In the PB-scale data era, the separation of data lakes and data warehouses often makes enterprises use multiple different systems together, which leads to high costs, data silos, inconsistent data standards, and inefficient operation and maintenance. Apache Doris 3.x’s improvements in lakehouse capabilities and its compute-storage separation architecture provide a way to solve this problem. This session shares how Xiaomi uses these features to replace different old engines and build a unified data analysis platform. We’ll focus on key points like: an overview of Apache Doris’ core lakehouse features, how it’s used and the challenges we faced at Xiaomi, the pain points of advertising business and how Apache Doris helped upgrade its architecture, multi-source federated query and Catalog management, unified semantic models for BI scenarios, and how we deploy Doris on Kubernetes, along with elastic scaling and observability practices. We’ll also share our future development plans. This session is designed to give a reference to enterprises facing similar architecture integration challenges, and help attendees learn practical experience in building a unified, efficient and stable lakehouse architecture with Apache Doris.

Speakers:

Congling Xia: Senior Software Engineer at Xiaomi, Contributor of Apache Kylin, Trino and Apache Doris

He is a Senior Software Engineer at Xiaomi’s Data Platform. He has years of experience in big data development and the operation & maintenance of various OLAP engines. His core focus lies in enterprise-grade OLAP engine engineering and service-oriented platform capability building. He is also an open source contributor to Apache Doris, Trino, Apache Kylin, with solid hands-on experience in lakehouse architecture, unified computing and cloud-native data operations.