Parallel File Systems
NFS is widely used in clusters and workstation environments, but the performance is restricted by the use of only one system operating as server. The parallel NFS, also referred to as pNFS (www.pnfs.com), will overcome this restriction by providing scalable parallel access to files distributed among multiple servers. For the success of pNFS there are two requirements:
- Implementation of the NFS v4.1 client in the linux kernel
- Availability of pNFS server components, particularly open source solutions
In 2011 a first evaluation of pNFS was made using the dCache server component. At that time, the integration of the NFS v4.1 client into the linux kernel was not yet available and the client ran a patched kernel. Also, the dCache server component was operational, but did not have the full parallel NFS functionality. The tests will be continued once the pNFS components are more mature.
Contact: Cristina Manzano firstname.lastname@example.org
The Jülich Supercomputers are running GPFS and Lustre parallel filesystems. GPFS is licensed by IBM and therefore mostly running on IBM-Hardware. On JUROPA , the alternative solution Lustre is used. Lustre has disadvantages and an unclear development future, therefore another option, the parallel file system FhGFS (www.fhgfs.com), has been evaluated on JUDGE. The tests have shown that FhGFS is operational, but with restricted hardware resources cannot compete with the enormous performance provided by the remote GPFS cluster JUST. The evaluation of FhGFS will be continued in the DEEP project.
Contact: Otto Büchner email@example.com