Pierre Zemb

Pierre Zemb

@pierrezemb.fr

🧑‍💻 Staff Engineer @clever-cloud.com 🤩 Distributed and Database Systems 🚀 Co-leader Finistdevs.org 💪 Squash player

900 followers 661 following246 posts

👀 Is it better for 100% of customers to experience a 5% failure rate, or 5% of customers to experience a 100% failure rate? ✍️ My notes on cell-based architecture, or how to densify multi-tenant infrastructure without sharing every failure 🔗 pierrezemb.fr/posts/notes-...

pierrezemb.frNotes about cell-based architectureCell-based architecture as the middle ground between one dense cluster and one cluster per customer, and why moving data is the core capability.

Zero-downtime Postgres upgrades with logical replication tapoueh.org/blog/2026/10...

tapoueh.orgZero-downtime Postgres upgrades with logical replicationThis is part 3 of a series about Postgres logical replication use-cases, and about how the feature set has evolved over the past ten years and ten releases …

Seeing Claude Code web debugging it’s own MR through deterministic simulation is always a joy

Pierre Zemb reposted

[𝗘𝗩𝗘𝗡𝗧] Deux jours, deux talks et l’équipe Clever à Clermont-Ferrand : direction @volcampio.bsky.social ! 🌋 @pierrezemb.fr & Alexandre Burgoni parleront d’ETCD et Kubernetes, @lostinbrittany.bsky.social d’architecture cloud. 📍 1 & 2/10 www.billetweb.fr/volcamp-2026 🧵👇 Thread

Jev introduces a new shape of LLM—System One, aka Decision Models simonwillison.net/2026/Sep/21/...

simonwillison.netJev introduces a new shape of LLM—System One, aka Decision ModelsLast week TypeSafe AI unveiled Jev, their first example of a new category of model that they are calling “System One models” (I’m with Maggie Appleton, I think “decision models” …

The architecture of Neki — PlanetScale planetscale.com/blog/the-arc...

planetscale.comThe architecture of Neki — PlanetScaleSharded Postgres, from the team behind Vitess

Any Nix package, live in your browser fzakaria.com/2026/09/04/a...

fzakaria.comAny Nix package, live in your browsertl;dr Try it at https://trynix.dev. Click hello, or python 3.6.2 from 2017, or two eras of hello at once, or a package that exists in no public cache. A Linux machine boots in the tab and you get a…

For the past two years, I’ve been building my own deterministic simulation framework in rust, and now I’m having fun learning Paxos with DST as the harness. 2026 is wild for engineers 😎

I just published a new blogpost about one of my favorite papers I discovered this year: "Explaining #FoundationDB's Architecture Through Compartmentalization" pierrezemb.fr/posts/founda...

pierrezemb.frExplaining FoundationDB's Architecture Through CompartmentalizationCompartmentalization explains FoundationDB's architecture: find the responsibilities that got coupled together, split them, and scale only the parts that can scale.

NVMe Polishes Its Specs, Brings Virtualization to Locally Attached SSDs - Slashdot hardware.slashdot.org/story/26/08/...

hardware.slashdot.orgNVMe Polishes Its Specs, Brings Virtualization to Locally Attached SSDs - SlashdotThe latest NVMe specifications add SSD-level virtualization that can simplify live VM migration by preserving storage identities across servers. The updates also introduce support for post-quantum…

#ApacheAvro frustrates a lot of people: you need two schemas just to decode a record. That annoyance is exactly why I like it. Here's why schema resolution makes rolling upgrades boring, in a good way. pierrezemb.fr/posts/why-ap...

pierrezemb.frWhy Apache Avro Needs Two SchemasAvro payloads cannot be decoded without the writer schema, and that two-schema model is what makes rolling upgrades and schema evolution work.

Pierre Zemb reposted

Martin Kleppmann

@martin.kleppmann.com

Engineering management after the cost of code collapsed: “Investing in machine-checkable correctness is now among the highest-leverage infrastructure work an org can fund” karimjedda.com/engineering-...

