BBR / BBRv3
Bottleneck Bandwidth and Round-trip time - Revolutionary Congestion Control
Performance
Maximum bandwidth utilization
Fairness
Fair resource distribution
Latency
Minimal transmission latency
BBR Performance Analysis
BBR Performance Analysis
Congestion Control Algorithm Metrics
Data Source: Google Research, Academic Publications
Last Updated: November 2024
Test Conditions: Standardized network simulation
satellite
BBR and CUBIC algorithm performance metrics across various network conditions
Note:
- Data based on standardized network simulation
- BBRv3 shows 12-31% throughput improvement
- Fairness Index: 0-1, where 1 = perfect fair distribution
- Buffer size affects algorithm optimality across different network conditions
What is BBR?
BBR (Bottleneck Bandwidth and Round-trip time) is a congestion control algorithm developed by Google that uses bottleneck bandwidth and round-trip time (RTT) measurements to optimize data transmission. Unlike traditional loss-based algorithms, BBR actively measures network bandwidth and latency.
BBRv3 is the latest version with improved multistream support, additional buffer overflow protection, and enhanced fairness in resource distribution. The algorithm is particularly effective in high-bandwidth, low-latency networks, providing stable performance and minimal jitter.
Publications:
BBR is described in Google publication "BBR: Congestion-Based Congestion Control" and implemented in Linux kernel. BBRv3 is described in Google document "BBR v3: A Modern Congestion Control Algorithm".
BBRv3 Benefits:
Technical Specs
Developer
Algorithm Type
Model-based
Implementation
Linux Kernel, QUIC
Version
BBRv3 (2023)
Evolution
BBR v1
2016 - Исходная версия с использованием bandwidth-RTT модели
BBR v2
2020 - Улучшенная справедливость в сетях с потерями, лучшая координация между потоками
BBR v3
2023 - Многопотковая поддержка, оптимизация для QUIC, улучшенная работа с ECN (Explicit Congestion Notification)
BBRv3 Performance Metrics
Metrics by Buffer Size
* BBRv3 achieves optimal performance at 2*BDP buffer size, balancing throughput and latency