Building a Unified High‑Speed Analytics Platform with StarRocks at Cross‑Express

Introduction: Cross‑Express, a large modern logistics company, faces massive data volume and real‑time requirements, with over 5 million daily tickets and more than 10 million API calls, demanding sub‑second response times.

Platform evolution: Initially used MySQL, then added Presto, ES, Impala+Kudu, TiDB, and ClickHouse, leading to multiple engines and complex BI development.

Engine selection: In 2021 the team benchmarked several engines on identical hardware and data, finding StarRocks superior in query performance and real‑time updates.

Engine convergence: The team retained ES for detailed queries and StarRocks as the primary engine, citing AP query performance, MySQL compatibility, primary‑key updates, federated external tables, low maintenance, and integrated data loading tools.

Benefits: Query latency dropped to milliseconds, daily API calls exceed 6 million, BI development time reduced, ETL eliminated, overall performance up 20%, and hardware usage cut from 15 nodes to 4.

Real‑time data warehouse: Replaced a 2‑hour batch pipeline with a Flink‑StarRocks pipeline, achieving sub‑5‑second end‑to‑end latency and 300 % faster queries compared with Presto; the primary‑key model improved performance another 200 %.

Multi‑scenario applications: Core business cost‑optimization uses StarRocks to compute optimal routes, reducing per‑order calculation from 90 seconds to 4 seconds and increasing throughput by 400 %.

Summary and outlook: StarRocks enabled convergence of five‑six legacy engines, delivering real‑time analytics, cost reduction, and plans to explore resource isolation, near‑real‑time solutions, and version unification.