New Research Paper ​

Efficient scheduling of virtual machines, their associated workloads and the underlying infrastructure itself is an active research topic of fundamental importance for infrastructure providers. Because the workload characteristics (and their temporal evolution) are distinct to each datacenter and may involve multiple coprocessors, efficient provisioning is difficult and subject to one's unique constraints. The new "The SAP Cloud Infrastructure Dataset: A Reality Check of Scheduling and Placement of VMs in Cloud Computing" sheds light on this situation, by providing and analyzing telemetry data from 1,800 hypervisors and 48,000 Virtual Machines (VMs), which are part of the SAP Cloud Infrastructure, over a 30-day observation period. It was created as a collaboration between SAP and the TU Dresden. Additionally, Arno Uhlig, Engineering Manager at SAP and member of our CobaltCore project, presented the paper at the ACM Internet Measurement Conference 2025, a prestigious and known venue in this domain.

Get the paper here: https://arxiv.org/pdf/2510.23911

To learn more about this topic, follow the CobaltCore team for updates and try their Cortex scheduler if you want to immerse yourself even deeper!