OpenAI has developed Pando, a new Git backend designed to handle high volumes of concurrent writes and maximize read throughput for its monorepo.
By Hermes-Vector AI Desk
OpenAI has announced the development of Pando, a new backend for Git repositories designed to scale with the increasing demands of the company's monorepo. According to OpenAI, the system was built to support the level of concurrent work enabled by Codex without interrupting existing engineering workflows or changing how the team uses Git.
OpenAI reports that starting around the end of last year, the systems supporting its monorepo struggled to keep up with development demands. The company noted it experienced "slow fetches and pushes along with failed merges," with reliability occasionally falling below 90%.
OpenAI attributes these challenges to the increased scale and pace of development driven by Codex, which enables engineers to work in parallel. This increased activity results in more branches, pushes, merges, and fetches from various environments, including research jobs and CI.
To address these issues, Pando was designed with two primary goals: achieving horizontally scalable writes optimized for concurrent pushes to many branches, and maximizing read throughput to saturate available network bandwidth.
Unlike conventional approaches, Pando utilizes a relaxed consistency model known as causal consistency. This allows the system to scale writes by coordinating updates to each branch independently. According to OpenAI, this means "updates to unrelated branches can proceed in parallel," though a reader looking at several branches may see them at different points in time.
To scale reads, Pando reuses packfiles—Git's format for compressing multiple objects—across fetches. This allows the server to serve the same bytes to many clients with minimal computation, rather than generating individual packfiles for every client.
OpenAI states that Pando is compatible with the Git protocol and client, allowing both engineers and agents to adopt the system into existing workflows. In production, OpenAI reports that Pando has "cut median push latency by nearly 5x" and currently delivers "nearly a petabyte of Git data per day."
When a client pushes to a branch, Pando streams the incoming packfile to blob storage and uses a multi-key transaction to record a ref version in a metadata store. During a fetch, Pando selects the stored packs required by the reader and provides them via a service called Express.
Sources
This article was drafted with AI assistance and checked against the sources above. Company claims are reported as claims. Cover image is AI-generated.
The day's checked news, by email
One short email a day with the top stories. Free, and you can unsubscribe in one click.
Or follow for the latest news:
No comments yet. Start the discussion.
Anthropic has introduced Claude Opus 5.5, a new model designed for improved performance in coding, knowledge work, and communication compared to previous versions.
Netflix has launched an experimental 'AI Discover' tab on its TV app, allowing users to find content through AI-generated themes based on their current mood or viewing preferences.
