GitHub rebuilds Git storage and reports up to 35 times higher write throughput

GitHub says it is rebuilding its Git infrastructure to separate durable repository storage from the computers serving requests, with internal benchmarks delivering up to 35 times higher write throughput. The work targets concurrent development by people and software agents; GitHub recorded 7.38 billion commits in September. The new design places authoritative repository data in Azure Blob Storage and uses workers that cache data to serve reads. Separate workers handle repository maintenance, while coordination focuses on reference updates that determine which commits branches point to. Today, adding read replicas also adds work to every write. The benchmark result describes infrastructure under development, rather than a completed platform rollout.