Observability with Prometheus and Beyond: Leveraging Cloud-native Technologies Outside of Cloud
summary
Prometheus runs as a single efficient binary with a pull model and a functional query language and compresses samples to about 1.3 bytes each; production instances ingest hundreds of thousands of samples/sec per core and one observed 15 million active time series. Long-term storage uses Thanos or Cortex; remote_write assumes pre-cleaned metrics. Loki stores label‑based logs (queries exceeding 40 GB/s reported) and Tempo links traces to metrics via exemplars, enabling direct jumps from a high‑latency metric into the exact trace. OpenMetrics standardizes the exposition format.
AI-written summary · about
chapters (18)
- 0:00Intro
- 3:09Dirty secret
- 4:36Buzzword alert!
- 7:12IDENTIFYING WOOD
- 7:29Complexity
- 8:59SRE, an instantiation of DevOps
- 11:28Shared understanding
- 12:24Services
- 14:31Alerting
- 16:39Prometheus 101
- 16:58Main selling points
- 20:56Time series
- 21:33Super easy to emit, parse & read
- 23:07Prometheus scale
- 26:28OpenMetrics
- 28:14Three pillars
- 29:04Loki 101
- 34:10From logs to traces
