HPC Solution Stack Featuring ClusterWare

Applications

Scyld ClusterWare is fully compatible with both RedHat Enterprise Linux and CentOS, supporting a huge variety of applications from all HPC disciplines such as Mechanical Computer-Aided Engineering (MCAE), Life Sciences, Computational Fluid Dynamics, Financial Services, Energy Services and Electronic Design Automation (EDA).

Resource Managers and Schedulers

Scyld ClusterWare includes the resource management/scheduling solution TORQUE and is compatible with many other popular schedulers. TORQUE is an open source resource manager/scheduler providing control over batch jobs. Originally based on the PBS project, significant advances in the areas of scalability, fault tolerance, and feature extensions have been incorporated. The version of Torque shipping with Scyld ClusterWare is fully supported by Penguin Computing. Scyld ClusterWare also supports other resource managers/schedulers e.g. Sun Grid Engine, Altair PBS Pro, Platform LSF.

For customers with sophisticated scheduling policies Penguin Computing offers the workload management solution Moab Cluster Suite

Tools and Libraries

For the development of distributed MPI based applications Scyld ClusterWare includes optimized versions of the MPI libraries MPICH, MVAPICH, and OpenMPI, built for common compilers. These modified MPI library implementations take advantage of Scyld ClusterWare's efficient process creation and management. Distributed applications linked against Scyld ClusterWare's MPI libraries can be started without depending on remote execution services. They are launched on the least loaded systems in a consistent environment that includes all environment variables set on the master node at startup time.

Scyld ClusterWare supports the GNU development tools (e.g. gcc, g77, gdb, ddd) that are included in the underlying Linux distribution. Compilers from Intel, Pathscale and the Portland Group are available from Penguin Computing. For the optimization of applications Intel's VTune™ and Intel's Trace Analyzer are offered.

Scalable Filesystems

The parallel high performance file systems Lustre from Sun Microsystems and PanFS from Panasas are supported on Scyld ClusterWare. Drivers are available from Penguin on request.

Monitoring

Scyld ClusterWare's monitoring tools include an optimized version of Ganglia, command-line tools and libraries including BeoStatus, and a new web-based management interface, Scyld Insight. Ganglia allows users to view live or historical statistics (such as CPU load averages or network utilization) for all machines that are being monitored. True to the philosophy of minimizing overhead on compute nodes Scyld ClusterWare's Ganglia implementation does not utilize local agents on compute nodes. With Scyld ClusterWare resource information for all compute nodes is already aggregated on the master node. Scyld ClusterWare's Ganglia retrieves the aggregated resource information from the master.

WebServices API

Scyld ClusterWare ships with the web service ‘beoweb’. Based on this service, the tool suite PODtools that also ships with Scyld ClusterWare allows for remote submission, monitoring and control of compute jobs from any Linux system. Future versions of beoweb will provide documented API’s that allow for integrating job and cluster management features with existing web applications.

45-day Evaluation

Give Scyld ClusterWare a try for 45 days at no cost. To register, simply complete the online sign-up form.