Netflix Serves 84% of Query Results from Cache with Interval-Aware Caching in Apache Druid
Scored daily by a customisable AI persona to surface the most relevant engineering leadership news.
Netflix Druid caching is a solid infrastructure case study but not agent-focused.
Netflix's interval-aware caching for Apache Druid serves 84% of query results from cache by decomposing rolling window queries into granularity-aligned buckets with exponential TTL, reusing historical segments via an external proxy layer while recomputing only the most recent interval. This reduced query load by 33%, improved P90 latency by 66%, and cut result bytes up to 14x for real-time analytics dashboards processing trillions of rows.
Adopt interval-aware caching with time-aligned segments and exponential TTL to avoid recomputing overlapping historical data in rolling window queries.
For engineers building data-intensive monitoring or experimentation dashboards, this technique directly reduces compute costs and latency by eliminating redundant scans on sliding window queries—a pattern common in real-time analytics at scale.