CCS provides several storage platforms to support research computing at the University of Kentucky. These systems are designed for high-performance computation, large-scale data workflows, and long-term research data retention. The storage offerings fall into three broad categories: parallel file systems, object storage, and network-attached storage.

This page provides a high-level overview of each system.

Parallel File Systems (GPFS)

Purpose: High-performance POSIX storage for active computation, scratch data, and project work.

CCS operates multiple GPFS file systems that provide high-throughput access to compute clusters through InfiniBand.

Capacity

  • Total usable GPFS capacity: ~6.0 PB
  • Total raw capacity: ~8.0 PB

Available Filesystems

  • /home – User home directories (10 GB quota)
  • /scratch – Temporary high-performance workspace (25 TB per user; 90-day purge)
  • /project – Shared project data (per-project allocation)
  • /pscratch – Large temporary workspace for projects (50 TB per project; 90-day purge)

Important Notes

  • Data stored on GPFS is not backed up.
  • Scratch and project-scratch areas are purged if files are not accessed for 90 days.
  • These filesystems are optimized for active compute, not long-term archiving.

Object Storage (Ceph)

Purpose: Large-scale, durable object storage for datasets, archives, and streaming-style workflows.

CCS maintains multiple Ceph clusters that provide S3-compatible object storage. This storage model is designed for large datasets, data sharing workflows, and applications that can make use of object interfaces.

Capacity

  • Total usable Ceph capacity: ~3.0 PB
  • Total raw capacity: ~5.3 PB

Usage

  • Object storage is available on request for research groups with workflows suited to S3 APIs or large-scale data retention.
  • Ceph is not a POSIX filesystem and is not intended as a substitute for GPFS during compute.

Important Notes

  • Ceph object storage is not backed up.
  • Object storage performance depends on the access protocol and client workflow.

Network-Attached Storage (NAS)

Purpose: Long-term storage of research datasets not actively being processed on HPC systems.

CCS operates several large-capacity NAS systems used for long-term retention and archival of non-sensitive research data.

Capacity

  • Total raw NAS capacity: ~15.7 PB
    (Combined across multiple systems)

Characteristics

  • Accessible from campus networks for storage, retrieval, and staging.
  • Suitable for datasets that need to persist beyond active computation.
  • Not designed for high-performance parallel job I/O.

Important Notes

  • NAS systems are not backed up.
  • Intended for long-term storage, not high-performance computing.

Data Protection Notice

No CCS storage system is backed up.
Users are responsible for maintaining their own off-site or redundant copies of important research data.

If you are interested in learning more about the different types of storage systems we offer, please visit this page.