Language:    

Soluzioni IT

HPC Cluster

a a a

The High Performance Computing was the first field of application of the first clusters. They are the most performant and biggest clusters. Their tasks (as e.g. Weather Forecast or Financial Forecast) are often split in smaller subtasks and then allocated to the different knots. The fast communication normally runs with MPI (Message Pathing Interface) as instrument to allocate the subtasks to the single knots. MPI uses therefor a faster network, as e.g. InfiniBand and Myrinet.

Programs based on MPI profit from shorter latent time of InfiniBand in comparison to TCP/IP based connections. Latent time is the time in a data transfer before the information appear on the resp. output port. For safing data on HPC Clusters you can chose withouth problems from the spectre of a simple NFS-Server to an iSCSI Server up to a complex LustreFS Serversystem.

As operating systems diverse standard Linux Distributions and Microsoft 2008 Servers are supported. We cope also with the complete cluster environment, as e.g. OSCAR or Rocks Cluster.