Andy Pavlo joins ClickHouse to establish ClickHouse Labs

Andy Pavlo 加入 ClickHouse，成立 ClickHouse Labs

I am excited to announce that I am joining ClickHouse to establish and lead a new research team called ClickHouse Labs. I want to share how it came about and what we plan to do. 我很高兴地宣布，我将加入 ClickHouse，成立并领导一个名为 ClickHouse Labs 的新研究团队。我想分享一下这件事的由来以及我们未来的计划。

How It Started

缘起

I started as a professor in the Computer Science Department at Carnegie Mellon University in 2013. I have spent my career seeking to understand the science of modern database management system (DBMS) internals. I make it a priority to track every new system that comes along, both in industry and academia, to understand their implementations. 我于 2013 年开始在卡内基梅隆大学计算机科学系担任教授。我的职业生涯一直致力于探索现代数据库管理系统（DBMS）的内部科学。我始终将追踪工业界和学术界出现的每一个新系统作为首要任务，以深入了解它们的实现方式。

I have known about the ClickHouse DBMS since it was first announced as open-source software in June 2016. My initial reaction to this news was that it had to be vaporware because it seemed too good to be true. ClickHouse had features that at the time were only found in a handful of closed-source, commercial analytical DBMSs. For example, ClickHouse was written in C++ and supported vectorized query execution using SIMD in 2016. Most prominent open-source analytical DBMSs in 2016 were JVM-based and did not support SIMD optimizations until years later. 自 2016 年 6 月 ClickHouse DBMS 作为开源软件首次发布以来，我就一直关注它。我最初的反应是这一定是“雾件”（vaporware），因为它看起来好得令人难以置信。ClickHouse 当时具备的功能在当时仅存在于少数闭源的商业分析型 DBMS 中。例如，ClickHouse 使用 C++ 编写，并在 2016 年就支持使用 SIMD 进行向量化查询执行。而 2016 年大多数主流的开源分析型 DBMS 都是基于 JVM 的，直到多年后才支持 SIMD 优化。

Since then, I have followed ClickHouse’s development closely. It has always been a leading system that was highly relevant to our academic research projects. You can even see me wearing my original ClickHouse shirt in my first remote lectures in 2020, when the pandemic forced us to move our database courses online. Given this history, I was honored when the ClickHouse co-founders invited me to establish this new research group at ClickHouse. The chance to work with one of the strongest engineering teams on the next generation of database technology was an opportunity that I could not pass up. This will be a next-level collaboration like when Killer Mike hooked up with El-P to create a hip-hop supergroup. 从那时起，我一直密切关注 ClickHouse 的发展。它始终是一个与我们的学术研究项目高度相关的领先系统。在 2020 年疫情迫使我们将数据库课程转为线上时，你甚至可以在我最初的远程讲座中看到我穿着那件原始的 ClickHouse T 恤。鉴于这段渊源，当 ClickHouse 的联合创始人邀请我建立这个新的研究小组时，我感到非常荣幸。能与最强大的工程团队之一共同致力于下一代数据库技术，这是一个我无法错过的机会。这将是一次跨越式的合作，就像 Killer Mike 与 El-P 联手打造嘻哈超级组合一样。

What Is ClickHouse Labs?

什么是 ClickHouse Labs？

The goal of ClickHouse Labs is to establish a best-in-class industry research organization focused on databases. It will not operate as an isolated research organization that throws ideas over the wall to engineering. Instead, we will work closely with ClickHouse engineers, customers, collaborators, and industry partners to develop and disseminate new ideas that keep ClickHouse at the bleeding edge. ClickHouse Labs 的目标是建立一个专注于数据库的一流工业研究机构。它不会作为一个孤立的研究组织，仅仅将想法“扔给”工程部门。相反，我们将与 ClickHouse 的工程师、客户、合作者和行业伙伴紧密合作，开发并传播新的理念，使 ClickHouse 始终保持在技术的最前沿。

