Fast Parallel Storage
Overview
This service provides high-throughput, low-latency parallel storage for HPC and AI workloads that require fast shared access to large volumes of data during training, simulation, and large-scale processing. It is intended as the primary high-performance storage layer for active workloads running on SLAIF compute resources. The service is based on parallel file-system technologies and high-speed storage infrastructure optimized for concurrent access from compute nodes, integrated with the HPC environment to support data-intensive processing and training pipelines. Currently a Lustre filesystem is available, which does not provide data isolation. Security and isolation will be provided by the new system in Q3/2027.
Outcome
High-performance data access during model training, simulation, and data processing, reducing I/O bottlenecks for active workloads.
Service Card
| Category | Infrastructure and data-related services |
| Domain | High-performance Data Infrastructure |
| Vertical | Cross-domain |
| Target users | AI Users, Researchers, HPC Users, SMEs, and Startups |
| Service type | Infrastructure Service / Managed |
| Outputs | Fast shared storage space for active HPC and AI workloads. |
| Requirements | Project registration, approved access credentials, and an assigned disk quota (if requested as part of the project application). |
| Limitations | Designed for active workloads rather than long-term retention; data lifecycle management and quotas apply. |
| Availability | Available (new system available in Q3/2027) |
| Licensing | |
| Visual content | |
| Visual caption |