Distributed-systems training expanded

A hands-on course for distributed systems has expanded from 15 tracks to 28, with 343 tasks spread across eight categories. (x.com) The project is “Building Distributed Systems from Scratch,” a free site by Mohit Mishra that teaches engineers by having them implement real networked systems instead of reading lectures alone. The site previously described its curriculum as 15 tracks and 75-plus challenges. (builddistributedsystem.com 1) (builddistributedsystem.com 2) Distributed systems are programs split across multiple machines, and they fail in ways single-computer programs do not: messages arrive late, machines crash, and two copies of the same data can disagree. Mishra’s site says its exercises run against simulated failures such as network partitions, message loss, and timing faults. (builddistributedsystem.com 1) (builddistributedsystem.com 2) The original version grouped lessons into four categories: foundations, consensus, scalability, and advanced topics. Those tracks covered remote procedure call protocols, gossip, conflict-free replicated data types, leader election, Raft, caching, load balancers, queues, sharding, distributed transactions, and MapReduce. (builddistributedsystem.com 1) (builddistributedsystem.com 2) The new release adds breadth and depth at the same time. Mishra said version 2.0 now spans eight categories and 28 tracks, and he separately published a 15-task track focused on remote procedure calls, retries with backoff, idempotency, and chaos testing. (x.com) (x.com) Those four ideas sit near the center of production systems. Remote procedure calls let one service ask another service to do work over a network; backoff spaces out retries after failures; idempotency makes repeated requests safe; chaos testing injects faults on purpose to see what breaks. (builddistributedsystem.com) (github.com) Mishra says the project draws on three well-known teaching and testing references: Maelstrom by Jepsen, Fly.io’s Gossip Glomers challenges, and the Massachusetts Institute of Technology course 6.824. The site says learners progress from simple message passing to full Raft consensus with log replication. (builddistributedsystem.com) (builddistributedsystem.com) The audience is software engineers who already know how to write single-machine programs and want practice with infrastructure problems that show up in systems such as Kafka, etcd, and CockroachDB. The site says the material is free to use. (builddistributedsystem.com) The update turns a small challenge set into something closer to a full curriculum. The pitch has not changed: learn distributed systems by building them, then watch them fail under test before they fail in production. (x.com) (builddistributedsystem.com)