Pierre Zemb reposted

@sunnytech.bsky.social

☀ "Et si on faisait du simulation-driven development ?" par @pierrezemb.fr Découvrez le Deterministic Simulation Testing (DST) pour simuler et diagnostiquer les scénarios et pannes de systèmes complexes de manière totalement déterministe. #SunnyTech

Pierre Zemb reposted

Replying to @clever.cloud

📍 @sunnytech.bsky.social — 02 & 03/07, Montpellier Retrouvez Julia Buchner, Pierre Zemb, @LostInBrittany, Alexandre Burgoni et Aurélien Hébert sur notre stand. @pierrezemb.fr y donnera aussi sa conférence : « Et si on faisait du simulation-driven development ? ». sunny-tech.io

sunny-tech.ioSunny TechSunny Tech est la 1ère conférence Tech organisée à Montpellier. Une conférence annuelle des technologies du numérique, créée par des passionnés, pour des passionnés avec l’ambition de rayonner nationa...

Honored to be featured by @eatonphil.bsky.social in The Consensus 🙏 We talked about the data services we are building at @clever.cloud, all API-compatible with Redis, PostgreSQL and etcd on top of FoundationDB. Give it a read 👇

Phil Eaton@eatonphil.bsky.social

I spoke with Pierre Zemb from French neo-compute company Clever Cloud. They're building data services API-compatible with popular systems like Redis, PostgreSQL, and etcd on top of FoundationDB and Pierre is leading the work. There is no paywall, give it a read. theconsensus.dev/p/2026/06/18...

Pierre Zemb reposted

@eatonphil.bsky.social

I spoke with Pierre Zemb from French neo-compute company Clever Cloud. They're building data services API-compatible with popular systems like Redis, PostgreSQL, and etcd on top of FoundationDB and Pierre is leading the work. There is no paywall, give it a read. theconsensus.dev/p/2026/06/18...

Me, denied by FoundationDB's simulation because my PoC wasn't reboot-proof 🤯

I just published 'Designing Fakes That Prove Correctness' ✍️ The blog also has a new theme, and it hides a quorum easter-egg: flip to single-replica reads and see what it was protecting you from 👀 pierrezemb.fr/posts/design...

pierrezemb.frDesigning Fakes That Prove CorrectnessYou fake the handful of methods your code actually calls, prove the fake stays honest, and then make it fail the way production does.

Building an AsyncIO executor for the 3DS blog.cat-girl.gay/3ds-async-pa...

blog.cat-girl.gayBuilding an AsyncIO executor for the 3DS (pt 1!)the Nintendo 3DS is a very fun device to write homebrew for, but one thing about it can be annoying: its multitasking is non-preemptive! in this little se...

Pierre Zemb reposted

[𝗖𝗞𝗘] Quand @zwindler.fr attend notre Kubernetes managé depuis des années, puis ouvre le capot dès la bêta publique, forcément on lit attentivement 👀 Merci Denis pour ce retour et vivement la partie 2 !

Denis@zwindler.fr

J'attends le #Kubernetes managé de @clever.cloud depuis des années. J'avais même déployé un CP K8s sur leurs "apps Linux" juste pour patienter 😅 CKE (Clever Kubernetes Engine) est enfin en bêta publique. Un premier post pour en parler (il y en aura d'autres) blog.zwindler.fr/2026/05/12/t...

blog.zwindler.frTest de CKE, le Kubernetes managé chez Clever Cloud (partie 1)La suite de ma saga avec Clever Cloud Ça fait maintenant quelques années que je teste les produits de Clever Cloud sur ce blog. En 2022, j’avais testé le Clever Kubernetes Operator, qui permet de prov...

Pierre Zemb reposted

@finistdevs.bsky.social

La session de mai du FinistDevs, c’est mardi 12 mai à 18h30 chez Zenika Brest. Au programme • Xavier Rollet: “Un environnement de dev sur mesure avec NixOS” • Benoît Morin / Ifremer: “Centre Datarmor”, HPC, IA, big data et infrastructures scientifiques Inscription : www.meetup.com/finistdevs/e...

meetup.comSoirée FinistDevs Mai 2026, Tue, May 12, 2026, 6:30 PM | MeetupSalut à toutes et à tous, La session de mai du FinistDevs aura lieu le **mardi 12 mai à 18h30**, toujours chez **Zenika Brest**. Au programme : deux talks, avec du Linux

I gave a lightning talk at #BugBash 2026, sharing the stage with people whose work shaped how I think about distributed systems. Still processing that. Wrote up why the correctness decade has started, and why you trust the simulator, not Claude 👀 pierrezemb.fr/posts/bugbas...

pierrezemb.frBugBash 2026, or how the correctness decade has startedA retrospective on BugBash 2026, and why software engineering looks brighter than ever.

