EtherFabric improves High Performance Computer Cluster performance by:

• Providing more CPU cycles to the applications
• Improving the performance of networked servers
• Increasing network bandwidth for applications
• Allowing existing applications to run unmodified

High Performance Computing, or HPC, is a branch of computer science that concentrates on using high performance computers and specialized software to solve complex problems. Parallel processing has driven the need for high performance interconnections in between the computers cooperating in a cluster. A High Performance Computing Cluster, or HPCC, which pools computing resources together to solve problems has become the preferred method of HPC.

EtherFabric’s features are performance enhancing benefits for the High Performance Computing Cluster.  The CPU efficiency of EtherFabric provides more CPU cycles to the application increasing the compute power of the server by allowing the program to execute faster.

The enhanced throughput and reduced the latency of EtherFabric improves networked servers performance and allow them to work together more efficiently. Efficient server to server communication is critical in the HPCC, where multiple servers must cooperate to solve complex computational problems.

EtherFabric can aggregate 2 Gigabit Ethernet ports together logically to accelerate a single TCP/IP application flow, doubling the useable bandwidth between servers. EtherFabric’s port striping mode delivers the high performance bandwidth that networked server need in the HPCC

Most importantly EtherFabric is 100% binary compatible and implements the exact Sockets and MPI interfaces such as Scali, LAM, MPICH used by applications.  This results in all existing applications gaining performance improvements by simply installing and EtherFabric NIC in the HPC servers.

Back to top