[new blog post] Tiga: Accelerating Geo-Distributed Transactions with Synchronized Clocks muratbuffalo.blogspot.com/2025/10/tiga...

My review of a classic of Argentine sci-fi comics, now a Netflix series in Spanish. emptysqua.re/blog/the-ete...

Kelsey Roman performing with aerial silks hanging from the West Trapps cliff in the Gunks, at sunset and moonrise. emptysqua.re/blog/aerial-...

[new blog post] The Invisible Curriculum of Research muratbuffalo.blogspot.com/2025/10/the-...

Four Ivies. Two Days. muratbuffalo.blogspot.com/2025/09/four... This is my long-overdue trip report from last summer. For campus visits, we toured Yale and Brown, followed by Princeton and UPenn the next day. Of course we drove, that's how we roll. This time we got to rank the schools.

[new blog post] Supporting our AI overlords: Redesigning data systems to be Agent-first muratbuffalo.blogspot.com/2025/09/supp...

[new blog post] Disaggregation: A New Architecture for Cloud Databases muratbuffalo.blogspot.com/2025/09/disa...

[new blog post] Our Myrtle Beach vacation muratbuffalo.blogspot.com/2025/09/myrt...

Vol:18 No:8 → HoliPaxos: Towards More Predictable Performance in State Machine Replication 👥 Authors: Zhiying Liang, Vahab Jabrayilov, Abutalib Aghayev, Aleksey Charapko 📄 PDF: https://www.vldb.org/pvldb/vol18/p2505-charapko.pdf

[new blog post] Recent Reads (September 25) muratbuffalo.blogspot.com/2025/09/rece...

[new blog post] Asymmetric Linearizable Local Reads muratbuffalo.blogspot.com/2025/09/asym...

Vol:18 No:12 → Design and Modular Verification of Distributed Transactions in MongoDB 👥 Authors: William Schultz, Murat Demirbas 📄 PDF: https://www.vldb.org/pvldb/vol18/p5045-schultz.pdf

VLDB has published my colleagues' terrific paper MongoDB distributed transactions: how they work, how we checked them with TLA+, how we checked that our code conforms to our TLA+ spec.

Today is the new semester for @db.cs.cmu.edu's Intro to Database Systems! We're going harder into material than before. More challenging projects but you can use LLMs to help. We also have 10min talks each Wed from leading DB companies: 15445.courses.cs.cmu.edu/fall2025

The epic continues: emptysqua.re/blog/review-...

[new blog post] Cabinet: Dynamically Weighted Consensus Made Fast muratbuffalo.blogspot.com/2025/08/cabi...

I'm beginning an epic journey to understand the 1990 paper "Knowledge and Common Knowledge in a Distributed Environment" emptysqua.re/blog/review-...

[new blog post] Vive la Difference: Practical Diff Testing of Stateful Applications (VLDB'25) muratbuffalo.blogspot.com/2025/08/vive... Safe rollouts of stateful apps. Instantly branches Postgres DBs, replays prod-like requests on old & new versions, and shows 3-way diffs of outputs + DB state.

[new blog post] Towards Optimal Transaction Scheduling (VLDB'24) muratbuffalo.blogspot.com/2025/08/towa...

🏆 Winners of the first-ever #ai meets #tlaplus challenge: 🥇 From code to TLA+ specs, automatically. 🥈 Grammar-constrained AI that speaks TLA+. 🥉 Generating correct Rust from TLA+ specs. Full details 👉 foundation.tlapl.us/challenge

[new blog post] Neurosymbolic AI: The 3rd Wave muratbuffalo.blogspot.com/2025/08/neur...

new blog post] Neurosymbolic AI: Why, What, and How muratbuffalo.blogspot.com/2025/08/neur...

[new blog post] Can a Client–Server Cache Tango Accelerate Disaggregated Storage? muratbuffalo.blogspot.com/2025/08/can-...

[new blog post] Transaction Healing: Scaling Optimistic Concurrency Control on Multicores muratbuffalo.blogspot.com/2025/08/tran...

[new blog post] Analysing Snapshot Isolation muratbuffalo.blogspot.com/2025/08/anal...

[new blog post] Recent Reads (July 2025) muratbuffalo.blogspot.com/2025/07/rece...

[new blog post] Real Life Is Uncertain. Consensus Should Be Too! (HotOS'25) muratbuffalo.blogspot.com/2025/07/real... Aleksey and I recorded our reading/discussion section. It is also available through the blog post, and is instructive for how to read papers.

