Phil Bernstein

概要

I am a member of the Data Systems Group (opens in new tab) in Microsoft Research Redmond (opens in new tab). I work on various aspects database systems, mostly related to database engines, transaction processing, and data integration.

I’m currently working on searching for approximate nearest neighbors over large vector databases. My previous project focused on improving database servers using disaggregated cloud resources, such as SmartNICs and disaggregated memory. From 2012-2019, I worked on a distributed systems programming framework, called Orleans (opens in new tab), which was released as open source (opens in new tab) in January, 2015 and is widely used (opens in new tab) inside and outside Microsoft. I gave a keynote about it at DISC 2014 (opens in new tab) (slides (opens in new tab)). I worked on projects to enrich Orleans to be an “actor-oriented database system”, which I described in a keynote at ICDE 2018 (opens in new tab) (slides (opens in new tab)). You can read about the following components: indexing (CIDR 2017 (opens in new tab)), geo-distribution (OOPSLA 2017 (opens in new tab)), and transactions (technical report (opens in new tab)).

My early research was primarily on transaction processing, and after a long hiatus, I resumed working in this area in 2006 as a co-designer of the database engine for SQL Azure (opens in new tab). I then focused on building Hyder (opens in new tab), a prototype transactional indexed-record manager that scales out without partitioning. I co-authored a survey of techniques for multi-master replication (SIGMOD 2013 (opens in new tab)) and two books on transaction processing:

• Principles of Transaction Processing (opens in new tab) (2nd edition), coauthored with Eric Newcomer, was published in June 2009 by Morgan-Kauffman Publishers, a division of Elsevier. Translated into Chinese, Japanese, and Korean.
• Concurrency Control and Recovery in Database Systems (opens in new tab), coauthored with Vassos Hadzilacos and Nathan Goodman, is downloadable for free from here (opens in new tab).

I gave a keynote about the history of transaction processing research at SIGMOD 2025. Here are the slides (opens in new tab).

My other main research interest is data integration. From 2000 – 2011 I led the Model Management Project (opens in new tab), whose goal was to make database systems easier to use for model-driven applications, such as design tools, message translators, and database translators. I also worked on object-to-relational mapping, especially in support of the ADO.NET Entity Framework (opens in new tab). Over the years, this work has been done in close collaboration with Sergey Melnik (Amazon), James Terwilliger, Eli Cortez (Microsoft), Suad Alagic, Alon Halevy (Google), Jayant Madhavan (Google), René Miller (Univ. of Waterloo), Kristin Peter Mork (Noblis), Rachel Pottinger (Univ. of British Columbia), Christoph Quix (Technical Univ. of Aachen), Erhard Rahm (Univ. of Leipzig), Adi Unnithan, and many great interns.

I’ve published many research papers on transaction processing, data integration, and other aspects of database management. You can find a nearly-complete list at the DBLP Computer Science Bibliography (opens in new tab).