GPU nodes
For model training, inference, and GPU-accelerated simulation.
- Nodes
- 16
- GPUs per node
- 8
- Memory
- 1 TB / node
- Interconnect
- 200 Gb/s
Research computing platform
GPU and CPU resources for research and teaching, with projects, allocations, and usage managed in a single place. Sign in and work from your first notebook to a full multi-node run.
Hardware
Two classes of compute and a shared data platform, scheduled as one system.
For model training, inference, and GPU-accelerated simulation.
For high-throughput and memory-heavy workloads.
Shared storage sized for large datasets and fast staging.
Platform
The portal is powered by Apache Airavata Custos: identity, projects, allocations, and usage in one place, from request to result.
Sign in through your organization's identity provider. No separate accounts to manage.
Request time, add collaborators, and track consumption against your award without leaving the browser.
PIs, co-PIs, allocation managers, and members, with the permissions each actually needs.
Per-project and per-user usage against every resource, kept current as jobs complete.
Connects to national allocation programs and identity registries, so awards flow in automatically.
Built on Apache Airavata Custos. Inspect it, extend it, or run your own.
Access
Allocations for research and teaching. Who the portal serves and how to get time.
Project allocations for data- and compute-intensive science in every domain.
Apply for timeCourse allocations and training programs on the same system research runs on.
Request course timeCollaboration opportunities for institutions building on shared research infrastructure.
Get in touchSign in, join a project, and follow the quickstart to your first job.