This video features Kevin Shinpaugh, director of IT and HPC at Virginia Bioinformatics Institute, discussing uses for HPC at VBI, Ray Bair, chief computational scientist at Argonne National Laboratory, on the Mira supercomputer, and Cycle Computing's Jason Stowe discussing the finalists for Cycle Computing's CycleCloud BigScience Challenge contest. The winners for the contest will be announced next year.

By Matthew Dublin

Researchers from the University of Tennessee and the Department of Energy's Oak Ridge National Laboratory are using supercomputer simulations to study a class of proteins involved in drug detoxification.

Jerome Baudry and Yinglong Miao, both of whom are jointly affiliated with ORNL and the University of Tennessee, have performed simulations using the Kraken supercomputer, a 1.17 petaflop Cray XT5 Linux system.

With Kraken, they were able to examine the motions of water molecules in a class of enzymes called P450s, which are sometimes responsible for processing a large fraction of drugs taken by humans.

The simulations modeled how water molecules move in and out of the protein's centrally located active site, enabling the team to clarify an apparent contradiction between experimental evidence and theory that had previously puzzled researchers. X-ray crystallography had displayed only six water molecules present in the active site, whereas experimental observations indicated a higher number of water molecules would be present in the enzyme.

"We simulated what happens in this enzyme over a time scale of 0.3 microseconds, which sounds very fast, but from a scientific point of view, it's a relatively long time," Baudry said. "A lot of things happen at this scale that had never been seen before. It's a computational tour de force to be able to follow that many water molecules for that long."

Water molecules (seen in red) move in and out of the active site (seen in blue) of a P450 enzyme. This class of enzymes is responsible for detoxifying a large fraction of drugs taken by humans.

The Kraken supercomputer:

Their research was published in the Biophysical Journal.

By Matthew Dublin

The international science grid this week has a post that describes how grid computing — not cloud computing — can be employed to aid proteomics research. The University of Nebraska-Lincoln's Robert Powers is using grid computing to increase the amount of data on the purpose of proteins with known structures. Powers has created the Comparison Protein Active Site Structures database, which contains all of the experimentally determined binding sites identified to date – roughly 36,000. The CPASS software uses that database to learn about proteins with unknown functions.

The CPASS software compares all possible orientations of the 3-D structure of the new ligand binding site with the entire database. But with only an in-house 16 CPU cluster at their disposal, each new site search took approximately 24 hours to complete. Over the last few years, the CPASS database had grown by some 40 percent.

To ramp things up a bit, Powers' group worked hand in hand with computing experts from the University of Nebraska-Lincoln's Holland Computing Center to help redesign the back end of database's interface so that users could submit jobs to the Open Science Grid via Glidein, a grid computing solution that is part of the Condor grid project. Condor is a grid paradigm that harnesses unused computer cycles in a network to build a computer grid.

Typical jobs on the CPASS database are now completed in about an hour.

By Matthew Dublin

Earlier this month at SC11, Clemson University's Walt Ligon presented an overview of OrangeFS, an open source file system for big data. OrangeFS is a highly scalable parallel file system the was originally developed from the Parallel Virtual File System, a file architecture designed to scale to petabytes of storage and data access rates at 100s of GB per second.

Ligon says that OrangeFS is relatively easy to install, use, and maintain because it's a "user level" file system which means that the interface is highly abstracted from the Linux kernel.

By Matthew Dublin

Gigaom has a post describing the lofty bandwidth dreams of billionaire physician and biotech entrepreneur Patrick Soon-Shiong, whose recently established Chan Soon-Shiong Institute for Advanced Health is aiming to build a super high-speed Internet connection for cancer research and treatment. The CSS Institute for Advanced Health is investing hundreds of millions of dollars into a nationally distributed computing network, with a focus on the National LambdaRail network /a>, which is comprised of over 12,000 miles of fiberoptics capable of 100 gigabits-per-second.

Cisco has also joined the efforts and will help build out the NLR's 100 Gbps capacity.

To achieve their goal, the institute has partnered with IO Data Centers, and will use IO's Phoenix data center serving as the primary hub for their operations. The data center houses a prototype sequencing appliance comprised of hundreds of high-end multicore Intel processors running on HP servers capable of running 50 gene-sequencing workloads a day.

By Matthew Dublin

Amazon Web Services has announced the release of a super-sized version of its cloud service called Cluster Compute Eight Extra Large, or 'CC2' for short. Each instance has two Intel Xeon processors, each with eight hardware cores, and comes with 60.5 GB of RAM and 3.37 TB of storage.

Currently, customers can expect to pay $2.40 per hour for one on-demand instance that comes loaded with either Linux or Windows Server 2008 R2. There is also a reserved instance offering priced at $4,146 plus $0.054 per hour through a one-year contract or users can bid for time for a lower price tag on the EC2 Spot Market, a sort of cloud computing auction block.

In order to show off their new extra larger offering, AWS put together a cluster on the fly that used 17,024 cores and managed to claim the number 42 spot on the latest version of the Top500 list released this week.