Capacity Management for Fun & Profit
summary
Elastic Cloud runs on AWS GCP and Azure with stateful VM allocators sold in RAM increments and a 58 GB block size, and capacity work moves from manual PR’d autoscaling YAML to automation. Problems include inefficient autoscaling that spins wrong VM sizes, manual bin packing (legacy code), ghost VMs that fail to bootstrap, and cloud quotas (GCP) and VMSS limits (Azure). Built tools in Go to tag allocators, automate allocator replacement, add quota alerting, and enable just-in-time capacity, scaling across 46 regions by Sept 2021.
AI-written summary · about
chapters (28)
- 0:00Intro
- 3:38State of mind
- 4:01Up on the downside
- 6:02Limitations in "auto" 11
- 7:28Autoscaling gone rogue!
- 8:35Out of capacity
- 9:20The other Covid surge
- 10:10Reduce, reuse
- 10:35Please don't litter the allocators
- 11:01To the limit
- 11:57Blocked & Unreported
- 13:15More going rogue
- 14:12Manual operations
- 15:15Partly cloudy
- 15:46Storm Clouds
- 19:18Just in time!
- 20:02Infrastructure API
- 21:09Improvement needed
- 21:56Tag, you're it!
- 23:15Ghost bustin'
- 24:00Fleet replacement
- 25:53Limiting limitations
- 26:22Catching quotas
- 26:49Alerting, Monitoring
- 27:29Automation, Engineering
- 28:04Data-backed Decisions
- 28:41Flexibility, Adaptability
- 28:59Team Work, Communication
related talks
29mCapacity Constraints Unveiled: Navigating Cloud Scaling Realities
18mResource allocation -- The usuals and beyond
20mGlobal Capacity Management
20mManage Scaling, While Keeping Costs Under Control. Challenge Accepted.
20mCloud Cost Management Updates
45mAutoscale — Pay When You Earn
47mScaling FinOps with a holistic approach to resource optimization
25mPreparing Apps for Dynamic Scaling