Interview with Ma Jin on Distributed Database DDB Performance Optimization and Architecture

From November 18‑20, 2016, the CSDN‑organized SDCC 2016 China Software Developers Conference was held in Beijing, gathering senior technical professionals to share cutting‑edge technologies and practical experiences.

During the event, we interviewed Ma Jin, a senior development engineer at NetEase Hangzhou Research Institute, who is responsible for the distributed database DDB and related projects.

CSDN: Please introduce yourself and the technologies you focus on.

Ma Jin: I joined NetEase Hangzhou Research Institute in 2013 and now lead the DDB project, also working on cache NKV, distributed transaction scheduler TCC, and other distributed systems, with a strong interest in distributed and storage technologies.

CSDN: What attracted you to NetEase and any memorable experiences?

Ma Jin: Visiting the NetEase campus before graduation, I was impressed by the environment and decided to stay, appreciating the ability to focus deeply on a single technology without departmental barriers.

CSDN: What technical difficulties did you encounter during DDB development and how did you overcome them?

Ma Jin: DDB has been online for about seven years, evolving through major shifts in internet technology. Adding new features like alias support required modifying hand‑written parsers tightly coupled with execution plans, which was risky and time‑consuming. We also faced version management challenges, leading to the creation of a Proxy layer to provide non‑intrusive, version‑controlled access.

CSDN: Where is DDB used in the system architecture and what responsibilities does it have?

Ma Jin: DDB serves as a data‑service layer, providing transparent sharding, read/write separation, and high availability. It also stores metadata for large systems such as NetEase's object storage NOS.

CSDN: Why did NetEase invest in developing a sharding database like DDB?

Ma Jin: DDB originated to support NetEase Blog in 2006 when MySQL could not handle the traffic. As user volume grew to tens of millions, a sharding middleware remained essential for large‑scale products.

CSDN: How did NetEase's private‑cloud project affect DDB?

Ma Jin: The private‑cloud initiative promoted the DDB Proxy model, enabling easier integration with various languages (Java, Python, Go) and pushing DDB toward a platform‑centric architecture.

CSDN: What are your views on the future of distributed databases?

Ma Jin: While NoSQL and NewSQL offer performance advantages, they often sacrifice flexibility. Traditional RDBMS‑based sharding solutions like DDB provide a reliable balance of scalability and functionality, and will likely continue to evolve, especially as SaaS offerings on cloud platforms.

CSDN: What will you focus on in your SDCC 2016 Beijing talk?

Ma Jin: I will introduce DDB’s architecture evolution, its value, and share performance‑optimization techniques to inspire the audience.