Case Study: CERN achieves 50,000 cores in production across three OpenStack clouds with Puppet

A Tale of Three OpenStack Clouds : 50,000 Cores in Production at CERN

CERN’s IT department, which supports about 11,000 physicists and the massive data stream from the Large Hadron Collider, faced the challenge of modernizing its infrastructure during an LHC upgrade window. Large High Level Trigger server farms would be idle but were attached to an isolated technical network, so CERN needed a way to repurpose those resources securely, allow experiments to choose their own stacks, and rapidly start/stop large numbers of virtual machines.

CERN and its experiments adopted OpenStack for cloud orchestration and Puppet for configuration management (CMS migrated from Quattor), leveraging community modules and collaboration with partners like Brookhaven. The result was three production OpenStack/Puppet clouds (ATLAS, CMS, and CERN IT) running tens of thousands of cores—about 60,000 cores managed (roughly 50,000 effective)—supporting simulation and analysis workloads with strong isolation and rapid VM turnover, and plans to scale to ~300,000 cores.

CERN

Tim Bell

Infrastructure Manager