knaxuslearn · build · ship

Roadmaps / Distributed Systems / Who this is for

Read first

Who this is for

Hours by path. A Senior who skims Week 1 and skips Week 0 does about 120 hours. A Staff reader doing Weeks 3, 4 and 9 to 16 needs 10 weeks, 80 hours: 27 weeks at 3 hours a week. The work version is a separate effort of about 30 hours of work time. A Principal does Phase 4 (32 hours) plus the work version, about 62 hours.

You areStart atSkipYour assignmentsYour milestone scopeAim for
Fresher (under a year of work; one language, Git, HTTP)Week 0, then Week 1Nothing. Do all the weeks, and plan 20Practice projects 1 and 2 in each phase, and checkpoints 1 to 4Core scope: 3 nodes, a consensus library for the log, you build everything around it2 Sound on every dimension
Senior (5 to 8 years, ships services)Week 2 (skim Week 1)Week 0All practice projects, checkpoints 1 to 4Adds your own Raft (election and replication), snapshots and ADRs 5 to 73 Proven on every dimension
Staff (10+ years, sets direction)Weeks 3, 4 and 9 to 16Weeks 1, 2 and 5 to 8Checkpoints 2, 3 and 4, and the work version (Checkpoint W)Adds a failure-testing policy, a consistency decision guide and a cost model3 on all, 4 on two
Principal (15+ years, sets strategy)The milestone section, Phase 4, then Interview topicsThe weekly tables of Phases 1 to 3The work version on a real system, plus a checklist others reuseAdds a decision on consistency and replication for several teams, with a measured outcome4 on at least two

Prerequisites for a fresher. If one is missing, do Week 0 first.

  • You can write a program with structs or classes, maps and loops in Go, Java or another language.
  • You can use a terminal and Git.
  • You know what a TCP connection and a JSON message are.
  • You know what a thread and a lock are.

Career change (for example a backend developer moving to infrastructure): start at Week 1. Thinking in failures and in orderings of events is the biggest difference from single-node work.