Global Capacity Management with Flux
summary
Flux uses RPC tracing to attribute backend capacity to products, feeds those attributions into a joint mixed-integer programming solver that simultaneously assigns service capacity and product traffic, and then executes plans via a capacity orchestrator tied to autoscaling and traffic systems. It runs in 20+ regions, has operated ~2.5 years serving thousands of services, produces plans in ≈5 business days and executes them in ≈10, and supports hardware refresh and global utilization rebalancing while shrinking disaster buffers.
AI-written summary · about
chapters (8)
- 0:00Flux and global capacity
- 2:34Solving regional imbalances
- 3:15Service modeling with RPC tracing
- 8:54Demand attribution gaps
- 11:20Capacity placement with MIP
- 14:32Iterative planning cycle
- 15:58Flux use cases and benefits
- 18:44Flux impact and future
related talks
16mGlobal Capacity Management With Flux
20mGlobal Capacity Management
36mManaging Thousands of Clusters and Their Workloads with Flux
05mFluxCapacitor: A Highly Scalable App
05mFluxCapacitor: A Highly-scalable, Highly-available Distributed Reference App
18mAvoiding and Breaking Out of Capacity Prison
27mGitOps Evolution Through the Lens of a Flux Maintainer
38mAutoscaling Flink at Netflix
