GPFS

GPFS, now IBM Spectrum Scale, is a high-performance distributed file system designed for scalable, shared-disk storage across clusters. It provides parallel access to data and metadata, enabling many clients to operate atop a single namespace with high throughput and low latency. The system is POSIX-compliant and supports multiple access protocols, including NFS and SMB, in addition to its native file system interface.

GPFS uses a clustered architecture with data storage servers (OSS) that host the blocks and metadata servers (MDS) that manage metadata and namespace operations. Clients mount the file system over the network, and a distributed locking and metadata management layer coordinates access to ensure consistency and fault tolerance. The architecture supports high availability through storage redundancy, failover of critical services, and optional multi-site replication.

Key features include scalable metadata management, automatic data distribution across disks and nodes, tiered storage integration (SSD, HDD, tape), and data protection through replication or erasure coding, as well as point-in-time snapshots and space-efficient backups.

It is commonly deployed in high-performance computing, data analytics, large-scale scientific research, and media/workflow environments.

Historically developed by IBM in the 1990s as GPFS, the software is marketed under the IBM Spectrum Scale name in many enterprise contexts. GPFS continues to be referenced in product documentation and user communities as a high-performance parallel file system for large-scale workloads.