A simple way to get more value from tracing
Explains how to get high value from distributed tracing with less effort, using a real-world implementation from Twitter as a case study.
DanLuu.com is the personal blog of Dan Luu, known for long-form essays that mix systems thinking with careful measurement and clear writing. The topics range from computer latency and input lag, testing versus informal reasoning, and concurrency bugs, to industry pieces on developer compensation and curated lists of programming blogs worth reading. Many posts include data, historical context, and reproducible reasoning, which is why the site is often cited in courses and shared across the developer community. The design is intentionally minimal, which puts all attention on the ideas.
136 articles from this blog
Explains how to get high value from distributed tracing with less effort, using a real-world implementation from Twitter as a case study.
How Twitter built LongTermMetrics (LTM), a system that identified and shipped millions in cost savings by analyzing JVM metrics.
An analysis of why many corporate engineering blogs fail to attract readers and the key processes used by successful ones like Cloudflare and Segment.
Analyzes the growth of command line options in Unix/Linux tools from 1979 to 2017, contrasting it with the original 'do one thing well' philosophy.
Explores how sharp income cutoffs in US tax and subsidy policies can create perverse incentives for people to intentionally lose money.
Argues that achieving the 95th percentile in many activities is not impressive, using examples from video games and real-world skills.
A software engineer critiques algorithms interviews, arguing they don't reflect real-world problem-solving, despite providing value by fixing such issues on the job.
A talk transcript explaining why files and filesystems are complex, error-prone abstractions for developers, using Dropbox as a case study.
An experiment testing if players with feminine usernames receive different in-game chat comments than those with masculine names in Overwatch.
Analysis of the 'fsyncgate' email thread discussing PostgreSQL's unsafe handling of fsync errors that can lead to data corruption.
A technical analysis measuring computer latency from 1977-2017, finding that some modern machines are slower than 40-year-old computers.
Analyzes decision-making quality in sports and board games, where clear data reveals the high cost of poor choices.
Analyzes Android device fragmentation using market share data to show how outdated devices accumulate over time.
The article critiques poor UI changes in apps like Zulip and Facebook, comparing them to confusing physical designs like doors.
Analyzing filesystem error handling by reproducing past studies on modern systems like ext4 and btrfs.
An investigation into keyboard latency claims, comparing modern gaming keyboards to older computers using high-speed measurements.
An introduction to CPU branch prediction, explaining its purpose and classic algorithms to understand modern research.
An explanation of Sattolo's algorithm for generating a random permutation with a single cycle, aimed at those without a strong math background.
The article explores the impact of input latency on user experience in tablets, VR, and computers, debunking the myth of instant response.
Debunks widely cited studies claiming the mouse is always faster than the keyboard, arguing for a task-specific approach to input efficiency.