Notes on systems and networks

Occasional writing about distributed systems, transport, storage, and the tooling around them. Mostly things I had to work out for myself.

• Measuring tail latency in a small clusterObservability

  Why the p99 you read off a dashboard rarely matches what users feel, and what to measure instead. August 2026.

• Queueing discipline, buffers, and the slow pathNetworking

  A practical look at bufferbloat on small links, why fq helps, and how to verify the change actually took effect. June 2026.

• A short note on congestion controlTransport

  What BBR and CUBIC are actually doing differently, and when the choice matters at all. April 2026.

• What fsync actually promisesStorage

  Durability, ordering, and the gap between what the man page says and what the hardware does. February 2026.

• Where to terminate TLSSecurity

  Tracing the tradeoffs between terminating at the edge, at the service, or not at all. December 2025.

• Backpressure is a design decisionArchitecture

  Most systems pick a backpressure strategy implicitly. Better to pick it on purpose. October 2025.