High-Performance Computing
For about a quarter century, SDSC has operated a wide range of allocated computational resources, and continues to make leading-edge resources available to the UC San Diego and broader UC community, in addition to others in academia and industry. To help researchers run on today's high-performance systems, SDSC experts guide users through a smooth transition from campus clusters to the center's high-performance computing environment, reducing time to results and taking science to the next level.
SDSC is home to the Triton Resource, a cost-effective and accessible high-performance computing system primarily for UCSD and UC researchers. Triton comprises a medium-sized (256-node) cluster that can tackle many research computing tasks, and a 28-node "large memory" cluster specifically for data-intensive computing projects. The system also includes a high-performance parallel file system for staging large data sets and access to high-bandwidth research networks such as CENIC. The Triton Affiliates and Partners Program (TAPP) offers various mechanisms for accessing the resource. It is also available to other universities, research institutes, government agencies, and industry on a space-available basis.
SDSC is a founding partner of the XSEDE, which integrates high-performance computers, data services and expertise across, and linked to, some of the nation's most powerful supercomputer centers, creating the world's most comprehensive distributed cyberinfrastructure for open scientific research. SDSC has begun to offer allocations to Trestles, with a theoretical peak speed of 100 TF (teraflops), containing 324 nodes, each with four 8-core Magny-Cours processors and 64 GB of memory. Trestles also incorporates 120 GB of local flash storage on every node. This resource is targeted towards XSEDE users requiring 1024 cores or fewer and long running jobs (up to multiple weeks).
SDSC also offers allocations to Dash, a unique, large-memory resource featuring solid-state disk drives that significantly reduce latency times for faster computations across a wide range of data–intensive research, along with vSMP and Flash Memory. Dash's unique design aggregates memory across nodes, making much larger memory capacity available to virtual nodes that are simpler for codes to access. Dash is a prototype of Gordon, a much larger cluster targeted at data-intensive computing that will be installed in the second half of 2011.
Also, SDSC is a resource partner in the FutureGrid project (also part of XSEDE), which provides a high-performance grid test bed that will allow scientists to collaboratively develop and test innovative approaches to parallel, grid, and cloud computing. SDSC hosts the Sierra resource, which has 7 TF of compute power, 96 terabytes of raw storage, and is connected to the FutureGrid network via a 10GB link. Sierra provides cloud environments (such as Eucalyptus and Nimbus) as well as supports HPC applications running on the "bare metal." Users can run stand-alone or distributed experiments on Sierra and other FutureGrid machines.University of California researchers who formerly had access to SDSC's OnDemand cluster can now utilize the Triton Resource for their computational needs. Triton features many of the same applications, including Star-P software, which provides a link between MATLAB tools on the client side and the high-performance parallel resource at SDSC.
UC San Diego departments and research projects across campus rely on the expertise and infrastructure of SDSC to enhance their capabilities and efficiency. By partnering with SDSC for cyberinfrastructure services, campus partners are able to leverage SDSC's economies of scale, high-quality services, and thorough knowledge to further their own research and productivity.
For more information about SDSC's services, contact services@sdsc.edu.
