The PVFS, an open-source parallel file system, is a joint project with the Parallel Architecture Research Laboratory at Clemson University and the Mathematics and Computer Science Division at Argonne National Laboratory. The latest version of PVFS i.e., PVFS2 available on pvfs.org
PVFS2 implemented transparently using a kernel module to use the Kernel Virtual File System and then mounted by the client like any file system. The file system is organized into stripes of data similar to a RAID array and distributed to different hosts on the network.
It distributes file data across multiple servers and provides for concurrent access by multiple tasks of a parallel application. It implements a distributed network file system using a management server that manages the file system on different I/O servers.
Management servers pvfs2-server maintain the file system information, which includes access permission, directory structure, and metadata information. Use the pvfs2-server service script to manage the server. On I/O servers, it creates storage space and turns the server on automatically. The I/O servers use the /etc/pvfs2-fs.conf configuration file.
The PVFS manager server uses pvfs2-fs.conf and pvfs2-server configuration files, created using configuration script pvfs2-genconfig. The script prompts for configuration information like protocol, port, log file, and storage directory.
A client requests to access the file submitted to the management server, which then sets up the connection between the client and the I/O servers that hold the required file's data. The access operations such as read and write carried out directly between the client and the I/O server.
Each client needs PVFS daemon, pvfs2-client, and its supporting module and library. At the client end, a user needs to install file system configuration information and access tools. File systems and mount configurations held in /etc/pvfs2tab, through which a user can place entries directly in the /etc/fstab file.
To mount PVFS file system, use pvfs2 type in the mount command with the -t option. To list and access files on file system use commands:
A direct Linux file system access provided with the MPI-IO interface requires recompiling using the libpvfs2 library. Parallel support in PVFS2 implemented with the message passing interface (MPI) supported by ROMIO and MPICH2 available from the MCS Division of the Argonne National Laboratory.
- BeeGFS is a parallel cluster file system, developed with a strong focus on performance and designed for effortless installation and management.