Performance tuning workflow
Performance tuning starts with the workload and the observed limiting stage. The complete workflow is now in Capacity and performance.
| Need | Guide |
|---|---|
| Estimate retained storage, replicas, and catch-up time | Capacity calculations |
| Find client, disk, replication, or downstream bottlenecks | Monitoring |
| Select batching, compression, and bounded sends | Producer APIs |
| Understand consumer parallelism and completion | Consumer overview |
| Assess flush and replica changes | Storage, HA design |
| Change retention or restart a node | Maintenance |
Use the actual service's configuration schema and current runtime-injected client APIs. Change one factor per comparison and preserve the intended durability contract. Example resource values are not universal tuning recommendations.