We had a query about joining orders and users that suddenly became slow as data grew. Nothing changed in the code, but latency doubled. After checking EXPLAIN ANALYZE, I noticed Postgres switched from a Hash Join to a Merge Join. Why it mattered Hash Join builds an in-memory lookup fast for smaller, unsorted data Merge Join works on sorted/indexed data better for large datasets In our case, adding the right index allowed Postgres to use Merge Join efficiently, and the query time dropped significantly. Same query. Different join strategy. Huge impact. Don’t just write JOINs. Understand how the database executes them. #PostgreSQL #DatabaseInternals #BackendEngineering #SystemDesign

We just transformed our entire backend from MongoDB to MySQL in a single day — and honestly, that’s been an incredible experience. This was a full EMR application, and we managed to recreate the same database structure by generating scripts from MongoDB to SQL and running them seamlessly. No tables existed initially, yet everything was set up and working smoothly. What really stood out was using Claude Code agents. The way you prompt them directly impacts how effectively they perform — and seeing that in action was one of the best learning experiences I’ve had. Even more impressive, the agents tested the backend across both MongoDB and MySQL simultaneously. A true multi-agent workflow handling different sections and validating the entire flow end-to-end. All of this — in just one day. Still processing how powerful this is.

DuckDB's DuckLake 1.0 has dropped yesterday. Lots of cool stuff and improvements. As it turns out, I haven't been paying close enough attention to how the project was evolving. With much less noise surrounding it after the pretty hyped launch, I thought it has quietly shrunk to just a fun toy bird people did. I was very wrong. Before getting into what 1.0 brings to the table, let's have a second to appreciate that Apache Spark can work with Ducklake. Remember when people sarcastically wrote in comments that the only thing supporting Ducklake is Duckdb? I 'member. Datafusion also supports it now, too. Neat! More importantly, 1.0 codifies team's effort to remain Iceberg-compatible on data level. There's even possibility to shallow-copy migrate tables from Ducklake to Iceberg and vice versa. Me, personally, I always thought that it'd be easier and more proper (?) to use views to mimic Iceberg meta structure directly in Ducklake catalog, and contribute directly to engines a Trojan Iceberg spec - the one that will actually do SQL lookups instead of GET on S3. Anyway, I dig the direction team has taken. But maybe someone can tell me why they have decided not to do it that way? But what really made me warm inside is the inlining. I think the first time I saw the implementation of this (or, rather something similar), was the RisingWave's streaming to Iceberg through Postgres-as-a-buffer. In both cases the concept is this: instead of creating a new Parquet file for each new insert or update, catalog database will store that data inside a conventional table until that data is flushed. Flushing writes a single Parquet file for the entirety of the batch. The same goes for deletions. And you can configure your buffer size to better fit the ingestion pattern. This helps out with small data files problem, if tuned accordingly. Future of approachable lakehouse is looking bright!

Learn what graphs in Postgres and Ticket to Ride have in common when you watch Ashutosh Bapat of Microsoft present on: "Exploring property graphs with SQL/PGQ in PostgreSQL". Beyond the Ticket to Ride game, the talk will give you an under-the-hood look at the SQL/PGQ implementation in Postgres, including the transformation of graph patterns into relational join trees and the complexities of path-variable bindings. You will leave with a clear understanding of the Postgres implementation of the SQL/PGQ standard from one of the authors of the Postgres patch (which is targeting PG19). When you join the POSETTE Livestream, be sure to join the virtual hallway track on Discord, where you can ask speakers questions while their talks are livestreamed. Instructions on joining the hallway track in the “Add to calendar” links on PosetteConf.com. Join POSETTE: An Event for Postgres 2026, Livestream 4 on Thu 18 Jun https://lnkd.in/gSxATjZj #PosetteConf

Should you reference or embed your data in MongoDB? That's an important question for many developers. This short video teaches you the answer from scratch using a simple pizza analogy! 🍕 Master the "Golden Rule" of MongoDB, and start building. https://lnkd.in/e_HJhixe

The "Golden Rule" of MongoDB is simple: Data that is accessed together should be stored together. Unlike relational databases that rely on normalization and complex JOINs, MongoDB gives you the flexibility to choose between embedding and referencing based on your application's specific workload and access patterns. Learn how ⬇️

The "Golden Rule" of MongoDB is simple: Data that is accessed together should be stored together. Unlike relational databases that rely on normalization and complex JOINs, MongoDB gives you the flexibility to choose between embedding and referencing based on your application's specific workload and access patterns. Learn how ⬇️

The "Golden Rule" of MongoDB is simple: Data that is accessed together should be stored together. Unlike relational databases that rely on normalization and complex JOINs, MongoDB gives you the flexibility to choose between embedding and referencing based on your application's specific workload and access patterns. Learn how ⬇️

The "Golden Rule" of MongoDB is simple: Data that is accessed together should be stored together. Unlike relational databases that rely on normalization and complex JOINs, MongoDB gives you the flexibility to choose between embedding and referencing based on your application's specific workload and access patterns. Learn how ⬇️

The "Golden Rule" of MongoDB is simple: Data that is accessed together should be stored together. Unlike relational databases that rely on normalization and complex JOINs, MongoDB gives you the flexibility to choose between embedding and referencing based on your application's specific workload and access patterns. Learn how ⬇️