NERSC 2016: Extreme Computation and Data for Science 1

Overview By the year 2016, scientific computing will see a continued exponential increase in computational power. Simulations at or near exascale level will be conceivable in a growing number of scientific frontiers. HPC facilities in 2016 will be dealing with an exponential increase in experimental and simulation data. Scientific discovery will increasingly be based on the creation, maintenance, and analysis of exa-to zetabyte data repositories that need to be stored, accessed, analyzed, processed, shared, and understood. Analytics (the techniques and technology in data analysis, visualization, analytics, networking, and collaboration tools) will be essential in datarich scientific applications. As DOE's Keystone high performance computing and stroage facility with the mission to accelerate the pace of scientific discovery by providing high performance computing, information, data, and communications resources for all open applied and basic science and engineering sponsored by the DOE Office of Science, NERSC 2016 will provide unique resources and assistance to the open science community to enable community members to make effective use of exascale HPC resources and data. To meet the computational challenge, NERSC will field a series of early production systems that increase performance for science. To meet the data challenges, NERSC envisions hosting community data repositories with integrated information management and analytics tools and services, as well as new exascale storage technologies. To meet the ever increasing electrical power demands of ultra scale computers NERSC will explore new, tightly coupled hardware/software designs that emulate the achievements of the low-power embedded computing market to result in computers that meet the computational needs of scientific applications while showing a dramatic increase in energy efficiency. This paper describes NERSC's approach to achieving this vision for two if the three areas -Computing and Data. NERSC's approach to ultra efficient computing is documented in other papers 1 and not discussed further here. NERSC Computational Services The Keystone Facility The need for increases in computational resources between today and 2016 is well documented in the DOE Greenbook, 2 the SCaLeS Report, 3 and the E3 Report