Yandex open-sources YaFF, a technology that can reduce server CPU usage by up to 20%

Yandex has open-sourced YaFF (Yet Another Flat Format), a new technology for efficient data transmission and processing in high-load services. YaFF allows applications to access data without spending resources on deserialization, helping reduce computing costs by up to 20%. The release is part of Yandex's ongoing commitment to sharing infrastructure technologies with the developer community.

Modern services and applications typically receive data from networks or storage systems in a compact serialized form. One of the most widely used standards for this purpose is Protobuf. While reliable and efficient for data transmission, Protobuf requires applications to deserialize data before it can be accessed. In large-scale systems, that process can consume up to 10% of a service's computing resources. An alternative solution, FlatBuffers, allows applications to read data without deserialization, but adopting it often requires extensive code changes. 

YaFF (Yet Another Flat Format) was developed to address this trade-off. The technology works alongside standard Protobuf infrastructure, allowing applications to read data directly without deserialization. This makes it possible to introduce YaFF into existing projects and reduce computing overhead without rewriting service code. The approach can be useful for banks, marketplaces, telecom operators, and cloud service providers — any organization where data-processing speed and infrastructure efficiency are critical.

The technology is already used in Yandex's advertising platform. In an environment that processes hundreds of thousands of requests per second, YaFF reduced processor load by 10–20%. The freed-up resources are used to handle more user requests without expanding the company's server fleet.

The YaFF source code and documentation are available on GitHub.

YaFF joins Yandex's growing portfolio of open-source projects, including Perforator, a continuous profiling system for data centers; YDB, a distributed SQL database; YTsaurus, a platform for distributed data storage and processing; SPQR, a PostgreSQL sharding system; and userver, a framework for high-load applications.

Sharing technologies and engineering expertise with the broader developer community remains an important focus for Yandex. The company welcomes community feedback and contributions to help guide the future development of YaFF.