Linux Clusters

Red Hat users benefit from a reliable software ecosystem, including:

• Certified applications from independent software vendors (ISVs)
• Excellent performance, flexibility, security, scalability and affordability
• Open source projects, such as Fedora™, and rigorously tested and mature technologies
• Dependable application interfaces and product support
• Homogeneous environment for seamless interoperability from desktop to exaFLOP — and with existing UNIX® and Microsoft® Windows® deployments

Middleware and Development Tools

Dell offers flexibility with the choice of standards-based message-passing libraries, including MPICH, MPICH-GM and MVAPICH, that support available interconnect options. Dell and its partners also offer a broad and comprehensive selection of development toolkits that include compilers and math libraries, system and development tools.

Provisioning and System Management Software and Tools

When deploying Linux-based clusters, whether small or large scale, it is important to factor in the right combination of provisioning and system management software and tools, to achieve the greatest performance and to maximize the system's overall capabilities.

Platform Computing — Platform Cluster Manager

The Platform Cluster Manager is designed to automate system initialization and software deployment, with a complete range of tools, including: device drivers, installers, cluster management tools, resource and application monitoring, interconnect support, Platform Lava — a powerful open source job scheduler — and integration with Load Sharing Facility (LSF).

The Platform Cluster Manager — Dell Edition — features a unique kit architecture designed to automatically install and configure software components, simplifying software installation and maintenance in cluster environments throughout the system life cycle. Kit features include:

• Cacti — reporting tools for gathering and graphing node performance metrics
• Ganglia — for resource monitoring
• HPC — collection of tools, libraries and utilities
• Lava — powerful workload scheduler compatible with Platform LSF
• LSF — a gold standard in commercial workload schedulers
• Nagios — for host, service and network monitoring
• NTOP — to monitor network bandwidth and analyze traffic
• OFED — collection of drivers to support servers, storage and facilities such as IP over InfiniBand (IPoIB)
• Dell — utilities, drivers and open manage agents to streamline installation and management

Clustercorp Rocks+ — Simplify the Management of HPC Clusters

The developers of Rocks started with the simple belief that it shouldn't take an expert or full-time administrator to build and manage an HPC cluster. Ten years of development and 10,000+ clusters later, Rocks has taken a clear leadership role in the HPC industry by unequivocally succeeding in the goal to 'make clusters easy.'

Clustercorp Rocks+ compiles the complete HPC software stack, including Dell-specific performance and reliability components, onto a single DVD. Clustercorp Rocks+ is backed by Clustercorp's development and support team, which includes the founders of the massive open-source Rocks community.

Clustercorp Rocks+Rolls: Everything You Want, Nothing You Don't Want

Clustercorp Rocks+ leverages Rolls, which empowers users with the option to build a variety of system types. A sampling of solutions includes:

• Compute Cluster (HPC Roll)
• Visualization Cluster (Viz Roll)
• Virtual Cluster (Xen Roll)
• Hybrid Cluster (Dual-boot)

Or add the components required for advanced infrastructures:

• Red Hat Enterprise Linux (plugged in as a Roll)
• Intel® Roll (Intel cluster-ready)
• OFED Roll (InfiniBand™ support)
• Moab Roll (Moab cluster suite) — green computing!
• TotalView Roll (parallel debugger)

For a complete list of Rolls and additional product details, visit www.Clustercorp.com (This product includes software developed by the Rocks Cluster Group at the San Diego Supercomputer Center at the University of California, San Diego and its contributors).

Interconnects

Dell integrates a breadth of interconnect options to facilitate node-to-node communication, data sharing and synchronization for a broad range of applications. Dell offers a variety of interconnect options to provide flexibility in balancing price, latency and bandwidth requirements, including:

• Gigabit Ethernet
• InfiniBand
• Myrinet®

Shared Disk File System

Shared disk, or clustered, file systems are important in Linux-based cluster architectures and can include either 'distributed' (across all nodes) or 'centralized' (metadata server) solutions. Designed to provide access to data and storage across shared nodes, regardless of data location, clustered file systems also allow for simultaneous read/write access, striping of data and metadata, and most importantly, they prevent data corruption.

Panasas ActiveStor Parallel Storage

Panasas® provides a comprehensive family of parallel storage solutions for primary and secondary storage applications. The ActiveStor Parallel Storage Cluster is based on PanFS™ Parallel File System and is designed to deliver exceptional performance, scalability and manageability.

• Up to 20TB of disk storage
• 600MB/s throughput
• Linear scalability of capacity and throughput
• Options include high-performance XC storage blades, ActiveImage snapshots and ActiveGuard high availability software

The advanced ActivStor solution turns files into smart data objects and dynamically distributes the data transfer operations to enable parallel data paths between nodes, improving performance and minimizing capacity bottlenecks. The ActiveStor suite of products is designed to manage petabytes (PB) of data capacity and growth, all within a single, easily managed name-space.