The Commit Log

Some of the most interesting reliability work in distributed systems starts with a simple idea: if a system can fail in production because of time, scheduling, networking, or unlucky interleavings, then those forces should be brought into tests under control.