Inside GitHub’s new Git storage architecture
Summary
GitHub unveils Rubicon, a new Git storage architecture designed to support thousands of concurrent agent writes per second while keeping reads cheap. It separates durable data in Azure Blob Storage from a replicated log (Ref-CAS) that orders reference updates, uses virtual replicas for reads and writes, and runs background compaction on dedicated nodes. The design aims to scale across repositories with improved throughput, lower latency, and a path to further optimizations like partial fetches and global replicas.