HPSs

HPSS, or High Performance Storage System, is a scalable storage management subsystem designed for high‑performance computing environments. It provides centralized control of data placement and movement across multiple storage media, combining disk caches with long‑term archive on tape or other archival media to deliver high‑throughput access to large scientific datasets.

The system typically consists of clients that submit data transfer and retrieval requests to one or more HPSS servers. A metadata and catalog layer tracks file information, permissions, and archival status, while the actual data may reside on fast disk pools or be staged to tape libraries as needed. Data flows are managed to enable parallel I/O and efficient use of available bandwidth, with storage backends organized into tiers to balance performance and cost. HPSS is designed to operate across clusters or data centers, supporting multi‑host access and centralized policy enforcement for data placement, retention, and migration.

HPSS originated in the 1990s as a collaborative effort among U.S. national laboratories and research partners to address the needs of large‑scale scientific computing. It evolved through multiple generations to improve scalability, reliability, and interoperability with other storage systems and workflows. HPSS has been widely deployed at national laboratories and research facilities to manage petabyte‑scale datasets, including long‑term preservation and large‑volume transfers.

Key capabilities include high‑throughput parallel data transfer, scalable metadata management, data lifecycle and migration policies, and integration with tape robotics for archival storage. While newer storage paradigms and cloud approaches have emerged, HPSS remains in use in some HPC environments, particularly where centralized control of large archives and cost‑effective long‑term storage are priorities.