Vector computing
HPC and AI compute
HPC storage
Resource infrastructure
Enterprise software
Resource disaggregation
Ensuring the reliability and reproducibility of HPC simulation results has become a critical requirement in both academic and industrial research. Yet in practice, the process that produced a result is often difficult to reconstruct — manual recording is time-consuming, incomplete, and prone to human error.
| The Challenge | What DPS4HPC Delivers |
|---|---|
| Information required to reproduce computational results is often not preserved — input data and processing steps cannot be traced after execution | Automatically captures the full relationships between input data, processing steps, and outputs — complete traceability with zero additional effort |
| Computational workflows become black boxes over time, and data management relies heavily on manual, error-prone effort | Automates provenance recording — DPS4HPC captures provenance automatically as jobs run, reducing human error and administrative burden |
| Reproducing past experiments or sharing computational processes with reviewers and collaborators is difficult and time-consuming | Enables accurate reproduction of past workflows using recorded provenance data — supporting peer review, validation, and seamless knowledge transfer |
DPS4HPC addresses the most common reproducibility and data management challenges in computational research environments.
| Use Case | Research Setting | What DPS4HPC Enables |
|---|---|---|
| Identifying Data to Preserve | Research generating large numbers of intermediate files across multiple experimental runs | Provenance graph shows exactly which inputs contributed to final results — preserve only what matters |
| Reproducing Experiments | Peer review, validation, and knowledge transfer when team members change | Past experiments can be accurately reproduced using automatically recorded job-level provenance data |
| Audit & Compliance | Research subject to funding body requirements, institutional governance, or international standards | Complete, tamper-evident record of how results were produced — ready for review at any time |
DPS4HPC fits into your existing HPC workflow without any changes. Users simply log in and submit jobs as usual — the system captures provenance automatically in the background.
DPS4HPC captures, stores, and visualizes provenance information automatically — without requiring any modification to how researchers run their jobs.
Zero-Effort Provenance Capture
Comprehensive Job-Level Recording
Search and Visualize Provenance Data
Reproducibility and Collaboration
* FAIR (Findable, Accessible, Interoperable, Reusable) is a set of guiding principles for research data management, widely adopted by funding bodies and research institutions worldwide.
DPS4HPC consists of the following components, working together with your job scheduler to capture, store, and visualize provenance information.
| Job Scheduler | Captures job-level metadata via a wrapper script integrated with the scheduler |
| Tracer | Captures process execution details and file I/O activity on compute nodes with minimal performance impact |
| Aggregator | Aggregates and correlates file I/O activity, process execution metadata, and job-level metadata to generate structured provenance records |
| Provenance Database | Stores provenance data and relationships for search, retrieval, and long-term access |
| Management GUI | Provides a web-based interface for search, graph visualization, and sharing of provenance information |
As AI inference, data processing, and shared research workloads continue to grow, organizations are investing heavily in GPU infrastructure. However, conventional server-centric architectures often lead to low GPU utilization, rising power and cooling demands, poor investment efficiency, and limited operational flexibility.
Request a quote: our team will confirm availability, compatibility and a deployment plan.
