Observability is Too Damn Expensive!
summary
Observability bills commonly hit 10–30% of cloud spend; 99% of indexed observability data is never searched and 30% of ingested data is never used at all. Use Prometheus query logs and OpenSearch slow logs to see what is actually queried, push cold data straight to S3 or Parquet instead of indexing, avoid reindexing by querying archives, transform noisy JSON logs into time‑series metrics, and reduce high‑cardinality metric labels to cut cost and improve performance.
AI-written summary · about
chapters (19)
- 0:00Introduction and title
- 1:06Why observability costs rise
- 2:48Microservices and data volume
- 3:50SRE practices and chaos
- 4:54Data complexity and retention
- 6:16Wasted data and indexing
- 8:41Unused data and retention policies
- 10:04DIY observability optimization
- 11:27Understanding data usage
- 12:51Defining use cases
- 13:53Indexing and archiving data
- 16:00Re-indexing vs direct querying
- 17:43Reducing log production
- 19:48Transforming data efficiently
- 21:12Indexing cautiously
- 22:16High cardinality metrics
- 24:00SAS cost optimization tools
- 25:44Key cost optimization questions
- 26:06Summary and closing
related talks
22mObservability is Too Expensive!
30mObservability is Too Expensive - Chris Cooney
35mDriving down the cost of observability
10mWhy your observability is costing you too much?
36mObservability? It's a Data Analytics Problem, You Fool!
16mShrinking the Observability Bill
15mThree key considerations for deploying best-of-breed observability
31mMy Logs Cost More Than My Services