Implementation Challenges: From HPC to Containers in the Academy
Talkat KubeCon + CloudNativeCon North America 2021 · on Wednesday, 2021-10-13· at 16:30· Petree Hall C + Online·
summary
Masaryk University runs distributed Kubernetes clusters built on RKE2 with Rancher dashboard and exposes NFS, Ceph RBD and S3 plus GPUs and InfiniBand to researchers; the national HPC has 20,000 CPU cores, 200 GPUs, 60 PB and ~1,000 active users. Persistent options tried include NFS (fast but no UID remap), SSHFS (UID remap but slow and fragile) and CIFS (UID remap, limited HPC support). Scheduling, fairness and storage-auth mapping remain unsolved but tens of thousands of Nextflow jobs already run to validate the platform.
AI-written summary · about
chapters (16)
- 0:00Intro
- 0:11Who we are
- 0:40Overview
- 1:38HPC: Resources
- 2:43HPC: Troubles
- 3:52HPC: Containers
- 4:38Containers: Infrastructure
- 5:29Containers: e-Infra cz
- 7:31Containers: Benefits for Users
- 8:11Containers: Challenges
- 9:47Challenges: K8s - HPC Integration
- 15:17Challenges: Queueing and Fairness
- 17:15Challenges: Scheduling
- 19:50Challenges: User Trust
- 20:42Future Plans
- 22:11Conclusion
related talks
31mManaged Kubernetes — Next Gen Academic Infrastructure?
46mBuilding Supercomputers Using Kubernetes and OpenHPC
41mBuilding & Operating a Large-scale HPC AI Cluster on Kubernetes
08mLightning Talk: Orchestrating Kubernetes Clusters on HPC Infrastructure
29mTowards Sustainable Multi-Cluster Management in Real-World Academic Infrastructure
19mKeynote: Building Bridges: Cloud Native and High Performance Computing
47mCloud Native and Sustainable, Reproducible Scientific Computing
37mRunning Mixed Workloads on Kubernetes at the Institute for Health Metrics and Evaluation