How Amazon ECS manages CPU and memory resources: Explains when, how to use task limits vs container limits within tasks.
Valhalla - Where Are We? by Brian Goetz: Change to the way java stores objects in memory for efficiency. Value classes (no identity) and bang lists. (Non-null guarantees) Allows for memory flattening - fewer pointers and headers per object.
Eileen Uchitelle - The Myth of the Modular Monolith - Rails World 2024

Challenges slide

My solar-powered and self-hosted website: The maker of drupal runs a website on a raspberry pi zero 2 powered by a solar panel and asks why so many websites need high availability. This default creates wasted energy and contributes to our current climate crisis.
Our cloud-exit savings will now top ten million over five years: More about 37signals cloud exit and dollars saved. Higher than expected and sooner it sounds like.
Solid Queue 1.0 released: Talks about new built-in rails lib for deferred tasks and also the scale they’re running with @ Hey! Nice writeup.
Incident management at major sporting goods e-commerce: Nice writeup of an evolved incident response process. They’ve created tools that help get a major incident rolling.
- A reporter selects product module, severity and impact
- An incident channel is notified with a details message (at this point a priority has been assigned to an incident determining next steps in their incident response workflow - eg p1, p2 triggers slack collaboration across teams)
- Ticket is created
- Slack warroom is created (Not for p4, p5 incidents)
- Incident commander is selected
- Communication is greatly improved, and they reduced time to declare an incident

Incident declare form

Incident declare form slack message to incident channel

First warroom message

Declare commander

Overarching workflow

P99 Conf

Conference page with videos: A really great online, pre-recorded video conference. I’ve watched p99 videos from previous years and have always enjoyed.
DTrace at 21: Reflections on Fully-grown Software: Bryan Cantrill reflects on 21 years of Dtrace. Huge contribution to our field around production debugging.
Patterns of Low Latency: Some techniques to avoid latency altogether or when it’s necessary how to make it less visible.
- avoid moving data
- avoid work
- avoid waiting (synchronization bad, context switching, blocked waiting)