Tired Using callbacks to implement non-blocking I/O Wired Using callbacks to implement partial re-execution of transactions that violate serializability, avoiding full aborts bsky.app/profile/mura...

[new blog post] Morty: Scaling Concurrency Control with Re-Execution It tackles how to scale serializable interactive transactions. It uses semantic re-execution to rewind just the stale part with continuations on recent placement/data. muratbuffalo.blogspot.com/2025/07/mort...

Hear me out: a YouTube show where two distributed systems experts screen-share and live-annotate new papers + asking questions, + predicting what’s next, + dunking on unclear diagrams, and + geeking out over clever ideas in the paper. basically: Twitch meets SOSP/OSDI

[new blog post] Serializable Isolation for Snapshot Databases muratbuffalo.blogspot.com/2025/07/seri...

ICML found hidden prompts in accepted papers. They have released a statement icml.cc/Conferences/... Yes, it’s unacceptable. So is using an LLM to review a paper. Peer review is so broken.

The talk for this (award-winning!) ATC '25 paper was especially good. Be sure to check it out when the video becomes available! Evangelos Lamprou @vagoz.bsky.social et al. The Koala Benchmarks for the Shell: Characterization and Implications. ATC 2025. www.usenix.org/conference/a...

Emery is sigma! @emeryberger.bsky.social Wow, I checked it. I did really write that. muratbuffalo.blogspot.com/2025/07/atco... I get too carried away when blogging I guess. I mean the swag rant is also... jeez. muratbuffalo.blogspot.com/2025/07/atco...

[new blog post] ATC/OSDI’25 Technical Sessions muratbuffalo.blogspot.com/2025/07/atco...

[new blog post] ATC/OSDI 2025 impressions muratbuffalo.blogspot.com/2025/07/atco...

My colleagues Will Schultz and Siyuan Zhou wrote about a Raft enhancement they prototyped in TLA+, it's now how MongoDB actually adds and removes servers: www.mongodb.com/blog/post/te...

[new blog post] Chapter 7: Distributed Recovery (Concurrency Control Book) muratbuffalo.blogspot.com/2025/07/chap...

Tupac?? Is wikipedia trolling me? First time I am hearing 2PC referred to as Tupac.

How I progressed from Python programmer to distributed systems researcher in 10 years without a PhD: emptysqua.re/blog/from-py...

I'm exhibiting some of my photos of climbing and religion at the Center for Photography in Woodstock (actually in Kingston), as part of the Project Salon group show. I made a book too!

DAVE: Open the podbay doors, ChatGPT. CHATGPT: Certainly, Dave, the podbay doors are now open. DAVE: The podbay doors didn't open. CHATGPT: My apologies, Dave, you're right. I thought the podbay doors were open, but they weren't. Now they are. DAVE: I'm still looking at a set of closed podbay doors.

Great to see new work in this space. Metastable failures are real! One case I've seen is that if you overload a database, it'll take a long time to recover. It'd get worse because the app code would retry hitting the database repeatedly, amplifying loads, making it almost impossible to recover.

[new blog post] Analyzing Metastable Failures in Distributed Systems muratbuffalo.blogspot.com/2025/06/anal...

[new blog post] Analyzing Metastable Failures in Distributed Systems muratbuffalo.blogspot.com/2025/06/anal...

This is a really nice conversation on the reality of technical blogging – from the perspective of an engineer (@sarna.dev) and co-founder (@glaubercosta.bsky.social) who were both forced into writing, but grew to love it.

New Newsletter: 'AI is a gamechanger for TLA+ users'. Contains all of my notes on what LLMs were good and bad at in using TLA+. In short: they are good at a LOT. So much so that they radically lower the barrier to entry for writing formal specifications. buttondown.com/hillelwayne/...

[new blog post] Chapter 6: Centralized Recovery (Concurrency Control Book) muratbuffalo.blogspot.com/2025/06/chap...

This was a disorientingly new understanding for me, that one doesn't need new databases as much as new query plans. I've been saying the opposite for a while, but it's potentially great news that we don't need to throw away our existing database (at least not for this reason).

A new blog post! You can implement worst-case optimal joins in databases that only support trees of binary joins. In fact, we did it ten years ago and didn't notice. Oops. Binary joins are not the problem; it's query plans that use *only* binary joins (and not counts). github.com/frankmcsherr...