Building an HPC infrastructure at a Liberal Arts College
summary
Design and prototype for an HPC cluster at Pomona College targets summer 2018 student research and a 2020 distributed grid; hardware is ordered but not yet deployed. Choices: two‑socket x86 servers (256–768 GB RAM), AMD Epyc or Intel Skylake CPUs, NVMe local scratch plus fast shared storage, Mellanox 10/40G networking, and GPU nodes using NVIDIA Volta (Tesla V100). Management stack uses Bright Computing, Slurm scheduler, Ansible for post‑provisioning, Docker containers, GitHub for versioning, hosted ELK and Prometheus for monitoring; one admin runs initial support.
AI-written summary · about
chapters (14)
- 0:00Start
- 4:53Challenges
- 7:08Must-Haves?
- 14:10NVIDIA-specific support
- 15:363 vendors considered and finalized
- 16:33Cluster Provisioning and Management
- 20:36Post-Bare-Metal Provisioning
- 22:10Scheduler
- 27:09Storage
- 30:04Software
- 30:18Kubernetes for scaling established workflows
- 30:55Exploring
- 33:09Pomona Use Case 4
- 33:47Semantic Web or Web 3.0
related talks
16mA novel Cloud High Performance Computing infrastructure for user
21mBuilding Distributed Platforms at Extreme Scale on Fast Changing Data
21mScala(ble) High-Performance Computing
09mThe DNA of Next-generation Data C
32mHPC Multi-Cluster Database Deployment: Service Level and Availability Driven Refactoring
37mRunning Mixed Workloads on Kubernetes at the Institute for Health Metrics and Evaluation
18mThe "New Normal" or An Evolution
08mThe high performance stack