A Domain Specific Language for Testing Consensus Implementations

Abstract

Distributed systems have been a crucial component in the growth of large scale databases that power most of modern internet. While the algorithms used to run these systems are proven to be sound, the gap between implementations and on-paper algorithms leaves room for errors. Recurring downtimes in popular services such as Facebook and Cloudflare are mainly due to bugs in these implementations. In this talk, I will present a new approach to define unit tests for testing these distributed protocol implementation akin to unit tests for sequential code. I will describe the design decisions that led to the language and elaborate on some of the bugs found in popular implementations.