We will also work with ClickHouse’s PostgreSQL team to help establish its burgeoning managed service as a market leader in performance and reliability. PostgreSQL and ClickHouse serve different workload requirements, but the combination gives us a broad foundation for investigating both transactional and analytical database problems. Our objective is straightforward but ambitious: conduct research with scientific value and then help transform the best ideas into technology that matters to users. 我们还将与 ClickHouse 的 PostgreSQL 团队合作，帮助其蓬勃发展的托管服务在性能和可靠性方面成为市场领导者。PostgreSQL 和 ClickHouse 服务于不同的工作负载需求，但两者的结合为我们研究事务型和分析型数据库问题提供了广泛的基础。我们的目标简单但雄心勃勃：进行具有科学价值的研究，然后帮助将最好的想法转化为对用户至关重要的技术。

I want to achieve the same level of impact associated with pioneering industry research organizations, such as IBM Research and Microsoft Research. Those groups demonstrated that industry laboratories can simultaneously advance fundamental computer science, influence commercial products, and train generations of database researchers. That is the tradition we want to continue. 我希望实现与 IBM Research 和 Microsoft Research 等开创性工业研究机构同等水平的影响力。这些组织证明了工业实验室可以同时推动基础计算机科学的发展、影响商业产品，并培养一代又一代的数据库研究人员。这就是我们想要延续的传统。

What is Next?

未来展望

The ClickHouse team already has an exceptional record of publishing deep technical material about its work. Since the establishment of the company in 2021, its engineers have produced detailed articles that explain the DBMS’s implementation. There is also the 2024 VLDB paper that describes ClickHouse’s core architecture. These works are so thorough that I assign them as readings to my students at Carnegie Mellon. ClickHouse 团队在发布关于其工作的深度技术资料方面已经拥有卓越的记录。自 2021 年公司成立以来，其工程师们撰写了大量详细的文章来解释该 DBMS 的实现。此外，2024 年的 VLDB 论文也描述了 ClickHouse 的核心架构。这些作品非常详尽，我甚至将它们指定为我在卡内基梅隆大学学生的阅读材料。

At the same time, there is a backlog of interesting ideas and optimizations that the ClickHouse engineering team has explored but has not yet had the time to validate fully and push into production. One of my immediate priorities is to help accelerate this process. We will then use that as a springboard to explore new ideas that push ClickHouse even further. 与此同时，ClickHouse 工程团队已经探索了许多有趣的想法和优化方案，但尚未有时间进行充分验证并投入生产。我当下的首要任务之一就是帮助加速这一进程。随后，我们将以此为跳板，探索能将 ClickHouse 推向更高水平的新想法。

One larger question we will investigate is how DBMSs like ClickHouse and PostgreSQL fit into emerging AI and agentic technologies. There are two sides to this problem. The first is determining what a DBMS should look like to better support agents. The second side is determining how agents can improve and automate the development of DBMSs themselves. Everything is on the table: new hardware, new algorithms, new data structures, new execution strategies, and new ways of building and operating DBMS software. 我们将研究的一个更大的课题是，像 ClickHouse 和 PostgreSQL 这样的 DBMS 如何适应新兴的 AI 和智能体（agentic）技术。这个问题有两个方面：首先是确定 DBMS 应该是什么样子，才能更好地支持智能体；其次是确定智能体如何改进并自动化 DBMS 本身的开发。一切皆有可能：新硬件、新算法、新数据结构、新执行策略，以及构建和运行 DBMS 软件的新方法。

Although I do not have answers to these problems yet (this is why it is research), the one thing I am certain about is that ClickHouse’s solid relational model foundation positions it well to evolve alongside these data-intensive workloads. I have spent my career studying how database systems are built and helping train the people who build them. With ClickHouse Labs, we now have the opportunity to create an organization devoted to advancing both. 虽然我目前还没有这些问题的答案（这就是为什么它被称为研究），但我确定的一点是，ClickHouse 坚实的关系模型基础使其能够很好地适应这些数据密集型工作负载并随之演进。我的职业生涯一直致力于研究数据库系统是如何构建的，并帮助培养构建它们的人才。有了 ClickHouse Labs，我们现在有机会创建一个致力于推动这两者发展的组织。