Case Study: DiDi cuts observability costs and speeds up log queries with ClickHouse

Didi Migrates from Elasticsearch to ClickHouse for a new Generation Log Storage System

DiDi, the global mobility platform serving more than 450 million users, needed a new log storage system that could handle petabyte-scale daily data, high-QPS trace queries, and diverse search patterns while reducing the cost and operational burden of Elasticsearch. ClickHouse was chosen to replace Elasticsearch for log storage and retrieval, using ClickHouse’s high-throughput MergeTree-based architecture to support DiDi’s observability workloads.

ClickHouse implemented a redesigned logging architecture with separate Log and Trace clusters, materialized views, distributed tables, optimized sorting and partitioning, and HDFS hot/cold data separation. The migration scaled to more than 400 physical nodes, reached over 40 GB/s peak write traffic, supported about 15 million queries per day, and cut hardware costs by more than 30%, while improving query speed by roughly 4x compared with Elasticsearch.

DiDi

Yuankai Zhong

DiDi