The release of the platform Libreqos 1.4 has been published. This platform is designed to organize a fair distribution of the existing bandwidth between users and reduce negative effects caused by intermediate package buffering (Bufferbloat) with network equipment. It can be used by providers or administrators of private networks to optimize traffic flows, maintain delays at a minimum level, and distribute the bandwidth taking into account priorities. The project code is written in SI, Python, and Rust, and is distributed under the GPLV2 license. The project is being developed under the leadership of Dave Takhta (Dave Taht), the co-founder of the project Bufferbloat, the creator of the Cerowrt distribution, and the author of numerous RFCs related to the processing of network queues.
LibreqOS allows users to reduce delays and increase the reliability of interactive sessions, games, online teaching platforms, VOIP traffic, and video calls even under heavy network load. For example, it can address issues such as video stuttering when someone on the same network begins downloading a 4K video. While the use of LibreqOS reduces the peak bandwidth available to one user, it significantly reduces delays and fairly distributes resources between all data exchange participants. In tests, the use of LibreqOS reduced delays when receiving data from 106 to 9 ms and when transferring from 517 to 23 ms. However, the continuous load speed decreased from 74 to 25 MBPS, and transfer speed decreased from 29 to 8 MBPS.
The foundation of Libreqos lies in the use of the CAKE network queues (Common Applications Kept Enhanced), the FQ_CODEL (Fair Queuing Controlled Delay) control mechanism, and the use of EBPF and XDP to perform handlers at the network of networking faith with the possibility of direct access to DMA package buffer. The CAKE algorithm replaces and simplifies the complex hierarchy of package processing disciplines, allowing for maximum throughput and minimizing delays even on slow communication channels with the provider, as well as when working on low-power devices.
LibreqOS also provides means to track Round-Trip Time (RTT), individual user references, access points, and sites. To analyze network conditions, a web interface has been developed, which enables clear evaluation of network traffic, tracking changes in load and delays