One of the ideas we wanted to explore more has been speeding up in-memory aggregations in PostgreSQL through vectorized execution. The opportunity to do so came up when we had our intern, Can, take this on as a project during his summer internship. The early numbers he has there are promising - suggesting a 3-4x increase in PostgreSQL performance for simple SELECT queries with sum/count/group by operations.
This is proof-of-concept work conducted within several weeks and not with the production-ready diligence that we always have on our projects - hence we made a very explicit point to add "_test" to the end of the project name. That said, it shows good promise for further increases in performance, and given the ideas there can be useful more broadly, we are happy to open source the project.
Please take a look at postgres_vectorization_test on GitHub, and let us know what you think! The readme also has more details on our approach, sample queries, performance comparisons, and instructions on getting it set up as a PostgreSQL extension: https://github.com/citusdata/postgres_vectorization_test
With this, my special thanks go to our intern, Can, for tackling this in the short amount of time he had, to Metin for mentoring Can, and of course, to Ozgun, for pulling everything together for an exciting summer project.
We are excited to announce the release of cstore_fdw 1.1, Citus Data's open source columnar store extension for PostgreSQL. The changes in this release include:
For installation and update instructions, please see cstore_fdw’s page in GitHub.
To learn more about what’s coming up for cstore_fdw see our development roadmap.
If you have questions about cstore_fdw, please contact us using the cstore-users Google group.
If you discover an issue when using cstore_fdw, please submit it to cstore_fdw’s issue tracker on GitHub.