The US Department of Energy's National Energy Research Scientific Computing Center (NERSC), located at the Lawrence Berkeley National Laboratory, is to install a Cray XT5 supercomputer, which will be upgraded to a future-generation Cray supercomputer. When completed, the new Cray will deliver a peak performance of more than one petaflops (quadrillion mathematical calculations per second).
Like NERSC’s current 355-teraflops Cray XT4 system, nicknamed 'Franklin', the new supercomputing system will help advance open science research in climate modelling, biology, environmental sciences, combustion, materials science, chemistry, geosciences, fusion energy, astrophysics, nuclear and high-energy physics, and other disciplines, along with scientific visualisation of massive data sets.
'As NERSC is the primary supercomputing center for DOE’s Office of Science, making Cray's latest technology available to our users will accelerate innovation across a wide range of scientific disciplines, helping scientists tackle problems of vital importance to our nation's future,' said Michael Strayer, associate director of DOE's Office of Advanced Scientific Computing Research.
According to NERSC Director Kathy Yelick, Cray was awarded the contract based on several factors, including performance and energy efficiency on a set of application benchmarks that capture the challenging workload of the 3,000 NERSC users.
'Because we serve such a large and scientifically diverse user community, it's critical that our systems deliver the best performance while running real-world applications – especially as users scale their codes to run on tens of thousands of processor cores,' Yelick said.
The new Cray system will provide many pioneering features, including the ability for users to customise the operating system for their own codes and to schedule jobs and access their data without logging in to the supercomputer. Yelick adds: 'Cray’s new cooling system and interconnect network technology mesh well with our research efforts into energy efficient computing and programming models.'
Consisting of products and services, the multi-year contract is valued at more than $50m. The full system is expected to go into production in late 2010.
