High Performance Computing (HPC) looks at the challenges of building and using scalable systems. Advances in device technology now allows the manufacture of single chips capable of billions of operations per second. However, the demands of scientific simulations of extremely small system (e.g. nanotechnology) or extremely large systems (e.g. global climate) now demand computational power orders of magnitude higher than this. By combining the many of the world's fastest processors together, it is now possible to design and build computer systems capable of trillions or quadrillions of operations. These HPC systems are enormously complex, and creating the environments to use them correctly remains a challenge.

NSF Petascale Computing: World-Class Science Through World Leadership in High Performance Computing

This project will construct the world's largest open computing system, dedicated to providing enormous computational power to researchers supported by the National Science Foundation. The new centerpiece of the Teragrid network, ASU will be involved in the design and operation of the machine as well as supporting scientists throughout the western United States as they scale their computing to this level. Researcher: Dan Stanzine. Collaborators: University of Texas, Cornell University

Programming from One Thousand to One Hundred Thousand Threads

Upcoming generations of processors will pack tens of processor “cores” onto a single silicon chip. HPC systems composed of these chips will in 3-4 years be capable of executing hundreds of thousands of simultaneous “threads” of a single program. In this project, the infrastructure required for programming at this scale is being created, including debuggers, compiler extensions, and run-time support. Researcher: Dan Stanzione. Collaborators: Intel Corporation, University of New Mexico

Thermal Aware Dynamic Resource Management for Datacenters

Large scale computing centers consume tremendous amounts of electricity, both to power the systems, and to remove the heat generated by these systems. Managing the temperature in data centers is important both to improve the reliability and lifespan of hardware and to optimally use power resources. In this project, we use models of data centers from computational fluid dynamics along with sensor data to integrate thermal information with resource management software to shape data center load for the best possible thermal behaviour. Researcher: Sandeep Gupta. Collaborator: Intel Corporation

Simulation Tools for Photonics Devices

Researcher: Dan Stanzione. Collaborators: Stephen Goodnick, Electrical Engineering

Dynamic Virtual Clustering

Researcher: Dan Stanzione. Collaborators: Oak Ridge National Laboratories, Cluster Resources Inc.

CyberPark

Researcher: Dan Stanzione. Collaborators: Mechanical and Aerospace Engineering (Kyle Squires, Bob Peck)