Update: Draft Technical Requirements were released to the vendor community December 17, 2012.
NERSC-8 Procurement Overview
The U.S. Department of Energy (DOE) Office of Science (SC) requires a high performance production computing system in the 2015/2016 timeframe to support the rapidly increasing computational demands of the entire spectrum of DOE SC computational research. The system needs to provide a significant upgrade in computational capabilities, with a target increase between 10-30 times the sustained performance over the NERSC-6 Hopper system.
In addition to increasing the computational capability available to DOE computational scientists, the system also needs to be a platform that will begin to transition DOE scientific applications to more energy-efficient, many-core architectures. This need is closely aligned with the US Department of Energy’s 2011 strategic plan, which states an imperative to continue to advance the frontiers of energy-efficient computing and supercomputing to enable greater computational capacity with lower energy needs. Energy-efficient computing is a cornerstone technology of what has been called exascale computing and represents the only way of continuing NERSC’s historic performance growth in response to science needs.
Read more from the NERSC-8 Mission Need Statement.
Partnership with Los Alamos and Sandia National Laboratories
The National Energy Research Scientific Computing (NERSC) Center and the Alliance for Computing at Extreme Scale (ACES), a collaboration between Los Alamos National Laboratory and Sandia National Laboratory are partnering to release a joint Request for Proposal (RFP) for two next generation systems, NERSC-8 and Trinity, to be delivered in the 2015/2016 time frame. The intention is to choose a single vendor to deliver two systems of similar technology.
Workload Analysis and Vendor Market Survey
At the start of each procurement NERSC surveys the HPC vendor market to understand current and future technologies and to gather input from vendors before the formal RFP is released. NERSC also shares high level goals, design targets and constraints in a Vendor Market Survey Presentation.
Also see a recent workload analysis presentation for information on the type of applications and algorithms that NERSC users run. The presentation also includes statistics on memory, threading, and library usage.
Assuring that real applications perform well on the NERSC-8 and Trinity platforms is key to the success of the systems. Because the full applications are large, often with millions of lines of code, are typically written using MPI only, and in some cases (LANL/SNL) are export controlled, NERSC and ACES have put together a suite of ‘mini-apps’ and micro-benchmarks for RFP evaluation and system acceptance.
See the benchmarks page