Cluster Computing:

INFINICON, SCALI ANNOUNCE APP-OPTIMIZED CLUSTER ENVIRONMENT

InfiniCon Systems and Scali announcedthe Application-optimized Cluster Environment (ACE), which provides breakthrough performance levels for commercial High Performance Computing applications such as Fluent, Eclipse, Magma, LS-Dyna, and others. In addition, the companies are offering an industry-first performance guarantee, assuring a minimum of 25 percent throughput improvement over standard cluster interconnect configurations.

"Clustered solutions for the technical computing market more than doubled sales from 2002 to 2003. Clusters are attractive to end users due to their highly competitive price/performance ratios, which are driven by leveraging industry standard processors, interconnects, and middleware," stated Earl Joseph, Vice President of Research, High Performance Computing, for analyst firm IDC. "Providing guaranteed levels of performance improvement on any application is an innovative marketing approach that will likely catch the attention of both buyers and competitors."

Record Performance Levels

New levels of performance for commercial HPC applications were achieved leveraging the combination of InfiniCon’s 10Gb InfiniBand-based systems, in conjunction with Scali’s MPI Connect middleware environment, to ensure a high performance, low latency network upon which industry-standard servers can be clustered. When used together in an Itanium" or Opteron" environment, these comprise an Application-optimized Cluster Environment (ACE) for which the companies performed benchmarks using commercial industry applications. Results for 16-node clusters showed performance results for Fluent’s FL5M3 that vaulted from 262 jobs/24hours to 1153 jobs/24hours – a 338 percent throughput increase. Using MagmaSoft’s MAGMAfill application, the throughput was even greater, at 543 percent measured with the same jobs/24hour metric; in tests using Eclipse E100, the increase was a respectable 56 percent. These results come on the heels of other recent performance tests, which show the ability to scale InfiniCon-based clusters to up to 144 nodes while retaining an industry- leading 80 percent or better CPU scaling efficiency.

Industry’s First Performance Guarantee

The ACE performance guarantee, in which InfiniCon and Scali guarantee at least 25 percent performance improvement for ACE configurations over standard, gigabit Ethernet-based clusters, is unique in the industry. "During our seven- year history, we have never felt as comfortable in offering a ‘guarantee’ as we do today," stated Hakon Bugge, CTO for Scali. "In coupling the power of InfiniCon’s InfiniBand-based offerings with our advanced MPI software, we are supremely confident in the performance and value offered to HPC customers."