pg_lake just went open source! (Apache 2.0) pg_lake is a set of extensions (from Crunchy Data Warehouse) that add comprehensive Iceberg support and data lake access to Postgres, with @duckdb.org transparently integrated into the query engine. Announcement blog: www.snowflake.com/en/engineeri...

No system hits the sweet spot of allowing for extensibility while maintaining systems safety. It would be nice if there was a standard plugin API (think POSIX) that allows compatibility across systems. Thanks to @marcoslot.com + @daveandersen.bsky.social for their collaboration on this project

At last @abigalekim.bsky.social's paper is out! Its the most complete eval of DB extensions/plugins ever. We analyze PostgreSQL, MySQL, MariaDB, SQLite, DuckDB, Redis. TLDR: Postgres extns ecosystem is fraught with footguns. Other DBMSs have fewer extns but less problems. DuckDB has cleanest API.

Five years ago I joined @crunchydata.com, shortly after I wrote about having unfinished business with Postgres. Today as part of Snowflake that journey is continuing. We've built some amazing things, but are just getting started. www.crunchydata.com/blog/crunchy...

Recording of my Data Council talk: www.youtube.com/watch?v=HZAr...

Generative AI comes up with details that would be hilarious, if it wasn't so mind boggling that it can come up with these details.

And there it is: Native logical replication from any Postgres server to Iceberg managed by Crunchy Data Warehouse. Speed up Postgres analytical queries 100x with 2 commands.

I gave a talk at the inaugural (and awesome) European Iceberg meetup in Amsterdam last night. It's an introduction to how and why we used Iceberg and DuckDB to build a Postgres Data Warehouse: www.youtube.com/watch?v=cEnq...

Move fast and build solid solutions that work across platforms. You can now use Postgres as a modern Data Warehouse anywhere, using any S3-compatible storage API. Query, import, or export files in your data lake or store data in Iceberg with automatic maintenance and very fast queries.

Excited to announce Crunchy Data Warehouse is now available for Kubernetes and On-premises. Need faster analytics from Postgres? Want a native Postgres data lake experience? Learn more about how it works: www.crunchydata.com/blog/crunchy...

Amazing result

We weren't really thinking of log management as a target use case, but Iceberg is ideal as the final destination for logs, and having transactions & built-in job scheduling & a fast query engine (& laser focus on developer experience) makes things really simple and cost-effective.

I got a number of questions on how we saved $30k a month on cloudwatch by moving logs directly to S3/Iceberg with Postgres so I wrote up how in a bit more detail - www.crunchydata.com/blog/reducin...

Excited to announce built-in maintenance for Iceberg via Postgres. Now within Crunchy Data Warehouse we will automatically vacuum and continuously optimize your Iceberg data by compacting and cleaning up files. Dig into the details of how this works www.crunchydata.com/blog/automat...

Imagine your potential customer as a serious company doing serious things, and willing to pay serious money if you can genuinely help them run their business without causing lot of new problems. Then go build products for that customer. This works.

Auto-vacuum for #Iceberg tables is now available in Crunchy Data Warehouse! We're always aiming for a 0-touch experience where possible, so we went out of our way to make Iceberg compaction & cleanup fully automatic without any configuration. Still pretty interesting to see a manual vacuum:

A big part of building Crunchy Data Warehouse was ease of use. How easy is it to load data from existing public datasets? Step 1: Point at your dataset and we'll load it for you Step 2: Query it Step 3: Profit

ChatGPT Plus had a good run, but looks like Le Chat is going to be my main assistant now. I like that it's fast, to the point, and quite clever. I was impressed with a SQL query it came up with today for finding contiguous ranges of integers. ChatGPT's version was 3x slower.

Postgres is increasingly becoming a versatile data platform, instead of just an operational database. Using pg_parquet you can trivially export data to S3, and using Crunchy Data Warehouse you can just as easily query or import Parquet files from PostgreSQL.

Deepseek R1 in an ollama "container app" on a managed Postgres server, because... why not?

5 years from now, no one's going to want slower, less reliable, or harder to use databases.

🎉 pg_documentdb is open source I created the initial version with Vinod Sridharan (an absolutely brilliant engineer) at Microsoft a few years ago and it's come a long way since. It reimplements Mongo API with exact semantics in PostgreSQL. Already used by FerretDB! github.com/microsoft/do...

Impressed by the latest ParadeDB release. Solving the right problems in the right way is really hard.

1/11. ParadeDB is now integrated with Postgres block storage. As far as we know, no one has integrated a search and analytics engine with Postgres storage before. This is a big deal. Here's why we did it, how we did it, and why you should care. 🧵

A lot of great recommendations on tuning PostgreSQL for analytical queries by @karenhjex.bsky.social www.crunchydata.com/blog/postgre...

