How Google migrated billions of lines of code from Perforce to Piper

In 2011, Google faced the challenge of scaling its source control system, which relied on a single Perforce server that had been in use for over a decade, supporting thousands of users and millions of commands daily. Despite the server's remarkable performance, it was stretched thin, prompting Google engineers to explore alternatives. The decision to maintain a monolithic repository was significant, as it set industry standards for handling code complexity at scale. After rejecting alternatives like SVN due to migration issues, Google developed Piper, a distributed system built on Google infrastructure like Bigtable, to replace Perforce. This migration, which took over four years, was complicated by dependencies on the Perforce API, leading to a clean room design approach to avoid legal risks. The transition to Piper was critical, as it was an all-or-nothing move, requiring complete migration before Google could benefit from the new system. The successful cutover in 2012 reduced Google's operational risk and enabled new systems, highlighting a period of innovative and daring engineering at Google.