Pierre Zemb reposted

Et hop, je mets la touche finale à mes slides alors que le train arrive à Montparnasse. C'est bon, je suis prêt pour @devoxx.fr demain matin ! Si vous voulez savoir comment marche un moteur de requête, c'est demain à 10h30 ! m.devoxx.com/events/devox...

m.devoxx.comDevoxx Mobile CompanionYour ultimate companion for Devoxx conferences worldwide. Browse talks, speakers, schedules, and manage your personalized conference experience.

On my way to #BugBash 2026 🤩 First Paris, then Washington DC ✈️ I will be presenting a lightning talk about how we borrowed #foundationdb simulation framework in #rustlang

Pierre Zemb reposted

@eatonphil.bsky.social

DataFusion 53 came out last week. I took a look at the individuals and companies who committed the most. (paywall has expired) theconsensus.dev/p/2026/04/03...

How #NASA Built #Artemis II’s Fault-Tolerant Computer – Communications of the ACM cacm.acm.org/news/how-nas...

cacm.acm.orgHow NASA Built Artemis II’s Fault-Tolerant Computer – Communications of the ACMMembership in ACM includes a subscription to Communications of the ACM (CACM), the computing industry's most trusted source for staying connected to the world of advanced computing.

I just published a post on why I think fakes are more powerful than both mocks and Testcontainers. A fake that is worse than production: dropping writes, returning stale data... It catches the bugs that containers faithfully hide. pierrezemb.fr/posts/why-fa...

pierrezemb.frWhy Fakes Beat Mocks and TestcontainersTrait-based fakes test partial failures that mocks and Testcontainers fundamentally cannot simulate. They are also the on-ramp to simulation-driven development.

I just gave a talk "Testing: prevention vs discovery" at the #FinistAI x #FinistDevs crossover meetup 🧪 DST is the ultimate feedback loop for LLMs. It lets them discover what they don't know. No recording, but slides are available: pierrezemb.fr/slides/2026-... #deterministic #simulation #LLMs

👀 How does a #fuzzer find a 3-byte crash sequence in 9,000 iterations instead of 16 million? 🛠 Coverage turns a multiplicative problem into an additive one. I built a minimal #rustlang fuzzer from scratch to understand how it works: pierrezemb.fr/posts/diving...

pierrezemb.frDiving Into Coverage-Guided FuzzingHow a 600-line Rust fuzzer discovers a 3-byte crash sequence in 9,000 iterations instead of 16 million

🎉 I will be speaking at #SunnyTech 2026 about simulation-driven development! #SeaTechAndSun

Rob Pike's 5 Rules of Programming "write stupid code that uses smart objects" is my favorite one! www.cs.unc.edu/~stotts/COMP...

cs.unc.eduRob Pike's 5 Rules of ProgrammingPike's rules 1 and 2 restate Tony Hoare's famous maxim "Premature optimization is the root of all evil."

Google Engineers Launch "Sashiko" For Agentic AI Code Review Of The Linux Kernel - Phoronix www.phoronix.com/news/Sashiko...

phoronix.comGoogle Engineers Launch "Sashiko" For Agentic AI Code Review Of The Linux KernelGoogle engineers have been spending the past number of months developing Sashiko as an agentic AI code review system for the Linux kernel

"Even in healthy clusters, MariaDB Galera Cluster exhibited Lost Update and Stale Read" Jepsen: MariaDB Galera Cluster 12.1.2 jepsen.io/analyses/mar...

jepsen.ioJepsen: MariaDB Galera Cluster 12.1.2MariaDB Galera Cluster is a replicated version of MariaDB, a popular SQL database. While MariaDB claims Galera ensures “no lost transactions”, it loses transactions in at least two scenarios. First,…

An Interactive Intro to CRDTs | jakelazaroff.com jakelazaroff.com/words/an-int...

jakelazaroff.comjakelazaroff.comJust a programmer trying to make a home for myself on the WWW.

📝 The datafusion-index-provider PoC from last year is now a real library in datafusion-contrib, running in production. 🔧 New post on building secondary index queries with standard DataFusion operators pierrezemb.fr/posts/datafu... #Rust #DataFusion

pierrezemb.frBuilding Index-Backed Query Plans in DataFusionWhat I learned about manually constructing physical query plans for secondary index queries, and the library that came out of it.

