GPFS-WAN User Guide
Introduction
GPFS-WAN (Global Parallel File System-Wide Area Network) is a 613-TB storage system. The system is physically located at SDSC and is mounted on the TeraGrid Linux clusters at SDSC, NCSA, and UC/ANL, as well as all DataStar p655 and p690 nodes and Blue Gene at SDSC.
GPFS-WAN is recommended for long- or short-term data storage of high-volume multi-site runs, as well as for large TeraGrid-based data collections. GPFS-WAN has three partitions, each with its own policy regarding access, allocation, and data preservation: 1) a Long term collections area; 2) a Projects Area; and 3) Scratch Area (for short term data analysis).
Users with computational allocations are not charged for the use of the Scratch Area or the Projects Area: however, the Projects Area requires that users submit a request via the Project Space Request form. To obtain a data resource allocation for Long-term Collections Area users must submit a proposal. See the GPFS-WAN page on the Teragrid Web site for more details.
For special projects requiring long-term disk cache residency, contact SDSC Consulting.
System Configuration
GPFS-WAN consists of p575 nodes with 16 GB of memory per node. Each node is divided into two Logical Paritions (LPARs) for a total of 32 NSD servers. There is a total of 2GB/s of bandwidth from each NSD server. The total capacity of GPFS-WAN is ~613 TB (RAID 6). All the nodes are directly and redundantly connected to the storage.
It is the user's responsibility to back up critical data. This storage system is very reliable; however, data can be lost or damaged due to media failures, system software bugs, hardware failures, and user mistakes. Because of the enormous amount of data involved, SDSC maintains only one copy of GPFS-WAN data on disk.
