on the compute node, but it will run. Developers will find both 32-bit and 64-bit Microsoft Message Passing Interface (MPI) libraries included in the Windows Compute Cluster Server 2003 Software Development Kit (SDK).

The application vendor is the best source for determining if your UNIX- or Linux-based application will run on Windows Compute Cluster Server 2003. Note that Microsoft Compute Cluster Pack in CCS can take advantage of 32- and 64-bit versions of Subsystem for UNIX-based Applications (SUA) on Windows Server 2003 Release 2 (R2), which may be required to run UNIX or Linux applications.

There are no plans at this time to release a 32-bit version of Windows Compute Cluster Server 2003.

Licensing for Windows Compute Cluster Server 2003 will be restricted to computers that are used as dedicated computational servers only. Customers wishing to use compute nodes for other purposes (for example, using a cluster head node as a database server) should purchase the appropriate version of Windows Server 2003 x64 Editions (Standard, Enterprise, or Datacenter) for installation on those nodes. The cluster management components encapsulated in the Microsoft Compute Cluster Pack support installation on all editions of Windows Server 2003 x64. Please see the Windows Compute Cluster Server 2003 Pricing and Licensing page for more information.

Yes. The Microsoft Compute Cluster Pack (in other words, the second CD of the two CDs that make up Windows Compute Cluster Server 2003) can be purchased separately and will run on 64-bit versions of Windows Server 2003 R2.

Yes. Windows Compute Cluster Server (CCS) 2003 supports the security requirements for these environments by leveraging Active Directory directory service for user authentication. CCS also provides a Microsoft Management Console plug-in that simplifies managing the head node and compute nodes, which make up a compute cluster. Now, tools such as Microsoft Operations Manager (MOM) and Microsoft Systems Management Server (SMS) can be used to manage CCS.

There are no special networking requirements for Windows Compute Cluster Server 2003. However, the product is designed to take advantage of high-speed networking technologies such as Gigabit Ethernet, Infiniband, 10 Gigabit Ethernet, Myrinet, and other interconnects by way of the Microsoft Windows Sockets Direct Interface.

Windows Compute Cluster Server 2003 comes with the Microsoft Message Passing Interface (MS MPI), an MPI stack based on the MPICH2 implementation from Argonne National Labs. Windows CCS 2003 will also work with other MPI stacks written to the MPI2 standard.

It is recommended that the Microsoft job scheduler be used with Windows Compute Cluster Server 2003. However, there are also third-party job schedulers that will work with Windows Compute Cluster Server 2003.

Yes. Use of the command line to create and submit jobs is supported. The command-line interface also supports submission of jobs to the scheduler by script or batch file.

Yes. Windows Compute Cluster Server 2003 compute nodes can be deployed using Remote Installation Services (RIS), which is included with Windows Compute Cluster Server 2003. For Windows Compute Cluster Server 2003 with SP1, RIS was replaced by Windows Deployment Services (WDS).

Customers can use Windows ntbackup to back up Windows Compute Cluster Server 2003 head nodes and compute nodes. Third-party backup applications that are certified to run on Windows Server 2003 products can also be used to back up Windows Compute Cluster Server 2003 nodes.

Each 'individual' cluster can have only one head node.

Windows Compute Cluster Server 2003 does not support head-node failover. However, high-availability functionality is being considered for future versions. Windows supports a number of techniques for recovering from a system failure, from using backup utilities to using disk mirroring.

There is no limit to the number of nodes in a Windows Compute Cluster Server 2003 cluster except for the number of hardware systems and node interconnects available, and the demands placed on the infrastructure by the applications running across the nodes.

Windows Compute Cluster Server 2003 comes with the Microsoft SQL Server Desktop Engine (MSDE 2000), which is installed by default and is used to track jobs. Due to licensing restrictions, customers cannot install the full version of SQL Server on compute nodes running Windows Compute Cluster Server 2003. Customers wishing to install full versions of SQL Server should purchase Windows Server 2003 x64 Standard, Enterprise, or Datacenter Edition. The Microsoft Compute Cluster Pack supports installation on these platforms.

Neither Windows Service for UNIX 3.5 nor Subsystem for UNIX-based Applications (SUA) (available with Windows Server 2003 R2) is required to use Microsoft Compute Cluster Pack in Windows Compute Cluster Server 2003. However, SUA can help you build and port high-performance computing applications to the Windows environment.

Supported programming languages for Microsoft MPI are Fortran77, Fortran90, and C. However, customers can use any language or scripting facility available from Microsoft or a Windows-oriented independent software vendor (ISV) to create and control applications on a compute cluster.

Windows Compute Cluster Server 2003 can be operated as a back end for a number of interactive parallel-computing programs with graphic user interfaces. How this is done and how tightly coupled the parallel processes are depends on the individual program.

No, but it should be kept in mind that an application started from a scheduler must be able to run unattended, with input from a file or a built-in default. This is a common requirement for any application run in batch mode or in the background. Interactive programs can run on Windows Compute Cluster Server 2003, but in this case, the scheduler is called by the application, which acts as a front end. A session-type application of this kind must have an interface to the scheduler and also a way to display the output.

Microsoft is currently working with its partners to port key applications to Windows Compute Cluster Server (CCS) 2003. Visit the high-performance computing partners page for a list of Microsoft CCS partners. For information on a specific application, please contact the vendor. For general information about an application area, please send e-mail to HPCinfo@microsoft.com.

Yes. In general, the lower the performance of the interconnect (both bandwidth and latency), the greater the requirement to limit inter-node communication—that is, to move from a tightly-coupled formulation of a problem to a loosely-coupled formulation. Many applications do not require tight coupling. For example, when you simultaneously run 100 SAS analyses, you are in fact running 100 independent jobs with no dependence on microsecond-level communication between the nodes during the computation. That said, most traditional massively parallel, scientific-related applications do rely on high-performance interconnect and will require significant investment in node interconnect.

There is little substantive difference, but UNIX-based solutions should be fully ported to Windows to realize the full benefits of the Windows operating system. There are several differences between UNIX-based operating systems and Windows. For example, I/O operations and threading are different on UNIX-based systems than they are on Windows. I/O intensive applications will benefit from using Windows native I/O APIs rather than UNIX style I/O APIs.

The head node hosts the scheduler, which dispatches tasks to the compute nodes. A parallel application (as opposed to multiple serial applications running in a parametric sweep) runs as a distributed application with a process on each of the nodes—one of them a master process that spawns the others and handles input and output files. When a massively parallel application has been started on all nodes, it can effectively be running peer to peer with its other instances on the other nodes.