Here is a collection of highlights from this week’s news stream as reported by HPCwire.
Internet2 Announces Winners of Fifth Annual IDEA Awards
Sandia’s Red Mesa Tasked with Finding Renewable Resources
Earlier this month, the Red Sky and Red Mesa supercomputers had their grand opening at Sandia National Laboratories in New Mexico with representatives from Sun/Oracle and Intel corporations taking part. This week, Sandia introduced more details about the Red Mesa system, which is being used for some of the world’s most challenging energy challenges.
The result of a collaboration between Sandia and the National Renewable Energy Laboratory (NREL), Red Mesa is being used to develop renewable energy resources, such as the power that can be generated from the sun and wind.
According the announcement:
The work for the first time brings defense-scale computing to bear on alternative energy projects that otherwise could take months or even years to complete if researchers had to rely on more limited computing resources or on physical testing.
At 180-teraflops, the Red Mesa supercomputer is the smaller counterpart to the Red Sky system. The two systems were constructed using the same architecture and components. Sandia engineers temporarily aggregated the two machines prior to the November 2009 TOP500 list, and together they reached a combined speed of 500 teraflops, earning the unified system a number 10 spot.
Not only is Red Mesa being used for so-called green energy projects, the machine itself is an example of green computing technology, with dramatically improved energy efficiency compared to other systems of its ilk. Its cooling mechanism, dubbed Glacier Door, uses an innovative cooling technique. There is a door capping each cabinet that keeps the cooling mechanism close the the heat source, with the novel result being that the air leaving the cabinets is slightly cooler than when it came in. The passive design does not require additional fans or electrical power to circulate air, further cutting down on energy use.
In addition to the improved cooling system, Red Mesa is outfitted with a power distributing system that significantly reduces power leakage and was designed for easier access to electrical wiring. The InfiniBand-based network fabric uses optical interconnect cables exclusively.
Sandia manager John Zepper said the upgrades “will save millions of dollars over the life of this machine.”
Oracle, the project vendor, intends to include the same types of innovations with its smaller systems.
Popular Workshops Announced
The Cornell Center for Advanced Computing has announced a National Science Foundation-sponsored workshop on sustainable funding for HPC centers will be available via the Web conferencing site WebEx. The “Workshop on Sustainable Funding and Business Models for HPC Centers” takes place May 3-5 at Cornell University. Participants include over 90 high performance computing center directors, CIOs, vice provosts for research, and CTOs.
The workshop has received so much interest that the onsite event is already at capacity; that’s where WebEx comes in. The main presentations and audio will be available via the Web and all you need is Internet access (and speakers or headphones). Questions can be relayed via the WebEx chat function.
More information, inluding a link to register for the WebEx sessions, is available at http://www.cac.cornell.edu/SRCC.
In an unrelated release, EM Photonics, developer of GPU-accelerated computing solutions, announced that it is offering CUDA training classes to help participants learn how to develop specific applications on the NVIDIA CUDA platform. The hands-on GPU training program covers the essentials of programming on the CUDA platform as well as more advanced techniques. Classes are customized to the participants’ skill level, and onsite classes allow trainees to apply what they are learning to their own tools and environment.
The standard training program lasts two days, but can be longer to cover more in-depth instruction. More information is available at www.emphotonics.com/services/cuda-training.