Hitchhiker's guide from traditional HPC Cluster to containerized ensemble run at scale
summary
Describes migrating GROMACS molecular dynamics workloads from traditional HPC to a cloud-native SLURM cluster using SPACK and AWS ParallelCluster, benchmarking GROMACS across AMD/ARM/Intel CPU instances and GPU/interconnect-equipped instances to find best performance and cost-performance. Explains evolving to run very large ensemble campaigns (order 20,000 replicas) globally by using containers to provide a consistent software environment across diverse cloud hardware and thereby dramatically shorten time-to-solution.
AI-written summary · about
related talks
47mCloud Native and Sustainable, Reproducible Scientific Computing
27mFrom Supercomputing to Serving: A Case Study Delivering Cloud Native Foundation Models
16mA novel Cloud High Performance Computing infrastructure for user
33mHPC on AWS: Innovation Without Infrastructure Constraints
38mBeyond Kubernetes: Adapting To Spe...
10mLightning Talk: Does Cloud Elasticity Pay off for HPC Workloads?
21mKeynote: CERN Experiences - Ricardo Rocha & Clenimar Filemon
07mCloud architecture for population-scale analyses of genomes and AI.