Bright Cluster Manager 5.2 shipping with support for CUDA 4, ScaleMP, SLURM, LSF and enhanced multi-cluster management
20 Jun 2011 San Jose -
Bright Computing, an expert in cluster management software, is now shipping Bright Cluster Manager release 5.2. This latest version of the product includes multiple new features and a broad range of expanded capabilities for managing HPC clusters. These capabilities include enhanced multi-cluster support, full CUDA 4.0 support for NVIDIA GPUs, tightly integrated support for on-demand SMP, additional options for integrated workload management, support for the latest releases of Linux distributions, cluster reliability improvement and an enhanced web-based user portal.
Bright Cluster Manager continues to lead the industry in both removing complexity from cluster management and in providing advanced capabilities for improving operating efficiencies. For example, Bright Computing has enhanced its unique multi-cluster support beyond its ability to easily monitor and manage multiple clusters from a single GUI or CLI. Bright Cluster Manager 5.2 enables any cluster settings, including monitoring rules, health checks, user information and access roles, to be exported from one cluster to one or more other clusters. This new feature greatly reduces the complexity of managing multiple clusters and delivers operational efficiencies in geographically distributed data centres.
Other enhancements include:
- Full CUDA 4.0 support, with the ability to monitor the full extended range of GPU metrics and visualize them within the Bright Cluster Manager GUI. System administrators benefit from maximum visibility and manageability of GPUs, quickly and easily, as part of the overall system management, radically reducing the time and effort required to get the most productivity from the cluster.
- The addition of SLURM and LSF as integrated workload management options, extending the previous choices of PBS Professional, Torque/Moab, Torque/Maui, Grid Engine. This integration brings substantial benefits, including automated workload manager installation and configuration, and robust workload manager failover. This integration also enables pre-job health checking, the only way to prevent the job-crashing "black hole node syndrome".
- Expansion of Bright's cluster health management framework with an extended range of user-selectable early-warning and intervention capabilities for any metric within the cluster.
- A customizable web-based, user portal separate from the admin GUI. This portal provides users with a dashboard-style overview of the status of the cluster and of their own usage of the cluster. Users can easily visualize a wide range of cluster metrics using configurable graphs. Furthermore, users have a clear overview of workload management information such as job queues.
- A Python interface for the cluster management SOAP API, making all cluster management functionality, as well as all cluster management configuration and monitoring data, accessible in Python scripts. C++ and PHP interfaces were already available. The interface to the SOAP API greatly improves the ease and ability to script and automate operations, as well as inter-operate with third-party software.
- Seamless integration of ScaleMPs vSMP Foundation into Bright Cluster Manager. This capability enables a user to dynamically aggregate multiple cluster nodes into a single virtual SMP node with just a few mouse clicks. This virtual SMP node is then provisioned, monitored and managed as part of the overall cluster. This capability not only delivers incredible flexibility to clusters needed to drive occasional applications optimized for SMP, it also offers significant hardware efficiencies and cost savings.
- Support for the latest Linux distributions, including Red Hat Enterprise Linux 6, CentOS 6 (coming), Scientific Linux 6 and SUSE Linux Enterprise 11 SP1. Older Linux distributions, such as releases 4 and 5 are also available and supported.
- The inclusion of several dozen customer-driven enhancements, spanning head node failover, provisioning, user administration, cluster monitoring and management. These enhancements, including several suggested by TOP500 system administrators, improve cluster performance, reliability and manageability.
Bright Cluster Manager 5.2 is the latest milestone in our continued leadership of cluster management innovation, anticipating and delivering on the evolving needs of our customers and partners", stated Martijn de Vries, CTO of Bright Computing. "We've enjoyed an enthusiastic response from our customers who are using our new release. Our partners have already won some major Bright deals as a direct result of the new features, including the recently announced TOP500 system at the University of Bordeaux."