Deployed Services & Ecosystem
Science IT deploys and manages a suite of cloud-native infrastructure components to power lab workloads:
Cluster Services Matrix
Section titled “Cluster Services Matrix”| Service | Technology | Description | Access / URL |
|---|---|---|---|
| User Portal | Astro + React | Self-service namespace and credential hub | berkelium.lbl.gov/portal |
| Container Registry | Harbor | High-speed local OCI image registry with vulnerability scanning | registry.berkelium.lbl.gov |
| Observability | Grafana + Prometheus | Live CPU, memory, NVLink, and GPU thermals dashboard | grafana.berkelium.lbl.gov |
| Secrets Engine | HashiCorp Vault | Secure credential and token management for research containers | vault.berkelium.lbl.gov |
| Distributed Ray | KubeRay Operator | Custom resource definitions (RayCluster, RayJob, RayService) | In-cluster CRDs |
| JupyterHub | Zero-to-JupyterHub | Browser-based interactive Python/Julia/R notebooks | jupyter.berkelium.lbl.gov |
| Object Store | Ceph S3 Gateway | High-throughput S3-compatible bucket storage | s3.berkelium.lbl.gov |
Harbor Container Registry Access
Section titled “Harbor Container Registry Access”Berkelium operates an in-cluster Harbor container registry with 100Gbps network connectivity to all compute nodes. Using this registry significantly speeds up pod launch times by caching base images locally:
# Log in with LBNL OneID credentialsdocker login registry.berkelium.lbl.gov
# Tag and push your research containerdocker tag my-pytorch-model:latest registry.berkelium.lbl.gov/sci-als/pytorch-model:latestdocker push registry.berkelium.lbl.gov/sci-als/pytorch-model:latest