Embracing Multi-Tenancy While Scaling MLOps
Talkat Multi-TenancyCon North America 2023 · on Monday, 2023-11-06· at 16:25· W192ab, McCormick Place·
summary
Demonstrates a working Flyte setup where two projects run identical ML workflows on one cluster, creating one pod per task and storing intermediate artifacts in S3 so another project reuses a completed task’s output; enforces isolation with namespaces plus RBAC and registers workflows per projects and domains; lets training tasks request task-level GPU allocation while preprocessing uses CPU-only pods; shows measurable resource savings from task re-use and proves cross-team S3-shared artifacts and namespace RBAC enable secure multi-tenancy.
AI-written summary · about
chapters (10)
- 0:00Multi-tenancy in MLOps
- 3:12Scaling ML infrastructure challenges
- 7:43Orchestrators for ML
- 10:11Flight orchestrator introduction
- 13:36Flight architecture
- 15:45Demo: multi-tenant sharing
- 20:03Task sharing in Flight
- 22:32Resource isolation with namespaces
- 24:18GPU/CPU allocation
- 25:00Conclusion and key takeaways
related talks
30mEmbracing Multi-Tenancy While Scaling MLOps
25mMLOps with Flyte
38mFlyte: Cloud Native Machine Learning & Data Processing Platform
39mFlyte: Production-Grade Data and Machine Learning Orchestration
28mRun Your Data and Machine Learning Workflows on Kubernetes with Flyte
01h 24mRobust and End-to-End Cloud Native Machine Learning & Data Processing Platform
23mManage Multi-tenant ML Workloads Using Istio
44mBuilding Multi-Tenant Compute Systems in the Enterprise