Today's SQL or Death Seminar Speaker: Monty Widenius (creator of MySQL + MariaDB + MaxDB) will present @mariadb_org's new query optimizer architecture. Zoom talk open to public at 4:30pm ET. YouTube video available after: https://db.cs.cmu.edu/events/sql-death-mariadbs-new-query-optimizer/

Today's SQL or Death Seminar Speaker: Jeff Shute from Google will present their Pipe syntax enhancements for SQL. Zoom talk open to public at 4:30pm ET. YouTube video available after: https://db.cs.cmu.edu/events/sql-death-pipe-syntax-in-sql/

Today's SQL or Death Seminar Speaker: Yash Mayya + Gonzalo Ortiz from StarTree.ai will discuss how to support multiple query languages at the same time in a query optimizer with Apache Pinot. Zoom talk open to public at 4:30pm ET. YouTube video available after: https://db.cs.cmu.edu/events/sql-death-apache-pinot-query-optimizer/

Today's SQL or Death Seminar Speaker: The Germans (Viktor Leis + Thomas Neumann) from @tu_muenchen want to replace SQL with Sane Intermediate Representation (SaneIR) for relational query execution. Zoom talk open to public at 4:30pm ET. YouTube video available after: https://db.cs.cmu.edu/events/sql-death-towards-sanity-in-query-languages/

Do you hate SQL and wish it would die & burn in hell? Or do you love SQL and wish it ran faster? If you answered 'yes' to either question then join our Spring 2025 @CMUDB Seminar Series: SQL or Death?
Mondays @ 4:30pm via Zoom.
Videos posted to YouTube: https://db.cs.cmu.edu/seminar2025/

Seminar Schedule:
Feb 10: Convex
Feb 17: The Germans (TUM)
Feb 24: Apache Pinot
Mar 03: Malloy
Mar 10: Google SQL Pipes
Mar 24: PRQL
Mar 31: StarRocks
Apr 07: Oxide OxQL
Apr 14: MariaDB
Apr 21: EdgeDB

Thanks to @andy_pavlo for pointing out the only known free query optimizer textbook:

https://www.microsoft.com/en-us/research/uploads/prod/2024/12/Extensible-Query-Optimizers-in-Practice.pdf

New @CMUDB Course: Database Query Optimization
This is a special topics course on how to build a SQL optimizer from scratch covering foundational and state-of-the-art implementations. All lectures available on Youtube: https://15799.courses.cs.cmu.edu/spring2025/

One topic we will discuss is the Cascades optimizer architecture. There has never been a good description of how to implement it but the Microsoft Research Database Group just published a book that describes SQL Server's implementation. Microsoft has made the entire book available for free: https://www.microsoft.com/en-us/research/publication/extensible-query-optimizers-in-practice/

Query optimization is the hardest topic in databases and this is the first time I am offering this course. I am going to make mistakes and say incorrect things in my lectures. Send corrections to db-mistakes@cs.cmu.edu

We're banging into the new year with my annual retrospective of the last year in databases! Highlights include license change blowbacks, Databricks vs. Snowflake gangwar, DuckDB's shotgun weddings, and how to buy a quarterback with database money to show that special somebody in your life that you're thinking of them! https://www.cs.cmu.edu/~pavlo/blog/2025/01/2024-databases-retrospective.html

Today's Database Building Blocks Seminar Speaker: Ruihang Xia will present the GreptimeDB time-series DBMS architecture. Zoom talk open to public at 4:30pm ET. YouTube video available after: https://db.cs.cmu.edu/events/building-blocks-greptimedb-ruihang-xia/

▶️ This talk is the makeup of the earlier cancelled event from Nov 25th.

My second blog post on Aurora DSQL, with a deep dive on how reads work: https://brooker.co.za/blog/2024/12/04/inside-dsql.html

Alexey Milovidov's recent ClickHouse talk at CWI is hilariously off-the-chain. I highly recommend watching it: https://www.youtube.com/watch?v=jmVxfGEN0QQ

You can hear me yelling at him from the audience at 1:09, 2:40, and 7:33.

The video for my "What Goes Around Comes Around... And Around" talk at CWI is now available: https://www.youtube.com/watch?v=8Woy5I511L8

📊Slides: https://www.cs.cmu.edu/~pavlo/slides/whatgoesaround2024.pdf
📄Paper: https://db.cs.cmu.edu/papers/2024/whatgoesaround-sigmodrec2024.pdf

Today's Database Building Blocks Seminar Speaker: Xuanwo will present the Apache OpenDAL framework for integrating object stores in database systems. Zoom talk open to public at 4:30pm ET. YouTube video available after: https://db.cs.cmu.edu/events/building-blocks-apache-opendal-databend-xuanwo/

Today's Database Building Blocks Seminar Speaker: Ruihang Xia will present the GreptimeDB time-series DBMS architecture. Zoom talk open to public at 4:30pm ET. YouTube video available after: https://db.cs.cmu.edu/events/building-blocks-greptimedb-ruihang-xia/

@CMUDB Correction: GlareDB is moving *away* from DataFusion! Their fascinating talk discusses the problems with building a DBMS using off-shelf parts. Like @duckdb, the new GlareDB rewrite borrows ideas from the Germans' HyPer system but it's written in Rust: https://www.youtube.com/watch?v=Sor3KZpmbHg&list=PLSE8ODhjZXjZc2AdXq_Lc1JS62R48UX2L&index=8

Today's Database Building Blocks Seminar Speaker: Sean Smith (Founder) will present the ups and downs in his journey of rewriting GlareDB to use Apache DataFusion. Zoom talk open to public at 4:30pm ET. YouTube video available after: https://db.cs.cmu.edu/events/building-blocks-glaredb-sean-smith/

Back at the CIT on Brown campus this past weekend. Some things changed over the past 10 years, some stayed the same. Glad to see the memorial marker for @andy_pavlo is still standing, in spite of campus protests last year.

Even though Newton's laws are deterministic, the behavior of many interacting bodies is so chaotic that it looks essentially "random."

Statistical mechanics effectively says: why bother with all those complex trajectories? Just go ahead and replace them with truly random motion.

Today's Database Building Blocks Seminar Speaker: Paul Dix (CTO + Founder) will present the third iteration of InfluxDB's architecture based on DataFusion + Arrow. Zoom talk open to public at 4:30pm ET. YouTube video available after: https://db.cs.cmu.edu/events/building-blocks-building-influxdb-3-0-with-the-fdap-stack-apache-flight-datafusion-arrow-and-parquet-paul-dix/

@norootcause @andy_pavlo
1) append-only storage to support jukeboxes of magneto-optical WORM discs, the hot storage technology of the early 90s
2) time-traveling queries (removed to reduce overhead)
3) recovery without needing write-ahead logging (it later turned out that WAL was faster and more correct)

Concurrency control was done originally using 2PL, and because that was not an interesting research problem, it only supported table-granularity locks, which was, of course, a performance nightmare.

Years later, when we were adding serializable isolation, we faced a lot of opposition from people who had trauma from the 2PL days and assumed any implementation that required any blocking would be that bad.