From full recomputations to incremental cache reuse: Apache Doris query cache in practice

This session was part of Apache Doris Summit 2026. Abstract: Benedict Jin will share how Incremental Merge in Apache Doris Query Cache tackles this challenge, making repeated analytical queries much faster under continuous writes. A seasoned software engineer and open-source contributor, Benedict brings extensive experience in distributed systems and database internals. In this session, Benedict will dive into how Apache Doris incrementally merges cached aggregation states with newly appended data, avoiding unnecessary full recomputations while preserving query correctness. He’ll also share performance benchmarks, implementation details, and how the approach handles real-world workloads, including compute-storage decoupled deployments. In one demonstrated benchmark, incremental merge reduced query time from 409 ms to just 18 ms, about 23× faster than full recomputation. A compelling look at how smarter cache reuse can help keep real-time analytics responsive as data grows. #apachedoris