JD Advertising built a massive ad‑data warehouse on Apache Doris, reaching nearly 1 PB and 18 trillion rows, then implemented a hot‑cold data tiering strategy—first a lake‑based approach, later a native tiering solution in Doris 2.0—reducing storage costs by 87% and boosting query performance over tenfold.
JD Advertising built a petabyte‑scale ad analytics service on Apache Doris, identified a hot‑cold access pattern, and implemented a native cold‑hot tiering solution (upgrading to Doris 2.0 and optimizing schema changes) that cut storage costs by ~87% and boosted concurrent query capacity over tenfold while simplifying operations.
This article presents JD Advertising's engineering experience with Apache Doris, describing the evolution from a data‑lake cold‑data solution to a native cold‑hot tiering approach, detailing performance regressions after upgrading to Doris 2.0, and outlining a series of optimizations for query speed, CPU and memory usage, schema‑change efficiency, and automated data migration and restoration.
