High performance computing has become increasingly important to both theoretical and experimental investigations across the sciences. Theoreticians require advanced computation and huge volumes of storage to increase understanding of the behavior of systems ranging from the sub-atomic to stars, galaxies and beyond, while experimentalists increasingly rely on advanced storage systems to accommodate and make sense of the mountains of data ingested from an ever growing array of specialized instrumentation.
SGI large shared compute resource solutions for multi-disciplinary research allow scientists to maximize application performance and research flexibility. They allow single users, research teams, departments, institutions and cross-institution entities to:
- Develop breakthrough solutions with advanced computing technology tailored to workload and budget
- Dynamically adjust the mix of applications to changes in research priorities
- Address huge amounts of data and move it quickly into and out of the system
- Focus on science, not system and data management
- Maximize ROI
- Easily manage large, shared compute resources
The SGI product and service solution set for Multi-Disciplinary Research is peerless in its comprehensiveness and functionality:
- SGI® UV systems run Linux® and are based on Intel® Xeon® processors. They enable compute, memory and I/O scaling within a single shared memory environment. These systems are unsurpassed in their memory and I/O capability, and deliver exceptional Message Passing Interface (MPI) application performance because of their low-latency and high-bandwidth SGI® NUMAlink® architecture. SGI UV systems implement globally shared memory so that any CPU can directly address any byte of memory or I/O device.
- The SGI® ICE integrated blade cluster system was designed for today's data intensive problems. This innovative platform raises the efficiency bar, easily scaling to meet virtually any processing requirements without compromising ease of use, manageability or price/performance. SGI ICE systems maximize flexibility since they can run single compute-intensive, memory-intensive or I/O-intensive applications that take advantage of the full capabilities of the system, or be used to run multiple medium to small simulations in throughput mode.
- Rackable® systems can be combined with SGI ICE systems in cluster environments to maximize the performance of multiple application workloads. SGI ICE and Rackable servers and clusters can be integrated into a single system management and data sharing environment that minimizes system management overhead. By freeing scientific staff to focus on science instead of system and data management, productivity and innovation increase.
- SGI storage systems, designed for data-intensive environments, enable fast delivery of information to the compute nodes, so researchers can minimize their time to insight. Our active archive, RAID, MAID and JBOD solutions, combined with a rich ecosystem of data management software solutions, are cost-effective, ensuring the feasibility of adding storage as a complete solution.
- SGI system management software supercharges the capabilities of the operating system to maximize performance on SGI computer systems, while enabling researchers to run jobs when resources become available, based upon an array of parameters that datacenter managers set up.
Image courtesy of SARA