Phoenix is a complete cluster management suite developed by Atipa to provide easy day-to-day management and operations of our clusters. Phoenix is complimentary with 3-years of standard support for all new cluster purchases.

• Fully compatible with Red Hat Enterprise Linux, CentOS, and Scientific Linux. Up-to-date and installed with the latest patches.

• Single, central point of management. All configuration, user accounts, and health monitoring are performed from the head node.

• Quick, intuitive node provisioning. 100% menu-driven installation process facilitates easy deployment of new nodes.

• Support for multiple node types. Pre-defined node types include standard compute nodes, large memory nodes, accelerator nodes, storage nodes, and Lustre* server nodes. Additional custom node types can be added if needed.

• Easily modified node software stack. Each node type has its own configuration files and installation scripts to allow fine-grained customization of the software stack for different node types.

• User account management. A single command creates new user accounts and access to all          cluster nodes.

• Parallel command execution. Simultaneously copy files to or execute commands on all nodes or subgroups of nodes in the cluster.

• Out-of-band system management. Remote power on/off and KVM-over-LAN via dedicated IPMI       2.0 network.

• Web and command-line based health monitoring. Web-based monitoring is available via ganglia, including custom plug-ins for CPU temperature and fan speed monitoring. Command-line checks are available via a single clusterQC command, customized for each cluster and node type to check hardware and software conformity.

• Customized to each system for optimal user experience. Custom clusters need custom software. Modifications or additions needed for specialty items will be included into a site-specific Phoenix stack as needed.

• Pre-bundled job scheduler, parallel libraries, and performance libraries. Open source job scheduler (torque/maui or Slurm), parallel libraries (Open MPI), Intel   Math Kernel Library (MKL) included. Others available on request.

• Pre-configured environment modules. Easily manage different versions of applications and libraries without conflicts. End-users can dynamically modify their environment with simple “module load” and “module unload” commands.

• Optional commercial schedulers, compilers and 3rd-party software. Advanced scheduling with Moab HPC Suite, transparent code optimization with Intel   or PGI compilers, improved software development with 3rd-party development tools.

• Performance tuning. All clusters are tuned for top performance for a wide range of applications.

• Certified Intel   Cluster Ready. Conforms with Intel   ’s Scalable System Framework cluster architecture.

®

®

®

®