Operating systems used in clusters

What fits into the existing IT?

Linux is tried and tested as a cluster operating system but now has some serious competition in the shape of Windows Compute Cluster Server. In the past, continuous availability, the large number of supported network devices and protocols and early support for multi-processor systems were strong arguments in favour of Linux but these now carry much less weight as the Windows Compute Cluster Server offers the same benefits.

What points need to be considered?
While most of the fundamental specifications of the Linux and Microsoft cluster operating systems are comparable, there are other criteria that are crucial when it comes to choosing the right operating system.
In general, it is important to ensure consistency between the front-end and the back-end.

• For example, if a user is familiar with the user interface and operation of Microsoft operating systems, he can still use the familiar Microsoft user interface in applications that are cluster ready and in which the HPC cluster therefore appears as a high-performance workstation.
• Internally, this enables calculations with Excel to be carried out far more quickly.
• HPC clusters can be more easily incorporated into an existing IT environment, for example if an active directory domain is already present in the environment, the cluster can simply be integrated into it, significantly reducing the amount of administration work.

Windows Compute Cluster Server

The new Windows Server 2008 offers users exactly what they would expect from a Microsoft operating system. Configuration is carried out using menus, while the display is based on graphics rather than text and users can find their way around right away.
The Windows Compute Cluster Server is based on Windows Server, with only three additional packages.

Components of Windows Compute Cluster Server

• The administration console...
• ...allows convenient management.
• ...provides a clear indication of the operating status.
• ...and allows the use of the compute nodes to be queried graphically.

• The Diagnostics window allows...
• ...function tests on individual nodes or node groups.
• ...performance tests to localise configuration errors.

• The Windows Deployment Service
• For updating or adding additional nodes
• The procedures are simple, similar to the familiar remote maintenance and remote installation procedures.
• Supports heterogeneous node architectures

• Job Scheduler
• The optimum tool for workload management
• Problem-free management of multiple queues, such as short or long-running jobs

• MS-MPI
• Allows fast data interchange via a high-speed interconnect
• Existing applications can quickly be ported into the parallel processing area

• Event tracing
• Allows time-synchronised event logs to be provided for easy tuning of applications
• Applications can be selectively pinned to individual CPU cores, making optimum use of current multi-core CPUs

For more information about Windows Compute Cluster Server (www.microsoft.com) >>

Linux operating systems

Advantages
For years, Linux has been the first choice for cluster operating systems.
Linux-based HPC clusters offer outstanding stability and reliability and have proven themselves thousands of times over.
The variety of commercial and open source software for cluster operation covers almost every conceivable type of application.
Management can be carried out entirely using adapted open source tools, such as Ganglia or OpenIPMI.

• Compatible with every code
  

A variety of free and commercial compilers are available for C, C++ and Fortran, along with a range of performance optimised libraries.

• Suitable for parallel and sequential calculations
  

Both open source and commercial MPI libraries are available for high-speed networks such as Myrinet or InfiniBand.
In conjunction with a compatible queuing system, such as the open software Sun Grid Engine, parallel jobs can be optimally scaled on the cluster.

Know-how

We are there for you

• From the conception and project design
• to installation, configuration
• and integration into your IT environment,
• installed turnkey on site and to your requirements.

Trust our years of experience in this business and let us help you optimise your structures.
Our experts will be happy to advise you.

Our promise

Our white papers on the subject of clustering

• High performance computing clusters (PDF)
• The Linux kernel 2.6 in cluster operation (PDF)
• The cluster configurator (PDF)

For further information about clustering and operating systems, speak to our experts.

A selection of our references

• transtec clusters are used for computing at the University of Stuttgart (PDF)
• transtec brings sport onto the net (PDF)