MPAK Technologies - Computing

Home

Products

Solutions

Support

Nodes

Interconnects

OS Software

Storage

Configurations


	Interconnects
	Depending on your application, I/O requirements, and or budget, BeoClusters are available with the latest fabric technology from low latency Infiniband, 10GbE, to Gigabit Ethernet.

	Infiniband
	A high-performance low-latency fabric technology designed for parallel computing and/or demanding I/O applications. MPI ping latencies as low as 1.2 microseconds and up to 1.9GB/s of uni-directional bandwidth. InfiniBand can scale from a small 8-node cluster to a large 1000+ node multi-site grid supporting different applications.

	QLogic HCAs/Adapters:
	QLogic InfiniBand HCAs feature the industry's highest message rate. Combined with extremely low MPI latency and very high effective bandwidth, this enables MPI and TCP applications to scale to thousands of nodes with unprecedented price-performance.
	QLogic Switches:
	The 9000-series 10Gbs (SDR) and 20Gbs (DDR) high-performance infiniband switches can be configured from entry level 8-node systems up to 1000+ nodes. The 9000-series switches are available in 24, 36, 48, 60, 72, 84, 96, 144, and 288-port configurations.

	Mellanox HCAs/Adapters:
	Mellanox ConnectX™ and InfiniHost™ HCAs deliver low-latency and high-bandwidth for performance-driven server and storage clustering applications in Enterprise Data Centers, High-Performance Computing, and Embedded environments. ConnectX IB simplifies network deployment by consolidating clustering, communications, storage, and management I/O and providing enhanced performance in virtualized server environments.

	Flextronics Switches:
	Flextronics offers a wide range of 10Gb/s SDR and 20Gb/s DDR Infiniband switches ranging from entry level cost-effective to high-performance high-port count.

	10Gbit and 1Gbit Ethernet
	Standards based protocols utilizing cost-efficient on-board gigabit ports to high-performance 10Gb with TCP-offload technology "TOE". Some application are perfectly suited for gige, even some MPI apps that don't require extremely low-latency. For a extremely cost-effective MPI solution, an optimized MPI such as ScaliMPI can be utilized to dramatically improve latency performance using standard gigabit ethernet.

	Return to top