If you've become your company's designated Postgres expert, learning to navigate Postgres source code can be a useful skill, even if you're not a C programmer. For instance, the docs do not specify how the optimizer works, but there is a detailed readme in the source: github.com/postgres/pos...

Buckle up because we're banging into the new year with my annual retrospective of the last year in databases! Highlights include license change blowback, Databricks vs. Snowflake gangwar, @duckdb.org's shotgun weddings, and buying a quarterback to impress your lover: www.cs.cmu.edu/~pavlo/blog/...

In the piece I cover: - the partnership between @cncf.io and Andela - the release of Akamai App Platform cc/ @mikemaney.bsky.social - the rebranding of Akka - the partnership between @gitlab.com and AWS - the release of Crunchy Data Warehouse cc/ @craigkerstiens.com @crunchydata.com

Woot, woot, #Debezium 3.0.5 is out, which amongst other things also contains my patch for adding support for #Postgres 17 failover slots 🥳. You now can fail over your Debezium CDC connector to a replica turned primary, without missing any events! debezium.io/blog/2024/12...

Well well well: www.crunchydata.com/blog/pg_incr... Incremental pipelines come to Postgres via Crunchy Data! This is like "dbt incremental", not true incremental view maintenance like @materialize.com or Snowflake's dynamic tables, but it's a neat step towards IVM.

We just launched tldraw computer

End-to-end demo of the new pg_incremental extension. There's raw events table and a summary table containing view counts. You then define a pipeline using an insert..select command, and keeps running that to do fast, reliable, incremental processing in the background.

There are many incremental processing solutions, but they seem to never quite do what I need. I decided to build an extension that just keeps running the same command in Postgres with different parameters to do fast, reliable incremental data processing. That's pg_incremental. 1/n

how do you all remember every UUID? I find it really hard. so I wrote them all down on every uuid dot com the list has fast search across all 2^122 values (so you can find your favorites) - hoping to add some social features like "trending UUIDs" soon!

Everything gets simpler once you have transactions on Iceberg tables. A useful pattern is to keep track of loaded files in a Postgres table the same transaction that loads the file into Iceberg, such that each file is loaded exactly once www.crunchydata.com/blog/iceberg...

The transactional piece is the real game changer, especially for data pipelines You can write these little procedures that load data from S3 or http in many different formats (using plain COPY) and store the fact that you loaded the file in the same transaction. Super simple, reliable, repeatable.

Want a complete Iceberg + transactional database system in one unified experience? (Postgres) Well we shipped just that last month. The ease of use, yet power underneath really is a game changer for data systems - www.crunchydata.com/blog/crunchy...

Exciting announcement. Played with this a bit today. It's like an Iceberg starter kit that bundles storage with the catalog and compaction. Systems looking to support Iceberg writes will have a relatively easy time targeting S3 tables catalog, because they don't need to add compaction. 1/n

It would be nice if databases had a more convenient way of saying "columns A and B are coupled and unique, one implies the other" and adjusted SQL syntax to it. now you need to always either group by both ID and email or do fake aggregations like any_value(email), which seems like an anti-pattern.

Can't wait to put my new decoration on my Christmas tree (when my husband isn't looking)! #PostgreSQL #crafting

We hosted our annual(ish) Databases Dinner in SF last week! There were 3 recurring themes among database practitioners: 1) The need for an alternative to Spark, 2) The desire for an alternative to SQL, and 3) Lots of praise for Postgres’ scalability and workloads moving to it.

In case you missed it last week, we released Crunchy Data Warehouse - transforming Postgres into a best in class data warehouse with full Iceberg support - www.crunchydata.com/blog/crunchy...

CrunchyData Warehouse all looks very impressive: www.crunchydata.com/blog/crunchy.... Impressed to see them build on industry primitives like Apache Iceberg, and materializing PostgreSQL data marts so seamlessly. Super interested to see how this works with fast moving datasets.

At cwi.nl/en/events/dijkstra-awards/cwi-lectures-dijkstra-fellowship we awarded the Dijkstra Fellowship to Marcin Zukowski for his contributions to DB architecture in a full room & great speakers: Allison Lee, Viktor Leis, @andypavlo.bsky.social @hannes.muehleisen.org @madelonhulsebos.bsky.social

The Database Capital of Europe

The promise of Apache Iceberg. It's like pong for databases. #databs

Some interesting things can be built with this. Next up, an efficient way to poll for new appends? 🤔

It did not make sense to me why Q28 from ClickBench was slower on Crunchy than on other systems, but we did not have enough time to thoroughly profile it at the time. Kept bugging me though. Turns out the problem was that we were running the actual benchmark, but almost no one else was 😅

I noticed that Q28 of ClickBench ran much slower when I ran it manually, than when running it with the provided benchmark scripts. Turned out Q28 was run incorrectly for pretty much every system measured by ClickBench: github.com/ClickHouse/C...