Database System Research and Future Trends – Talk by Prof. Zhou Xuan (ECNU)

On October 27, 2020, Meituan’s Top Talk invited Prof. Zhou Xuan, professor and deputy dean of the School of Data Science and Engineering at East China Normal University, to give a presentation titled “Database System Research at ECNU”. The talk covered the evolution of database systems, recent research directions of the ECNU database team, and future trends.

Speaker Biography : Prof. Zhou earned his Ph.D. from the National University of Singapore in 2005, worked at Germany’s L3S and Australia’s CSIRO, taught at Renmin University, and joined ECNU in 2017. His research focuses on database systems and information retrieval, with awards including the National Science and Technology Progress Award (Second Class) and the Ministry of Education’s Science and Technology Progress Award (First Class).

Abstract : The ECNU database group, a leading academic and industrial research team, will share recent ideas and results, including hardware adaptation, distributed transaction processing, HTAP, and modular system design.

00 Introduction : The speaker highlighted the different motivations of academia (theoretical research) and industry (product‑centric development) and argued for joint efforts in database system research.

01 Changes in Database System Forms

1.1 Drivers of Change

Application demand – growing scale, diverse domains, and increased pressure on data management.

Software development models – shift to micro‑services and polyglot persistence, leading to varied database choices.

Hardware platforms – cloud infrastructure, heterogeneous processors, and storage.

1.1.1 Application Demand – illustrated by historical shifts from tape to disk, then to cloud‑scale services.

1.1.2 Software Development Model – micro‑service architectures decouple services and allow different databases (MySQL, MongoDB, Elasticsearch) per service, emphasizing the need for flexible data stores.

1.1.3 Hardware Platform – cloud platforms demand elastic, scalable databases; separation of compute and storage, and emerging hardware (e.g., NVM) affect design.

02 ECNU Database System Research

2.1 Team – Over 20 faculty members, ~10 focusing on database kernels, with a 20‑year history and industry collaborations.

2.2 Research Results

2.2.1 Distributed Transactions : Discussed the debate on the usefulness of distributed transactions, explored techniques to reduce blocking (MVCC, OCC, adaptive validation), decentralized timestamp allocation (Posterior SI), and lock‑violation strategies (DLV) to shorten lock holding time.

2.2.2 Database System Decoupling : Analyzed high coupling in traditional DBMS implementations, proposed Transparent Concurrency Control (TCC) that separates operation‑level and transaction‑level concurrency control, and demonstrated performance close to tightly coupled systems.

2.2.3 New Hardware : Examined non‑volatile memory (Intel Optane) characteristics, its role as a cache between compute and storage nodes in cloud databases, and the integration with RDMA for shared caching.

03 Related Papers : A list of representative publications from the ECNU lab (not exhaustive).

Conclusion : The talk emphasized that future database systems will be more modular, hardware‑aware, and tailored to specific workloads, requiring close collaboration between academia and industry.