Optimizing price performance with feature management and observability (PRT259)
summary
Honeycomb migrates production workloads from x86-64 to AWS Graviton2 on EC2 and Lambda using progressive delivery with LaunchDarkly feature flags and Honeycomb observability. The migration yields 30% fewer instances, 30% lower latency, and a 10x increase in workload capacity; one early Lambda rollout spikes cost 50x until rolled back by flipping a flag. Feature flags stream changes in ~200 ms and enable safe, measured rollouts that work.
AI-written summary · about
chapters (14)
- 0:00Introduction
- 7:24What is observability
- 9:15How we use feature flags
- 10:12Feature flags are not a big bang
- 11:57Graphs
- 12:30Ingest
- 13:07SLO
- 14:33Graviton
- 16:13Lambda
- 17:03Better observability
- 18:22Cost per query
- 19:03Feature Flags
- 22:37Playing on Fridays
- 24:22Fear of deploys
related talks
13mGraviton2: How Honeycomb Reduced Infra Spend by 40%
32mUsing Serverless Functions for Real-time Observability
33mSupercharging Observability with Feature Flagging
29mCooking up o11y w/ Feature Flagging
46mArchitecting for Success and Runway
50mObservability and the Development Process
29mGitOps Sustainability with Flux and arm64
25mBest practices for advertising and marketing technology on AWS