INDUSTRY

TORONTO, CANADA - Platform Computing Inc. announced Platform HPC for IBM, a new Grid-enabled job management solution for IBM high-performance computing (HPC) clusters. Platform HPC for IBM is designed specifically for midrange to high-end distributed HPC clusters running on IBM eServer xSeries and pSeries systems. This out-of-the box offering maximizes the performance and capability of parallel workloads across AIX and Linux servers, enabling government, research, life sciences and manufacturing users complete compute-intensive tasks faster and more reliably at reduced costs. "IBM HPC users need an easy way to optimize usage of their servers without having to compete for available resources. The trend toward server consolidation has only intensified this challenge, making it increasingly difficult for multiple users and groups across an enterprise to effectively share the HPC resources," said Paul Hill, vice president, marketing and business development, Platform. "With Platform HPC for IBM, IBM HPC users can maximize their application performance and achieve greater computational capacity to meet their grand challenges sooner, deliver higher quality products faster, and reduce their time to discovery." Platform HPC for IBM provides a set of scheduling policies rich enough to prioritize the workload and arbitrate usage of the heterogeneous hardware within clusters of both AIX and Linux servers, and across multiple sites. As a Grid-enabled enterprise computing solution, Platform HPC for IBM enables groups and individual users to share globally distributed HPC resources, using sophisticated job scheduling and sharing policies across all HPC assets. "A complete infrastructure requires many technology components, which is why IBM is working closely with key business partners like Platform to build Grid solutions for its customers," said Dan Powers, vice president, Grid computing strategy for IBM. "IBM offers the most extensive experience in the HPC market while Platform offers extensive experience in providing job scheduling and workload management software for a Grid environment." Key features of Platform HPC for IBM include: * Advanced HPC Scheduling Policies: High-speed interconnect aware scheduling combined with policy-based pre-emption, advance reservation, project-based fairshare, and custom scheduling ensure resources are allocated to the right users and the right projects at the right time, while maximizing throughput of mission critical workload. * Heterogeneous Cluster Support: Integrates with mixed 32-bit and 64-bit AIX and Linux clusters for greater application flexibility and reduced resource administration. It also provides a single point of control for submitting and monitoring workload across IBM and non-IBM hosts for optimal hardware utilization. * Commercial Grade Reliability and Availability: Unprecedented levels of job reliability can be achieved through a combination of automatic scheduler fail-over, duplicate job event logging, OS-level checkpoints and restarts, and automatic job migration and re-queuing. * Parallel Application Integration and Support: Out-of-the-box integration with parallel applications, along with accounting for parallel HPC applications, enhanced job control and run-time resource limit support and collection, ensure that complex parallel applications can be deployed, tracked and managed as easily as sequential applications. * Open, Scalable, Secure Architecture: Job management is further extended through application level access to resources. Modular scheduler architecture, system authentication, support for plug-in APIs and external resource collectors ensure secure access to and site specific support for over 500,000 active HPC jobs and more than 2,000 multiple processor hosts. Platform provides High Performance Computing (HPC) solutions to educational institutions, government and supercomputing laboratories, automotive firms, life sciences companies and oil and gas customers worldwide, including the U.S. Department of Defense, NASA, Oxford University, UCLA, the University of Glasgow, the European Laboratory for Particle Physics (CERN), and the National Scaleable Cluster Project (NSCP) which includes the University of Pennsylvania, the University of Illinois at Chicago, the University of Maryland at College Park, Osaka University and the University of Tokyo.