Linux Networx hooks up with IBM GPFS

Linux Networx announced last week that it had signed an OEM agreement with IBM to distribute IBM’s General Parallel File System (GPFS).

The agreement between IBM and Linux Networx is the first for IBM and follows Big Blue’s move last week to open up the source code to GPFS to third parties.

Linux Networx will integrate GPFS with its Linux Supersystems and will sell, tune, optimize and support GPFS. Linux Networx last month expanded its Supersystem range with a midrange LS-1 Supersystem and a high-end LS/X supercomputer. GPFS will also be available within Linux Networx’s Advanced Technology Clusters.

Linux Networx clusters are currently available with the Lustre File System, but customers wanted a choice of open-source file systems.

So far, the National Energy Research Scientific Computing Center has implemented GPFS on a 320-node Linux Networx cluster. Each node contains two processors. 

GPFS provides a scalable parallel file system that can support hundreds of terabytes of storage within a single file system. It is designed for bandwidth-intensive applications such as computational fluid dynamics, crash analysis, structural engineering, flow simulation and interactive 3-D visualization.  

IBM’s file system has its roots in Unix, where it was used starting in the 1990s on clusters of Unix servers. It is capable of supporting multiple terabytes of storage and more than 1,000 disks within a single file system. According to IBM, GPFS runs on AIX 5L on Power processors, on Linux on IBM AMD and xSeries systems, and on Linux on Power platforms.

GPFS on Linux Networx is available now. An eight-node, dual-socket system with GPFS starts at around $8,000.