In section Releases

Yandex Releases YaFF to Cut Server CPU Costs by 20 Percent

For high-load digital platforms, the constant overhead of deserializing data can drain massive computing resources. Yandex has now open-sourced YaFF, a new format designed to bypass this bottleneck, potentially saving large-scale services hundreds of millions of dollars annually by allowing direct data access without the need for extensive code migration.

Yandex Releases YaFF to Cut Server CPU Costs by 20 Percent

Modern backend systems rely heavily on technologies like Protobuf to transmit data between servers and applications. While effective, Protobuf requires a deserialization step that can consume up to 10 percent of a server's CPU capacity. While alternatives like FlatBuffers exist, they often force engineering teams into complex and costly infrastructure migrations to implement.

YaFF, or Yet Another Flat Format, functions as a drop-in solution that integrates with existing Protobuf schemas and APIs. By enabling applications to read data directly, it removes the deserialization tax entirely. Yandex has already validated the technology within its own advertising platform, which processes hundreds of thousands of requests every second. Internal testing showed a 10 to 20 percent reduction in CPU load, effectively expanding system capacity without adding hardware.

Developers can now access the project under an Apache 2.0 license on GitHub. This release joins a growing suite of Yandex open-source infrastructure tools, including the YDB database, the YTsaurus processing platform, and the userver framework, all designed to optimize large-scale production environments.

Share:on TelegramXFacebook

Subscribe to our newsletter

Once a week — the best stories from our editors, no ads or push notifications. Delivered Sunday morning.

Comments (0)

Leave a comment

No comments yet. Be the first!