The IBM® General Parallel Filesystem (GPFS™) is known for its performance, scalability, availability, and manageability features. It is the file system used in many High Performance Computing (HPC) solutions, scaling to thousands of servers and serving multi-petabyte file systems. GPFS is also used in many non-HPC areas, such as database solutions, but typically has been deployed at a much smaller scale in those environments both in number of servers and size of the underlying storage subsystems.
This IBM Redpaper™ publication focuses on the use of GPFS in conjunction with server virtualization, where logical partitions (LPARs) access the external storage through a relatively new Fibre Channel storage virtualization technique called N_Port ID Virtualization (NPIV). From the GPFS perspective, this case is positioned between the two above extremes: Virtualization might increase the number of GPFS nodes to the scales typically seen in HPC clusters while the number of physical servers is still relatively small (as in other non-HPC environments).
Table of contents
Architectural decisions for GPFS and storage connectivity
Hardware components for the proof of concept
Storage virtualization planning
Configuring the infrastructure
Summary and outlook