Many of the the lessons learned at @scylladb.com!

🚀 Thrilled to announce my interactive #WebAssembly course! Learn WebAssembly from the ground up with: 🔍 Custom-built debugger & VM 📊 Live stack & memory insights 💻 Hands-on exercises 🛠️ Build real projects Level up with WebAssembly! ✨ 👉 learn-wasm.dev

We at @tur.so are hiring for a Platform Engineer position. If you battled Kubernetes and lived to tell the story, and is a decent coder in Go (and maybe Rust), this position is for you. Location agnostic, but must be in the US timezone. turso.tech/careers/seni...

Today we have decided to make @penberg.org 's experimental project, an official Turso project, following its great success. What if instead of just forking SQLite, we were to completely rewrite it - in Rust? That's our moonshot - codename Limbo turso.tech/blog/introdu...

What material or papers should I read to understand alternatives to Rust's borrow checking with lifetimes for memory-safety? I hear some people mention linear types, for example. What should I read to understand how they provide memory-safety?

I am using Buffer to cross-post between here and X, but seems like threads are not either supported by Buffer or Bluesky so some posts are limited to X 🤔

Is there a way to test Github actions locally? It sucks so much to play whac-a-mole with the YAML config to fix one problem at a time...

With recent batch of improvements from Pere, I can now write to a SQLite file (it's WAL actually) with Limbo and read with SQLite itself. Nice!

Rust folks! In Limbo, I am implementing the SQLite C API with Rust. Right now I have the whole of SQLite API in a single file (https://buff.ly/3URtylE ) and like to split it up. But I seem to be unable to convince cbindgen to pick up symbols from other files. What am I doing wrong?

I believe you are thinking of: ⬅️ Why Threads Are A Bad Idea, John Ousterhout, 1995 ➡️ Why Events Are A Bad Idea. Rob von Behren, Jeremy Condit and Eric Brewer, 2003

How should I package Limbo for JavaScript? I now have a ”limbo-wasm” package that bundles the core database compiles to Wasm and does I/O with Node filesystem API. But this wont work in the browser. Another package or make ”limbo-wasm” detect runtime? 🤔

My memory tells me there are two papers. First one that talks about how event-based concurrency is better than thread-based concurrency and a second one that makes tries to make the opposite claim. But I can't find them anymore! Did my mind hallucinate them?

The fourth #FutureFrontend meetup focused on deterministic simulation testing and @penberg.org went into great detail in this new approach to find bugs you couldn't even imagine. You can find the recording (1.5h) online at www.youtube.com/watch?v=29Vz....

Published a new blog post on explaining the basics of disaggregated storage systems. The architecture is relatively new considering the rich history of databases, but pivotal enough to be adopted by many databases like Aurora, Neon, Snowflake, etc. https://avi.im/blag/2024/disaggregated-storage

The wgpu crate for Rust looks promising for cross-platform graphics programming! What are the best resources to learn more about WebGPU?

Hah, setting up @penberg.org handle on this thing was really easy!

I watched @penberg.bsky.social 's talk "Patterns of Low Latency" at P99 conf. This talk is dense with practical insights and to the point. I was wondering why watch a talk on latency, but his background as a Linux Kernel contributor, on OSv and ScyllaDB made worth it. Here are my notes

I am about to start doing slides for my next week’s presentation on deterministic simulation testing (DST). What kind of things folks want to know about?

Hey BlueSky, I run the Berlin Systems Group. The 5th edition of the meetup will happen on November 14th - this time we are finally diversifying. Only 1 database talk and the other is going to be a deep dive into LLVM passes! Register here: lu.ma/nq4bzyjb #databases #compilers #berlin #meetup

I had convinved myself that disaggregated storage is the way to go for OLTP, but now i am starting to convince myself that ”diskless” (object storage as the primary storage) is where disaggregated storage converges to (on cloud anyway). All we need is little bit cheaper writes on AWS S3 and others…

I learned that BlueSky uses one SQLite database per user, as God intended. Where can I read more about its high level architecture?

I can't stop thinking about this blog post where they replaced Redis with SQLite—and surprisingly, SQLite was faster! What's interesting is that Redis was running locally, SQLite was storing the data on disk. So it was memory (Redis) vs disk (SQLite), but Redis needed to communicate through IPC.

I am looking forward to P99 starting today!

GPUs are essential for modern AI workloads, but how do they execute code and why are they so much for efficient than CPU for such workloads?

What are the seminal database management systems everyone should know about? For example, there's Integrated Data Store (IDS) (1964), IBM's System R (1974), INGRES (1974) and arguably Postgres (1986) and Dynamo (2007) belong to the list as well. But what else am I missing?

I have some really basic understanding of GPU programming with shaders and CUDA, but am mostly clueless of actual GPU architecture and how code gets executed on it. What books or blog posts would you recommend me to read to catch up?

I try to do an online course on Coursera or similar platform about once a year. My recent courses have been "Neural Networks and Deep Learning" (2023), "Hardware Description Languages for FPGA Design" (2020), and "The Introduction to Quantum Computing" (2019). What would you recommend for 2024?

This is super good and everyone working with LLMs should read https://www.honeycomb.io/blog/hard-stuff-nobody-talks-about-llm

The third day in a row GitHub is having issues. I don't remember the service being *this* unstable, ever. One incident? Happens. Two in a row? Bad luck. Three in a row - what's going on?

The Amazon Prime Video team was able to reduce cost by moving from Serverless backed by Lambda to monoliths running on VMs. "Moving our service to a monolith reduced our infrastructure cost by over 90%. It also increased our scaling capabilities."

Great story about how Avinash Sajjanshetty discovered a bug in a research paper on Hekaton multi-version concurrency control: https://avi.im/blag/2023/internet-mvcc/ As Linus's law says "given enough eyeballs, all bugs are shallow".

Hey all, I am Pekka, mostly will be talking about distributed systems, databases, and operating systems. Curious to see what kind of role Bluesky will have with me also still being on Twitter and Mastodon. Twitter is the main one for me, but lots of interesting folks migrated to Mastodon for good.