Engineer Reads

Mon, 01 Jan 0001 00:00:00 +0000

Engineering Reads — 2026-07-29#

The Big Idea#

Applying rigorous frameworks to our engineering intuition reveals the hidden realities of software systems. Whether using statistical curves to expose how the slowest 1% of requests secretly consume the majority of your server capacity, or applying formal logic to untangle domain modeling, adopting a more mathematical lens pays outsized dividends in system design.

Deep Reads#

Lorenz and Little: How Much Does Your Tail Cost? · Marc Brooker · brooker.co.za Tail latency is often treated purely as a customer experience problem, but it is actually a massive driver of system cost and capacity demand. Brooker demonstrates this by using the empirical Lorenz Curve to calculate how much requests above a certain latency percentile contribute to the overall mean latency. Because Little’s Law dictates that a request’s contribution to mean latency equals its contribution to system concurrency, a p99 latency that contributes 52% of the mean indicates that the top 1% of your slowest requests are occupying 52% of your busy threads. While the exact mapping from concurrency to cost gets complicated in queued or event-based systems, and estimating true percentiles from raw samples introduces statistical noise, the core mechanics hold true. Systems engineers optimizing for cost should read this to understand why ignoring or casually trimming off heavy tails is a mistake when those outliers are precisely what drives your infrastructure bill.

Formal Logic on MacWorks

Engineer Reads

Engineering Reads — 2026-07-29#

The Big Idea#

Deep Reads#