{"resourceId":"lbl-berkelium-kubernetes-science-platform-20260925","versions":[{"version":"external-0eda27531243145ce9fcf56c3afec4cdb37a9dbe84ec2e22deb1ad0972864a06","resource":{"id":"lbl-berkelium-kubernetes-science-platform-20260925","title":"Berkeley Lab ScienceIT launches Berkelium: Kubernetes platform for containers, AI model serving, and reproducible scientific services","organization":"Lawrence Berkeley National Laboratory (ScienceIT)","sector":"National laboratory research computing","geography":"United States (Berkeley, CA)","publishedAt":"September 25, 2026","publicationDate":"2026-09-25","eventDate":"2026-09-25","sourceName":"Berkeley Lab IT / ScienceIT","sourceLabel":"National laboratory IT announcement","sourceUrl":"https://it.lbl.gov/meet-berkelium-a-flexible-new-kubernetes-based-platform-for-modern-science/","evidenceClass":"independent-reporting","outcomeClass":"emerging","topics":["infrastructure","developers-agents","operating-model"],"finding":"On 25 Sep 2026 ScienceIT announced Berkelium, a new Kubernetes cluster for building, testing, automating, and sharing scientific work. It complements traditional HPC (Lawrencium) for interactive apps, automated workflows, persistent services, and containerized environments that support reproducibility. Documented use cases include AI training and model serving (CBorg already runs services on Berkelium), automated instrument-to-workflow data pipelines, and browser-shared results. Training is scheduled 15 Oct 2026.","sledRelevance":"University and lab research computing teams adopting AI agents need a service/Kubernetes tier beside batch HPC so model endpoints, persistent workflows, and containerized reproducibility are first-class—not bolted onto Slurm alone.","evidence":"ScienceIT announcement page dated September 25, 2026 opened via HTTPS; complementarity to Lawrencium, container reproducibility language, and CBorg-on-Berkelium stated in body.","architectureImplications":"Two-tier research platform: Lawrencium-class batch HPC for heavy simulation plus Kubernetes (Berkelium) for containers, persistent services, AI endpoints, and instrument pipelines.","governanceImplications":"Launching a shared Kubernetes science platform implies access control, OneID-style auth, and training before researchers treat persistent AI services as production.","securityPrivacyImplications":"Persistent model-serving and instrument pipelines raise credential, data-egress, and multi-tenant isolation requirements beyond batch job sandboxes; announcement does not publish a full security model.","caveats":"Announcement/launch article without measured throughput or failure rates; hardware counts on related portal pages not used as primary evidence here; training still upcoming.","streamIds":["research"],"roles":{"sales":"Sales framing: Research computing leaders stuck between batch HPC and ad-hoc AI notebooks need a governed Kubernetes science tier. Berkelium’s launch pitch is complementarity—Lawrencium stays for heavy simulation while Berkelium covers containers, persistent services, AI training/model serving (CBorg already hosted), and instrument data pipelines with reproducibility via container packaging. Position as platform enablement and training (15 Oct session), not a TOP500 replacement or measured discovery accelerator yet.","engineering":"Engineering framing: Deploy a Kubernetes research cluster beside existing HPC, with GPU pools for training and model serving, container registries, and clear pathways from laboratory instruments into automated workflows and databases. Integrate institutional identity, namespace quotas, storage classes, and endpoint patterns similar to CBorg-on-Berkelium. Publish placement rules so MPI-heavy simulation stays on Lawrencium-class systems while interactive AI services use Kubernetes. Proof of value is reproducible environments, stable services, and reduced VM sprawl—not peak FLOPS.","delivery":"Delivery framing: Stand up researcher namespaces, onboarding documentation, GPU model-serving patterns, and the scheduled October training; migrate two to three persistent AI or instrument-pipeline services off fragile VMs onto Berkelium-class Kubernetes. Acceptance should include written HPC-versus-Kubernetes placement guidance, working institutional identity login, owner runbooks for each migrated service, and a short reproducibility check that the same container image reruns for a second user—not a launch blog post alone."},"roleDetails":{"sales":{"customerProblem":"Modern science needs persistent AI services and containerized workflows that traditional batch HPC queues handle poorly.","stakeholders":["ScienceIT/research computing","Lab CIO","AI platform owners (e.g., internal LLM services)","Instrument facility leads","Reproducibility/compliance staff"],"discoveryQuestions":["Which services need 24/7 endpoints vs batch jobs?","Where do instrument pipelines break today?","What identity and quota model exists?","Who owns model-serving SLOs?"],"valueHypothesis":"A Kubernetes science platform beside HPC can host AI model serving and automated pipelines with container-level reproducibility, if access and training are funded.","potentialEngagement":"Pilot migration of one internal AI service and one instrument pipeline onto a research Kubernetes cluster with training.","unsupportedClaims":["Replaces Lawrencium-class HPC","Measured discovery speedup","Guaranteed multi-lab portability","Fully specified security accreditation"]},"engineering":{"fit":"Kubernetes research cluster for containers, persistent services, AI training/serving, and instrument pipelines complementary to batch HPC.","architectureIntegration":"Connect identity (OneID-class), GPU operators, registries, storage classes, and ingress for model endpoints; keep heavy MPI on HPC.","prerequisites":["GPU node pool","Institutional auth","Container registry","Storage classes","Ops on-call for persistent services"],"constraints":["Launch announcement without capacity KPIs","Training still scheduled","Single-lab context","Security model not fully public"],"security":"Enforce namespace isolation, secret management, egress controls for model APIs, and data classification for instrument feeds.","proofOfValue":["Service uptime for AI endpoints","Successful container reuse across users","Pipeline latency from instrument to store","User completion of onboarding training"]},"delivery":{"work":"Launch cluster access, publish HPC-vs-K8s guidance, run training, migrate pilot services.","dependencies":["Identity provider","GPU capacity","Facility networking","Service owners for CBorg-like apps","Training staff"],"ownership":"ScienceIT owns platform; service owners own app SLOs; facilities own instrument data feeds.","skillsAdoption":"Teach researchers container packaging and when to use Kubernetes vs batch HPC; train ops on GPU serving.","governanceCheckpoints":["Access policy","Data egress review for model APIs","Pilot go-live review"],"acceptanceCriteria":["Onboarding path live","At least one AI serving or pipeline pilot in production","Documented placement rules"],"risks":["Researchers misuse K8s as unmanaged VMs","GPU contention with batch AI","Credential sprawl on persistent services","Overpromise reproducibility without image provenance"]}},"retrievedAt":"2026-09-26T05:20:57-05:00","enrichedAt":"2026-09-26T05:21:15-05:00","enrichmentBasis":"retrieved source","accessibilityWorkforceImplications":"October training and container self-service can widen who can run modern AI/service workloads; skill barrier shifts from cluster modules to container literacy.","procurementImplications":"Budget Kubernetes ops, GPU serving, and training alongside HPC refresh—not only more batch nodes.","operatingModelImplications":"Explicit two-tier operating model (batch HPC + Kubernetes science platform) with ScienceIT ownership of the service tier.","sourceVerification":{"openedUrl":"https://it.lbl.gov/meet-berkelium-a-flexible-new-kubernetes-based-platform-for-modern-science/","referenceExcerpt":"This supports reproducibility, which is an essential part of the scientific process, while giving researchers greater control","promptVersion":"sled-research-v4.0","model":null,"basis":"agent-reported inspection"}}}]}