10 years of OpenStack at CERN (From 0 to 300k cores)
summary
CERN describes building a production OpenStack cloud since 2013 that scales to hundreds of thousands of cores, with concrete wins and pain points: deployment on Grizzly, images moved from AFS to Ceph-backed Glance, Ceilometer later removed for scale, and a long migration from Hyper-V to KVM. Operations now run 80+ Nova cells, three RabbitMQ clusters, Ironic managing 5,000+ bare-metal nodes, Magnum serving ~500 clusters, three regions to dodge Neutron limits, and recent additions like preemptible instances and plans to run the control plane on Kubernetes.
AI-written summary · about
chapters (9)
- 0:00Intro
- 0:52CERN Visualization Infrastructure
- 2:57Early days of OpenStack
- 4:23CERN Agile Infrastructure
- 8:31CERN Production Infrastructure
- 12:11Milestones
- 14:47Operation challenges
- 16:20Collaboration
- 17:08Summary
related talks
15mThe CERN Cloud Infrastructure: 5 years in 15 minutes
20mCERN OpenStack User Story
01h 50mKeynote
09mSuperuser Awards!
31mIs OpenStack Still Needed in 2022?
30mOpenStack Works, so now what?
01h 47mDistributed Power (Full Day 2 includes CERN, Expedia, TapJoy and Telco Panel)
18mOpenStack Playing for an Open World