Load is a shape, not a number
Why average requests per second hide the failures that matter, and how to reason about bursts, queues, and tail latency.
Engineering notebook
Short write-ups about distributed systems, performance under pressure, and the unglamorous work of keeping services predictable.
Why average requests per second hide the failures that matter, and how to reason about bursts, queues, and tail latency.
Retries feel harmless until they synchronize. A practical look at backoff, budgets, and retry storms in small systems.
What a health endpoint should actually prove, and why “the process is running” is rarely the same as “the service works”.
Work & Load is a small notebook for engineers who care about how systems behave when the graph goes up and to the right.
Topics include capacity planning, observability, queueing, failure modes, and the operational details that rarely make it into architecture diagrams.
Written in Frankfurt. No trackers, no pop-ups, no newsletter.