Keeping a Hypervisor Fleet Up to Date with Minimal Customer Disruption
summary
A production scheduling service models host reboots and VM migrations as interval variables and uses OR-Tools CP-SAT to optimize who and when to update, aiming to finish 10,000-host rollouts in 90 days (~100 hosts/day) while enforcing capacity budgets, per-customer migration limits and data‑center concurrency caps. The fleet maintenance planner emits daily schedules consumed by the control plane, provides a provable lower bound so operators see optimality, and replaces manual spreadsheet planning, cutting planning toil and improving rollout predictability.
AI-written summary · about
chapters (9)
- 0:00Hypervisor fleet maintenance
- 1:02Large-scale update challenges
- 2:47Maintenance scheduling issues
- 5:56Mathematical optimization solution
- 8:04OR-Tools CP-SAT solver
- 10:11Fleet maintenance planner
- 11:59Control plane integration
- 13:03Solution quality evaluation
- 14:47Summary
related talks
50mOps issues
28mDesigning for Failure: How to Manage Thousands of Hosts Through Automation
35mEmbrace Fleet Reboots and Make Them Boring
17mPIMS: Fleet-Wide Datacenter Maintenance with Minimal Capacity Buffer and Predictable Lat…
33mScalable methods to patch and upgrade your hybrid fleet of Oracle databases
30mEnsuring a Consistent Installation Time for ROSA with Hosted Control Planes
25mAutomating Kubernetes Upgrades for Zero Downtime
14mDeep Optimization of VMM Live Upgrade
