Downsampling: Trading Granularity for Speed in Time Series Data
Downsampling trades resolution for speed by summarizing old, high-granularity metrics into coarser ones. This makes long-range queries faster and cheaper, common in systems like Thanos for long-term Prometheus data.
Downsampling trades metric precision for query speed by 'rolling up' high-granularity points into coarser ones, like turning per-second data into per-minute averages. It's essential for long-term storage in monitoring systems like Thanos, making dashboards that query months of data remain fast. The main footgun is downsampling too soon; you must wait for enough raw data to accumulate (e.g., 40 hours) before you can create accurate, lower-resolution aggregates.
Read the original → thanos.io
- #monitoring
- #sre
- #thanos
- #time series
Get five bites like this every day.
Tezvyn delivers a daily feed of 60-second tech bites with quizzes to lock in what you learn.