How to reduce expenses on monitoring with VictoriaMetrics
summary
A set of concrete cost wins with VictoriaMetrics: swapping a Prometheus-like binary yields a 7-day benchmark drop from 83 GB to 33 GB of disk, memory from ~8 GB to 4.5 GB, and 16× faster average alert-query latency; the VM agent uses an enhanced remote write that cuts inter-zone traffic about 4×. Pre-compute at the edge with streaming aggregation to shrink query time (demo shows ~8× faster) and lower cardinality, use the cardinality explorer and significant-figures tuning to drop storage bytes/sample from ~1.2 to ~0.8 (≈30%). All features are open source.
AI-written summary · about
chapters (13)
- 0:00Introduction to VictoriaMetrics
- 1:05Monitoring architecture overview
- 2:29Cost factors in monitoring
- 3:14Binary swap benchmark
- 8:54Network traffic reduction
- 11:03Significant figures for compression
- 12:48Query tracing and analysis
- 15:41Cardinality exploration
- 18:53Streaming aggregation demo
- 21:27Streaming aggregation summary
- 22:10Complexity penalty in systems
- 24:39Resources and references
- 25:22Q&A session
related talks
40mScale Your Monitoring Solution With the VictoriaMetrics Ecosystem
31mObservability: Dust Off Prometheus with VictoriaMetrics
20mAggregating Metrics In-Flight: Challenges, Opportunities
38mAggregating Metrics In-Flight: Challenges and Opportunities
19mCutting Through Metrics Cardinality Noise with VictoriaMetrics
36mVictoriaMetrics: scaling to 100 million metrics per second (EN)
28mThe Sixth Sense: Finding “Dead” Metrics and Silent Alerts in Your Monitoring
05mLightning Talk: How to detect and delete unused metrics?