Matthias

🎯 Bespoke OLAP makes workload-specific database engines faster [ADRS Blog #21] We feature Bespoke OLAP, a fully autonomous synthesis pipeline that outperforms general-purpose engines like DuckDB while being ~11.78× faster on the main benchmark comparison. Bespoke OLAP uses LLM-guided code generation to build a complete, custom C++ engine for OLAP workloads for just ~$120. ✍️ Read the blog: ucbskyadrs.github.io/blog/bespoke-o… 📖 ADRS Blog Series: ucbskyadrs.github.io 📄 Bespoke OLAP Paper: arxiv.org/abs/2603.02001 👩‍💻 Bespoke OLAP Code: github.com/DataManagement…

@isDineshHere Thanks for the shout-out!

This marks the beginning of "Build your own TigerBeetle" trend, and I'm here for it.

"Rare examples such as @TigerBeetle demonstrate that extreme specialization can pay off when the target workload is narrow yet widely shared" Bespoke OLAP: DBMS on-demand, for your specific workload. Also the fact @DuckDB has pretty much become a standard to benchmark against

@isDineshHere We database people already worked on this. What came out is the bespoke olap paper, which is from my lab! Btw, we achieve even greater speedups with our approach! arxiv.org/abs/2603.02001

Dear Algorithm, pls show this to database people What do Database folks think about this?

This code now landed upstream and will be part of the 7.0 kernel when released. The liburing helpers, man page, and test cases have been merged into the master branch.

💿 io_uring is everywhere right now But for DBMSs it isn’t magic dust: • drop-in swap epoll/libaio → io_uring: ~+6–10% • redesign for async + batching + io_uring features: ~2× end-to-end • Postgres case study: +14% when “done right” arxiv.org/pdf/2512.04859… #io_uring #linux

I'm excited about the database performance io_uring will unlock. Last year I benchmarked Postgres 17 vs 18 to test the initial io_uring upgrades. I was surprised to see they weren't always a clear win for TPC-C. This paper studies the potential, and the future looks good.

Yes

Our paper is currently on Hacker News, feel free to join in if you have questions or comments. @melhindiCS @m4ttze and I are happy to take them.

Paper too long? We compiled a high-level and concise overview when io_uring is beneficial: toziegler.github.io/2025-12-08-io-…

@_devott @TigerBeetleDB @melhindiCS @Tobias__Ziegler @viktor_leis @cbinnig Latency-wise we can go down as low as ~9.3 microseconds for UDP and ~10.3 microseconds for TCP. DPDK for reference is ~7 microseconds on our system. However, DPDK lacks all the nice TCP/IP features. (more Details in the paper)

@TigerBeetleDB @m4ttze @melhindiCS @Tobias__Ziegler @viktor_leis @cbinnig This seems like it’s meant for efficient cpu and nic usage over getting higher throughput. DPDK and other frameworks that offer kernel bypass will probably be faster than this (latency wise) with high throughput. If this is for low latency, this is prob not gonna be useful

How modern database systems can leverage the Linux io_uring interface for efficient, low-overhead I/O arxiv.org/pdf/2512.04859… @m4ttze @melhindiCS @Tobias__Ziegler @viktor_leis @cbinnig

@AhamPiyush Paper: x.com/Tobias__Ziegle…

io_uring vs epoll in linux for Network IO

@__vishwanath__ @Tobias__Ziegler @cbinnig @melhindiCS @viktor_leis @axboe Yes! We evaluated them in this paper, including zero-copy receive on 400Gbit/s NiCs!

@Tobias__Ziegler @cbinnig @m4ttze @melhindiCS @viktor_leis @axboe 6.12/6.13 also got a bunch of Zero copy features which is great for network I/O bound applications!

Interested in io_uring? Check out our new paper, joint work with my former colleagues from TU Darmstadt (@cbinnig and @m4ttze) and from TU Munich (@melhindiCS and @viktor_leis). arxiv.org/abs/2512.04859…

Starship goes interdimensional during Flight 10. One of the most insane vapor cones I've ever seen.

Good morning

Mechazilla has caught the Super Heavy booster!

HARMONY OF RESILIENCE: Recorded in space and sent to Earth via @SpaceX’s @Starlink constellation, Polaris Dawn crewmember and violinist @Gillis_SarahE invites you to enjoy this music moment in support of @StJude & @ElSistemaUSA → polarisprogram.com/music

Stage separation!