For cases where data processing speed and hardware efficiency are critical

The Yandex team has made its latest development publicly available — it has published the documentation and code for YaFF (Yet Another Flat Format) on the GitHub platform. This technology is designed for high-load services and allows for optimizing data operations.

The peculiarity of YaFF is the ability to directly read information without the need for unpacking it. Traditional formats, such as Protobuf, require a resource-intensive data unpacking operation, which can consume up to 10% of computing power. YaFF solves this problem by working on top of the Protobuf format.

As emphasized by Yandex, YaFF can be used in existing projects and save computing resources without rewriting the service code.

According to the company, the use of YaFF in Yandex's own advertising system allowed for a 10-20% reduction in processor load when processing hundreds of thousands of requests per second. This makes it possible to process more data without expanding the server infrastructure.