Best Research Paper, VLDB 2023
We are proud to announce that our paper titled "DBSP: Automatic Incremental View Maintenance for Rich Query Languages" co-authored by two of Feldera co-founders, Mihai Budiu and Leonid Ryzhyk, has won a Best Paper Award at the VLDB 2023 conference.
While it is always nice to see your work recognized, this award holds special significance for our team. This is because the paper it acknowledges describes the mathematical theory that underpins our Feldera Incremental Compute Platform.
When we started working on the technology that became [Feldera][3], we already knew firsthand that computing on continuously changing data was hard. A robust, efficient, and general-purpose query engine could not be created solely through feats of engineering but required a solid theoretical foundation. We set out to build such a foundation, leading to the creation of **DBSP**, a new theory of incremental and streaming computation and the not-so-secret sauce enabling Feldera to achieve:
- Correctness, by providing provably correct algorithms for evaluating queries
over changing data. - Performance, by establishing complexity bounds on various operators and
informing optimizations. - Extensibility: DBSP offers a precise recipe for adding new operators while
preserving the end-to-end correctness of the system.
As practitioners, one of our favorite aspects of the DBSP theory is that it lends itself naturally to an efficient implementation. Our open-source implementation of DBSP in Rust powers the Feldera query engine. It is also useful on its own as a library for computations over changing datasets. We are constantly working on improving the library and welcome both users and contributors.