TLA+ By Example tlabyexample.com

tlabyexample.comTLA+ By ExampleLearn TLA+ specifications through interactive examples in your browser

Pierre Zemb reposted

Martin Kleppmann

@martin.kleppmann.com

The second edition of Designing Data-Intensive Applications, by myself and @chris.blue, is finished and sent off to the printers! Ebooks should be available in the next week, and print books in 3–4 weeks. Sigh of relief. 😅 (BTW, this is a good opportunity to support your favourite local bookshop!)

👀 Turns out I'm #163 on the most active GitHub committers list in France A non-trivial amount of credit goes to #ClaudeCode, my favorite coding partner these days! committers.top/france

committers.topcommitters.top - Most active GitHub users in FranceA list of most active GitHub users in France.

✍️ I just published "Simulating Leader Election on top of FoundationDB" 🧪 Can LLM-generated code be trusted for distributed systems? Only if you have the right feedback loop. pierrezemb.fr/posts/simula... #FoundationDB #rustlang #simulation

pierrezemb.frSimulating Leader Election on top of FoundationDBCan LLM-generated code be trusted for distributed systems? Only if you have the right feedback loop.

💡 In #FoundationDB, your reads determine whether YOU conflict. Your writes determine what OTHERS conflict with. ✍️ Wrote about OCC patterns for layer engineers: phantom conflicts, snapshot reads, atomic ops, and versionstamps. pierrezemb.fr/posts/fdb-tr...

pierrezemb.frFoundationDB's Transaction Model for Layer EngineersHow optimistic concurrency control works in FoundationDB, and how to design data structures that stop fighting it

I recently had an intense #squash match this morning, lasting an hour and ending with a 3-2 victory ✌️ Throughout the entire game, I maintained a heart rate consistently above 190 beats per minute 🥵

✍️ I keep having the same conversation with colleagues about using LLMs for code. Finally wrote it down. Planning, context, feedback loops, and why simulation + LLMs is a powerful combo. pierrezemb.fr/posts/llms-f... #llm #rustlang #testing

pierrezemb.frWhat I Tell Colleagues About Using LLMs for EngineeringLLMs amplify expertise, they don't replace it. Here's what works: planning, context, feedback loops, and building systems that let AI discover bugs.

Pierre Zemb reposted

@fasterthanlime.hachyderm.io.ap.brid.gy

RUST REFLECTION MVP HAS LANDED https://github.com/rust-lang/rust/pull/146923 THIS IS NOT A DRILL

github.comReflection MVP by oli-obk · Pull Request #146923 · rust-lang/rustI am opening this PR for discussion about the general design we should start out with, as there are various options (that are not too hard to transition between each other, so we should totally jus...

🎄 Before heading to holidays, I wrote my 2025 year in review. ✍️ Back to engineering after 2 years of management. Built moonpool, a #FoundationDB simulation framework in #rustlang. Published 20 posts. Learned to work with LLMs. A year of going deeper. pierrezemb.fr/posts/2025-y...

pierrezemb.fr2025: A Year in ReviewReflections on returning to engineering, discovering simulation as a superpower, learning to work with LLMs, and the compounding value of years of distributed systems work.

🔧 I just published a blogpost about specs in the #LLM era. Model checkers like #Fizzbee work for algorithms. But for APIs and business logic? The tooling doesn't exist yet. pierrezemb.fr/posts/specs-...

pierrezemb.frSpecs Are Back, But We're Missing the ToolsLLMs revived specifications. Model checkers like Fizzbee work for algorithms. But for APIs and business logic? The tooling doesn't exist yet.

What Does a Database for SSDs Look Like? - Marc's Blog brooker.co.za/blog/2025/12...

brooker.co.zaWhat Does a Database for SSDs Look Like? - Marc's BlogMaybe not what you think.

🧪 How do you test #rustlang code that runs on top of #FoundationDB? ✍️ I wrote about designing simulation workloads: the patterns that find real bugs, the rules you can't break, and the pitfalls pierrezemb.fr/posts/writin...

pierrezemb.frDesigning Rust FDB Workloads That Actually Find BugsPatterns and principles for writing Rust simulation workloads that catch bugs before production does.

This is an awesome podcast episode about debugging a Grown-up #ZFS Data Corruption Bug podcasts.apple.com/fr/podcast/o...

podcasts.apple.comGrown-up ZFS Data Corruption BugÉpisode de l’émission · Oxide and Friends · 26/11/2025 · 1 h 21 min

Currently having a lot of fun modeling some internal spec in #fizzbee!