My colleague Sanjay Ghemawat & I have done a fair bit of performance tuning of various pieces of code. We wrote an internal Performance Hints document ~2 years ago as a way of identifying some general principles & we've recently published a version of it externally. Doc: abseil.io/fast/hints.h...

DRAM prices mooning will hurt everyone and might cause a consumer tech recession on its own in the next 12 months even without the AI bubble popping. Hyperscalers are going to jack up prices for cloud resources, electronics manufacturers (Apple, Samsung, etc.) will raise prices, this is gonna hurt.

The testing practices (UT, IT or any other tests) define how mature and well (g)rounded engineering org is. #Engineering

One of the many joys of using AI for programming is the creation of huge PRs on complex topics that the authors barely understand, but still suggest "because they work". Here's a great example from #OCaml github.com/ocaml/ocaml/... Kudos to OCaml's maintainers for handling this so gracefully.

It seems that only #MicrosoftAzure supports XDP native? @awscloud.bsky.social seems to support with ENA(full or limited?) while #GoogleCloud public documentation has no mention of XDP native support and infact says its not supported in both vertIO or gVNIC interfaces. Good job #MicrosoftAzure!

Great write up. Sometime back I was going around Katran by Meta. The checksum calculations and other references sounded very familiar :)

One of the huge perks of working at a software company is, when there's something about the product that bothers you personally, you can get your hands dirty and find out why it's basically impossible to get fixed

Some great finding and discussion at www.reddit.com/r/java/comme... I was casually looking at record's showing up in profiling reports and then bump into this. Thanks to @lukaseder.bsky.social

We're growing! We're working with some awesome enterprise customers with tonnes of data and are trying to scale the team. We're hiring for many high-leverage roles, including sw-eng / sre, solution engineering, and GTM roles. Fully remote team. DM me if you have any questions!

PostgreSQL 18 is released!

Anyone have any recommendations for links specifically about Rust lifetimes, ownership, move, borrow, etc?

Just learnt that “syzygy” is a real English word

🎉 NEW PARTNER: @tigerbeetle.com Introducing the new TigerBeetle Connector for Redpanda Connect! Get financial transaction updates into #Kafka-compatible topics for real-time #analytics insights in downstream systems, faster. Learn more about this powerful integration and how to use it👇

💯 This is exactly the reason we should have all the engineers on the pager. It doesn't matter if you are staff/principal. Need to be on-call.

This is an absolute banger. The talk is a "benchmark" for how tech talks should be. Amazing work, and reference to staff and principal level problems!!

As an industry, we lack at writing quality reads. Be it design, debugging story, incident postmortem or a new proposal. 90% of the docs are how to apply a helm file, yaml file, change a config. The "why" part is completely missing in the docs. #Software #engineering

Getting into the world of raft! Mostly because we are not able to understand consul well enough so that we can be confident of running it at some crazy scale. thesecretlivesofdata.com/raft/ was good. I remember @eatonphil.bsky.social has an impl in less 1k loc & I'll have to refer & do the same!

Already chatGPT has moved to expert level?? People saying "I confirmed with chatGPT"!!

One of the biggest mistakes that we can do is to set wrong examples. One such example - indiscriminate mocking in the UT's. It becomes a subconscious pattern that most just follow it.

this is obviously the use case where io_uring best showcase its capabilities 😲 tangled.sh/@rockorager....

There are good cloud examples and there are bad ones. I have been part of many cloud negotiations and I echo every bit of the writeup. Companies that can watch their costs and care about it, will always get their act together on cloud.

Please read this and save yourself from possible txnid wraparound!

I am subscribing to it the moment it goes live!

Microsoft open-sourced a sizable chunk of the GitHub Copilot VS Code extension today (MIT licensed) - I did some spelunking to try and find the LLM prompts that make it all work simonwillison.net/2025/Jun/30/...

left: the interview questions right: the actual job

deepwiki.com is pretty neat. Found it pretty good for the public repos.

Git pull seems to be running on dial up connection.

With this Apple Container Framework stuff, I need to clarify something that I see confusing a lot of people: In Docker Desktop/colima/lima/Rancher Desktop today a single VM runs all of your containers. With Apple Container Framework, each container is it's own VM.

