|
InfiniCon's InfinIO family of InfiniBand-based solutions which provides up to a 30Gbps, low-latency infrastructure for building ultra-scalable computing fabrics was integrated into a number of real-world application environments.
InfiniCon's InfinIO technology was featured in a demonstration of Hewlett-Packard's Linux-based, ProLiant DL140 servers clustered together to run application sets that illustrate the effects of computational fluid dynamics modelling software from Fluent Inc., as applied to the design and testing of jet aircraft. In this environment, multiple commodity servers process a single large computational job as a single image, offering higher levels of performance and availability at lower cost than is traditionally available from proprietary systems.
Leveraging the power of InfiniBand architecture, the InfiniCon fabric comprised of high-performance, low latency switching systems and host channel adapters provides a 10Gbps network to allow the clustered nodes to work in
tandem. In addition, the performance-optimized implementation of Scali MPI-Connect middleware from Scali ensures that high performance is combined with high-availability features that protect against faults in the network, interconnect, or routing tables.
InfiniCon's InfinIO 7000 Shared I/O and Clustering System was being showcased as part of Intel's HPC Express demonstration, which features the next-generation industry standard PCI Express technology implemented into a 72-node, InfiniBand-based high performance computing cluster configured with open source and off-the-shelf hardware and software components from ten vendors. The InfinIO 7000 is used to provide a 10Gbps interconnect and embedded fabric management for a segment of the cluster, illustrating InfiniBand's power to accelerate data throughput and optimize application performance as Xeon processor-based server systems are scaled.
In addition, InfiniCon's InfinIO Switching Systems was utilized in a demonstration by Penguin Computing to provide the low-latency, 10Gbps interconnect for an Opteron cluster running Scyld Beowulf cluster software. InfiniCon was also being showcased in HPC clustering demonstrations within the AMD Pavilion in partnership with Appro and Rackable Systems.
|