Recap of Tongcheng Travel’s 7th Big Data Technology Salon – Talks on StarRocks, Paimon, Iceberg, Data+AI, Vector Retrieval, Real‑Time Computing, and Hotel Ranking

On December 23, Tongcheng Travel hosted its 7th Big Data Technology Salon in Beijing, featuring experts from StarRocks, Baidu, Tencent, Alibaba, JD.com and Tongcheng sharing cutting‑edge lake‑warehouse, real‑time computing, compute‑storage separation, Data+AI and machine‑learning practices.

StarRocks Architecture Evolution – Ding Kai, StarRocks community TSC member, presented the background, design goals, key technologies, performance effects and future roadmap of StarRocks’ compute‑storage separation architecture.

Lakehouse Paradigm with StarRocks and Paimon – Wang Riyu, StarRocks committer, discussed the development of StarRocks+Paimon, scenarios, technical principles, performance benchmarks and future plans, highlighting optimizations from version 1.x to 3.x.

Paimon Lakehouse Practice at Tongcheng Travel – Wu Xiangping, Apache Hudi & Paimon contributor, introduced Tongcheng’s lakehouse system, Paimon architecture, optimization techniques, cost and freshness benefits, and shared troubleshooting tips and future directions.

Million‑scale Real‑time Upsert with Iceberg – Chen Wandong, Tencent Cloud DLC engineer, described the Smart Optimizer service built on Apache Iceberg, productization of upsert, large‑scale case studies, performance improvements and outlook for batch‑stream convergence and AI‑driven data services.

Data+AI Evolution for Smart Travel – Xu Kaisheng, Tongcheng Travel AI team lead, presented travel recommendation business, challenges, model evolution, multi‑task learning and prospects for large‑scale travel models.

Vector Retrieval in AI Scenarios – Ke Fei, Baidu Intelligent Cloud director, explored vector database fundamentals, engineering practices, case studies, and Baidu’s optimizations for large‑model vector search, including indexing and multi‑modal retrieval.

Real‑time Computing Performance Governance at JD Logistics – Kang Qi, Apache Flink & Calcite contributor, shared JD Logistics’ massive real‑time workloads, Flink performance and usability enhancements, reliability measures, and expectations for Paimon + StarRocks 3.x.

Multi‑Task and User Interest Modeling for Hotel Ranking – Zhao Xiaochuan, search algorithm expert, explained hotel ranking scenarios, foundational capabilities, multi‑task modeling, dynamic weighting, and model consistency across online and offline environments.

The salon emphasized open sharing of big‑data infrastructure, lake‑warehouse innovations, AI integration and real‑time analytics, with presentation slides available at https://github.com/TongchengOpenSource/TechSharing.