[arXiv] Fast Userspace Networking for the Rest of Us arxiv.org/abs/2502.0... "if it doesn't work on the cloud does it even matter?" but applied to kernel bypass networking.

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.

Was looking at refactoring a SQL (#postgres). Haven't written a lot of SQL in last many years and was bit rusty. I knew what I wanted and copilot with Claude 3.7 gave the right query. Of course, validated for correctness. #AI can be large time saver when used properly.

The flow never ages! For the past ~20 years(my time in the industry), the first thing to check is stats(irrespective of DB). Have had pleasure(?) of asking to run this as part of production incidents a few times!

1️⃣ Stale stats: ANALYZE hasn’t run; filter looks un-selective. 2️⃣ Truly un-selective: Stats are right; scan really is cheaper. 3️⃣ Complex estimates: Many joins / ORs / CTE / functions scramble row estimates. 4️⃣ Handcuffs on: RLS limits optimizations or GUCs forbid the path. 5️⃣ Planner bugs: Rare but...

From hours to minutes is brilliant! Yet to see the video, but what if there are bunch of business logic between bronze and gold? Most likely @feldera.bsky.social would handle that? Also, how it would compare cost wise with batch?

This looks and reads very neat. Time to get a feel of it!

This is great write up. One of the challenges with testing something for a scale system is to do it without introducing regression and not breaking the bank while making sure most of the scenarios are still covered.

Newsletter: I am sick and god damn tired of everybody pretending that generative AI is the next big thing. The media is complicit in accepting fantastical nonsense - both in the numbers put out by OpenAI and the silly jobs created by Anthropic - and it has to stop. www.wheresyoured.at/reality-check/

If you call yourself a "software engineer" you better be able to: 1. Flip a linked list 2. Flip it again 3. Bop it 4. Twist it 5. Twist it 6. Pull it 7. Bop it

0x.tools xCapture v3: Linux Performance Analysis with Modern eBPF and DuckDB 👀👀 tanelpoder.com/posts/xcaptu...

ooh, @b0rk.jvns.ca has a wonderful blogroll jvns.ca/blogroll/

Async-Profiler v4.0 Heatmaps and Native memory profiling and VMStructs-based stack walker github.com/async-profil...

This is an excellent take on the current state of affairs in #Bengaluru. Near my home, good wide asphalted road has been taken up for white topping. The road doesn't see any buses or heavy vehicles. As expected, the topping is far from complete. www.youtube.com/watch?v=lV7s...

If your codebase uses org.json and you wanted to try a better alternative like widely adopted fasterxml, then you are in luck - github.com/FasterXML/ja... FasterXML continues to amaze and heartfelt thanks to @cowtowncoder.bsky.social

If you are on Java8+(which you should be, think again if you aren't), then a cleaner way of timing out a CompletableFuture is orTimeout. In #Java 8 you would have to create a future, schedule timeout on it and do CompletableFuture.anyOf().

This blog post is a great summary of GHA pain points. The major issue is too much hidden complexity—good discussions in the HN thread. I like some advice in the comments: - Write as much CI logic as possible in your own code. - Be able to test your pipelines locally. - Use your own runners.

Java 24 / JDK 24: General Availability: mail.openjdk.org/pipermail/jd... Features: openjdk.org/projects/jdk... #java24 #jdk24 #openjdk #java

After analysing the cpu usage of concurrent marking threads of G1, it looks like for larger heaps (>50G), the ideal setup seems to be newsizemax + old regions + humongous regions < IHOP. Did I get the heuristic completely wrong here? #java #jvm #garbagecollection

What are the options for off-heap maps in 2025? Is it Chronicle-Map, Mapdb, rocksdb, oak or anything else? #java #off-heap #jvm

I enjoyed this post on supply chains and days pipelines. The analogy is surprisingly robust.

Ingress is free in cloud is a false claim unless you are getting data into one of the PaaS offerings(eg: storage, PaaS DB exposed to internet). Can get ingress for free if the volume can be managed by single VM or you are OK with DNS load balancing. #CloudComputing #hyperscalers