A small lab for network and systems work.

What happens here

Mostly unglamorous work: measuring links over long periods, watching how systems behave when conditions are imperfect, and keeping a small number of services running for years rather than weeks.

The interesting problems tend to be the boring ones. Why a connection is slow at one time of day and fine at another. What actually happens to a service across a reboot. Whether a change that looked harmless left something behind.

Current interests

• Long-baseline measurement

  Continuous latency and throughput sampling, kept long enough that a single bad afternoon stops looking like a trend.

• Behaviour under failure

  What a system does when a dependency is slow, absent, or lying, rather than what it does when everything works.

• Reproducible operations

  Changes that can be described, reversed, and verified afterwards, with the evidence written down at the time.

• Small, durable services

  Fewer moving parts, boring dependencies, and a preference for things that keep working while unattended.