Ricardo Rocha, CERN | KubeCon + CloudNativeCon NA 2020
summary
CERN describes moving its data pipeline onto containers and Kubernetes to handle detector cameras that take 40 million frames/sec and can produce up to 1 petabyte/sec per detector before hardware filters reduce that to tens of gigabytes/sec, accumulating over half an exabyte to date. The infrastructure team reports a demonstrated public cloud bursting via Kubernetes and plans to adopt GPUs and other accelerators plus ML for filtering and simulation to absorb an expected 10x data increase with upcoming upgrades.
AI-written summary · about
