Alibaba's Real-Time Data Warehouse Practice with Doris: 50 Billion Records Daily

I recently came across the real-time data warehouse practice based on Doris shared by Alibaba Cloud's database team, and it's packed with valuable insights.

Business Scenarios

  • Real-time dashboards for Taobao/Tmall
  • Real-time ROI analysis for ad campaigns
  • Real-time optimization of logistics routes

Daily data volume: 50 billion records. Storage per single table: PB-level.

Key Technical Points

1. Data Layering

ODS (Raw Data) → DWD (Detail Data) → DWS (Summary Data) → ADS (Application Data)

Doris primarily handles query workloads at the DWS and ADS layers, while the ODS/DWD layers use Flink for real-time ETL.

2. Partitioning and Bucketing Strategy

  • Partition by day, retaining 30 days of hot data
  • Choose high-cardinality fields (e.g., user_id) as the bucketing key
  • Keep each bucket at 1-3GB to ensure query parallelism

3. Materialized Views for Acceleration

Create materialized views for high-frequency aggregation queries (e.g., "GMV in the past hour"), reducing query latency from 3s to 50ms.

Takeaways for Learners

Doris is gaining significant traction in China. Compared to ClickHouse, Doris offers better standard SQL compatibility and operational friendliness. If you're looking to pursue a career in data infrastructure, Doris is definitely worth learning.


References:

About Zihao Zhang

Data Platform Engineer. Distributed systems, OLAP databases, AI Agent development.

Comments

Comments are closed.

Ask Me Anything
Hey! I'm Hank's digital avatar. How'd you find your way here?
⚠️ AI-powered · May be inaccurate · Powered by DeepSeek
Chat Logs