Products, Deployments, and News in Supercomputing

[ High Performance Computing ] The atmosphere at SC11 was big data-focused and full of new tools. 


By Bio-IT World Staff 

January 10, 2012 | SEATTLE—November’s Supercomputing 11 (SC11) conference* in Seattle was full of news and product announcements across the industry. Here’s a sampling of what caught Bio•IT World’s ear.  

Top500 announced their newest list. The Top 10 order didn’t change, but the space between systems did. Japan’s “K Computer”, at the number one spot, completed a build out to make it four times as powerful as China’s Tianhe-1A system in the number two position.  

The Cornell Center for Advanced Computing received an HPC Innovation Excellence Award from the International Data Corporation for enabling hepatitis C virus research on a remote experimental MATLAB computing resource located at Cornell University. 

AccelerEyes launched ArrayFire, a freely-available GPU software library supporting CUDA and OpenCL devices. “ArrayFire is our best software yet and anyone considering GPU computing can benefit,” said James Malcolm, VP Engineering, in a press release. “It is fast, simple, GPU-vendor-neutral, full of functions, and free for most users.”  

Jason Stowe (of Cycle Computing) was working the Amazon Web Services booth. AWS theorized a bit on the cloud’s role in high performance computing in a blog post mid-week, and announced a new type of instance—CC2, or Cluster Compute Eight Extra Large—at the event. 

Cycle Computing offered $10,000 worth of free CycleCloud time plus technical support (at SC11 Amazon kicked in their own $2,500) to help a research group address an “un-askable question” as part of its BigScience Challenge. The finalists named at SC11 were Alan Aspuru-Guzik and Johannes Hachmann from the Harvard Clean Energy Project; Jesus Izaguirre from the University of Notre Dame; Soumya Ray from Harvard Medical School; Victor Ruotti from Morgridge Institute for Research; and Martin Steinegger from TU Munich ROSTLAB. 

Bright Computing is a company worth watching for their cluster management capabilities in the cloud and new status as an Amazon Web Services Solution Provider. Bright just announced comprehensive support for cloud bursting: with a few mouse clicks, customers can either create new clusters in the cloud or add cloud-based nodes to existing clusters that Bright manages like part of the local cluster.  

Convey Computer announced a partnership with Nimbix (link to PDF) to deliver Convey’s unique hybridcore architecture and bioinformatics personalities as a cloud-based HPC solution, and a doubling of its Graph500 performance (link to PDF).  

Cray won a University of Illinois’ National Center for Supercomputing Applications contract, and will provide a Cray XE6 system and a future upgrade of the recently-announced Cray XK6 supercomputer with GPU computing capability for National Science Foundation’s Blue Waters project. Once fully deployed, the system is expected to have a sustained performance of more than one petaflops. Cray also announced the Sonexion 1300 system, an integrated file system, software and storage product.  

Along with Cray, NVIDIA announced a parallel-programming standard—OpenACC—with the Portland Group (PGI) and CAPS enterprise. OpenACC allows parallel programmers to provide simple hints—“directives”—to the compiler, identifying which areas of code to accelerate, without requiring programmers to modify or adapt the underlying code itself.  

The DataDirect Networks booth was busy as the company released new storage offerings tied to its Storage Fusion product line. The company’s Storage Fusion Architecture was chosen by Munich-based Leibniz Supercomputing Center to power its SuperMUC HPC system.  

Globus Online celebrated its first birthday at SC11, and is doing effective large file transfer. The United Kingdom’s National Grid Service has adopted Globus Online as the preferred data movement method for its users.  

NETLIST introduced its new 32GB Virtual Dual Rank HyperCloud Planar-X RDIMM, enabling up to 768GB in a standard two-processor server, and demonstrated 1333 MT/s (mega transfers per second) on a standard server.  

Oak Ridge National Laboratory, Myricom, and Juniper Networks demonstrated “wide-area” 100 Gigabit Ethernet (100 GbE) between the ORNL and Juniper Networks booths. Juniper Networks and Myricom are providing ORNL the infrastructure required to connect to DOE’s Advanced Networking Initiative (ANI) 100G testbed, which links several national laboratories.  

ScaleMP pushed virtualization with thenewest version of its server virtualization for aggregation software platform, vSMP Foundation 4.0, and showcased the product line with partners HP, IBM, AMD, and AdvancedHPC.  

SGI hosted an impressive booth and introduced their Next-Generation ICE X Scale-Out Bladed HPC Cluster. The SGI-Cloudera partnership is also worth watching. SGI will now distribute Cloudera software pre-installed on SGI Hadoop Clusters.  

Platform Computing made a big splash, as they always do. Since being acquired by IBM in mid-October, Platform’s offerings could get even more interesting.  

Bull had a solid presence and launched a new generation of ultra-dense petascale supercomputers, though it’s not yet clear how much that will affect life sciences. •  

This article also appeared in the January 2012 issue of Bio-IT World magazine.