The Weekly Top Five features the five biggest HPC stories of the week, condensed for your reading pleasure. This week, we cover Argonne's new 10-petaflop supercomputer, big rig aerodynamics, Austria's new 150-teraflop supercomputer, Whamcloud's partnership with Bull, and Bright Computing's deal with Dell.

10 Petaflop IBM Supercomputer on Order at Argonne Lab

IBM has been selected to build a 10-petaflop supercomputer for Argonne National Laboratory. The advanced computing machine, named "Mira," will boost researchers' ability to tackle advanced scientific challenges with real-world significance. Projects at the top of Mira's to-do list include designing ultra-efficient batteries for the next generation of electric cars, modeling detailed climate change scenarios and simulating the beginnings of the universe.

The IBM Blue Gene/Q supercomputer sports more than 750,000 cores and uses advanced chip designs and energy-efficient water cooling. With 10 petaflops of processing power, Mira is 20 times faster than Argonne's current supercomputer, Intrepid.

According to Rick Stevens, associate laboratory director for computing, environment and life sciences at Argonne National Laboratory, the new supercomputer "will help address the critical demand for complex modeling and simulation capabilities, which are essential to improving our economic prosperity and global competitiveness."

Argonne sees Mira as a stepping stone to exascale computing, machines that will be 100 times more powerful (than Mira). According to the release, "Mira will offer an opportunity for scientists to become more familiar with the capabilities an exascale machine will offer and the programming changes it will require. For example, scientists will have to scale their current computer codes to more than 750,000 individual computing cores, providing them preliminary experience on how scalability might be achieved on an exascale-class system with 100s of millions of cores."

When Mira comes online in 2012, scientists around the world, from industry, academia and government research labs, will be given the opportunity to apply for time through the DOE's Innovative and Novel Computational Impact on Theory and Experiment (INCITE) and the ASCR Leadership Computing Challenge (ALCC) programs.

Additional in-depth coverage of Mira is available here.

ORNL's Jaguar Supercomputer Helps Big Rigs Go Farther

Oak Ridge National Laboratory's Jaguar supercomputer is being used to improve the aerodynamics of long haul tractor trailers, and in the process save billions of gallons of fuel each year. South Carolina-based BMI Corp. partnered with ORNL researchers to develop the SmartTruck UnderTray System, "a set of integrated aerodynamic fairings that improve the aerodynamics of 18-wheeler (Class 8) long-haul trucks." After installation, the typical big rig can expect to achieve a fuel savings of between 7 and 12 percent.

With Jaguar, the time it took BMI to process its complex models was reduced from days to hours, and the system's advanced simulation capabilities obviated the need for time-consuming and costly physical prototypes. The time-savings meant the company was able to finish the project two years ahead of schedule, going from concept to production in 18 months instead of the 3 1/2 years they had originally anticipated.

The technology has significant financial and environmental implications. Mike Henderson, chief executive officer and founder of BMI, noted that if all 1.3 million Class 8 trucks in the US were outfitted with the basic UnderTray package, the average fuel economy would go from 6 mpg up to 6.5 mpg or more, saving the companies 1.5 billion gallons of diesel fuel and $5 billion in fuel costs each year. Better fuel efficiency means fewer CO2 emissions as well, 32.7 billion pounds or 16.4 million tons fewer.

Energy Secretary Steven Chu commented on the success of the collaboration, which was made possible through ORNL's Industrial High-Performance Computing Partnerships Program:

"The Department of Energy's supercomputers provide an enormous competitive advantage for the United States. This is a great example of how investments in innovation can help lead the way to new jobs, new ways of cutting our carbon emissions and new opportunities for America to succeed in the global marketplace."

Austrian Universities to Share New Supercomputer

MEGWARE, a leading German IT company, has been selected to build Austria's fastest supercomputer at a cost of EUR 4.2 million. The computer will be used by researchers at the Vienna University of Technology, the University of Vienna and the University for Soil Management. A Europe-wide tender process called for a "first-class energy efficiency of the entire system and an enormously high raw data throughput." The universities chose MEGWARE based on the merits of the company's design and its comittment to using energy-efficient technology.

The 150-teraflop supercomputer, which will be known as "Vienna Scientific Cluster 2" (VSC-2), will provide Vienna scientists with the power to make important scientific advances. VSC-2 includes more than 1,300 MAGWARE-designed servers, each of which is equipped with two AMD Opteron Magny Cours 6132HE processors, providing the machine with 21,000-plus processsor cores.

The new supercomputer will have a signicant lead on its predecessor, VSC. With 150-teraflops, the new system will be five times more powerful. VSC-2 is also more energy-efficient, employing state-of-the-art water cooling technology to reduce power demands.

MEGWARE representative Jörg Heydemüller expressed satisfaction with the project: "The fact that this significant IT order was awarded to a company from Chemnitz is very important for the location. It shows again the historic links existing between the Free State of Saxony and the microelectronic industry, as the project is implemented with the exclusive use of processors made by AMD."

Bull, Whamcloud Share Lustre Dreams

Whamcloud's busy Lustre devotees have joined with European IT vendor Bull to advance the Lustre cause. The duo have formed a strategic partnership towards developing open-source file system and contributing the improvements back to the community. According to Whamcloud, the company's shared goal is to prepare Lustre to become a file system of choice for exascale systems.

Eric Monchalin, HPC software director at Bull, comments on the company's involvement, "Bull has been integrating Lustre in its Extreme Computing solutions for many years, and our experts are regular contributors to the evolution of the Lustre file system. Whamcloud has gathered some of the top talents of the Lustre ecosystem. This combination of Bull and Whamcloud skills will allow our customers to continue to get the best performance out of Lustre, while being assured that the file system they chose is fully maintained and supported."

Whamcloud CEO Brent Gorda expressed excitement about the opportunity to partner with Bull, and looks forward to "driving IO and storage development worldwide in an open-source, hardware-agnostic environment."

Bright Cluster Joins Dell's HPC Family

Bright Computing announced this week that its cluster management software, Bright Cluster Manager, would now be an option on Dell's High Performance Computing Clusters (HPCC). There are already more than 50 Dell installations around the world using Bright Cluster Manager, including three TOP500 sites. Other sites include a number of universities, businesses, and government research centers.

CD-adapco uses the combined solution to run its engineering applications, and IT director Philip Jones is pleased with the product's performance: "Bright Cluster Manager is a comprehensive cluster management solution that provides all the functionality that we need here at CD-adapco. Our key applications -- STAR-CCM+ and STAR-CD -- were easy to install and run well on the cluster. Bright Cluster Manager has many features that make it easy for us to manage the cluster and allow us to focus on running our CFD and CAE applications. For example, the image based provisioning makes it very easy to tailor software images and propagate changes to the compute nodes."