Administration And Operation; Higher Performance/Scalability; Recoverability - IBM RS/6000 SP Handbook

9.2.1 Administration and operation

GPFS provides functions that simplify multinode administration and can be
performed from any node in the SP configuration. These functions are based on,
and are in addition to, the AIX administrative commands that continue to operate.
A single GPFS multinode command can perform a file system function across
the entire SP system. In addition, most existing UNIX utilities will also run
unchanged. All of these capabilities allow GPFS to be used as a replacement for
existing UNIX file systems where parallel optimization is desired.
All file systems provide a number of data management services, such as the
ability to duplicate files, remove files, rename files, and so forth. As a POSIX file
system, GPFS appears the same as any other POSIX file system but with
exceptional capacity and read (or write) performance. The user performs
ordinary file system operations, such as copying a file, and so forth, on GPFS in
the same manner as they would on other standard file system.

9.2.2 Higher performance/scalability

By delivering file performance across multiple nodes and disks, GPFS scales
beyond single-server (node) performance limits. Unlike NFS and JFS, GPFS file
performance scales as additional file server nodes and disks are added to the SP
system.
In a parallel environment GPFS can outperforms:
PIOFS
Network File System (NFS)
Distributed File System (DFS)
Journalled File System (JFS

9.2.3 Recoverability

GPFS can survive many system and I/O failures. Through its use of the RS/6000
Cluster Technology capabilities of PSSP in combination with IBM RVSD, GPFS
is able to automatically recover from node, disk connection and disk adapter
failures. GPFS will transparently failover locked servers and other GPFS central
services. Through its use of IBM RVSD, GPFS continues to operate in the event
of disk connection failures. GPFS allows data replication to further reduce the
chances of losing data if storage media fail. Unlike PIOFS, GPFS is a logging file
system that allows the recreation of consistent structures for quicker recovery
after node failures. GPFS also provides the capability to mount multiple file
systems, each of which can have its own recovery scope in the event of
component failures.
268 RS/6000 SP and Clustered IBM ^ pSeries Systems Handbook