Blog

System Design Interview.

3 posts
  1. 01

    Load-Testing a URL Shortener (opens in a new tab)

    Part 2 · Medium

    Pushing the linkpulse redirect path with k6 until something gave. A fixed VU test measures concurrency, not throughput, and at 3,000 requests a second one Node process ran out of CPU while Redis and Postgres sat at about 25%.

  2. 02

    Building a URL Shortener: Six Design Decisions and the Alternatives I Rejected (opens in a new tab)

    Part 1 · Medium

    Six decisions behind a URL shortener that runs in public, each with the alternative I rejected: a 302 rather than a 301, caching misses as well as hits, a sliding window rate limiter in Lua, and refresh tokens with reuse detection.

  3. 03

    I Built a Twitter-Style Feed to Find Out What Fan-Out Actually Costs (opens in a new tab)

    Part 2 · Medium

    Push and pull feeds built over the same data and measured. Fan-out costs about 12 microseconds per follower, a timeline entry about 80 bytes in Redis, and the read advantage was far smaller than expected.