Improvement of RTT Fairness Problem in BBR Congestion Control Algorithm by Gamma Correction

Abstract

Google proposed the bottleneck bandwidth and round-trip propagation time (BBR), which is a new congestion control algorithm. BBR creates a network path model by measuring the available bottleneck bandwidth and the minimum round-trip time (RTT) to maximize delivery rate and minimize latency. However, some studies have shown that there are serious RTT fairness problems in the BBR algorithm. The flow with longer RTT will consume more bandwidth and the flows with shorter RTT will be severely squeezed or even starved to death. Moreover, these studies pointed out that even small RTT differences will lead to the throughput of BBR flows being unfair. In order to solve the problem of RTT fairness, an improved algorithm BBR-gamma correction (BBR-GC) is proposed. BBR-GC algorithm takes RTT as feedback information, and then uses the gamma correction function to fit the adaptive pacing gain. This approach can make different RTT flows compete for bandwidth more fairly, thus alleviating the RTT fairness issue. The simulation results of Network Simulator 3 (NS3) show that that BBR-GC algorithm cannot only ensure the channel utilization, but also alleviate the RTT fairness problem of BBR flow in different periods. Through the BBR-GC algorithm, RTT fairness is improved by 50% and the retransmission rate is reduced by more than 26%, compared with that of the original BBR in different buffer sizes.

BBR-With Adaptive Pacing Gain Rates (BBR-APG): An improved TCP-BBR Congestion Control Algorithm on RTT Fairness Design Problem

“…To our knowledge, all published BBR variants blindly apply the fixed pacing rates of 1.25 and 0.75 when probing up for more available bandwidth and during probe down intervals, respectively, in the probebw phase. The abovedescribed pacing rates generate a persistent queue based on our experimental results and other studies' evaluations analysis [8,9]. Short RTT will be limited beforehand and progressively release bandwidth to the larger RTT flows since long RTT traffic has bigger RTprop values; therefore, their cwnd will be greater than short RTT flows'.…”

“…After Google published the TCP-BBR, several studies have evaluated its performances under various network setups and conditions [15][16][17], while others have tried to improve various aspects of the BBR [8,[25][26][27][28].…”

“…Fig. 1 A relationship between delivery rate, RTT, buffer inflight data and operating points based on [11] BBR has four main phases to carry out its principal function: sending data at the fastest rate possible without creating a queue, as shown in Figure 2 [8,14].…”

“…Contrary to traditional TCP congestion control variants, BBR regularly measures Btlbw and RTprop of the network traffic and utilizes the measured results to adjust its transmission pattern. Therefore recording higher throughput with minimum transmission delay [8]. In addition, BBR uses bandwidth-delay product (BDP) estimates to reduce bufferbloat occurrences rather than applying a notation of packet loss as an indicator of a congested network.…”

Improvement of BBRv2 Congestion Control Algorithm Based on Flow-aware ECN

“…Mathis et al [7] claimed that BBR opens a new era in congestion control and obsoletes Jacobson88. However, some evaluations [12][13][14][15][16][17][18] found that there are some problems in BBR, such as the unfairness when BBR and Reno/CUBIC share bottlenecks, the intraprotocol fairness, and RTT fairness. For the problem of the BBR algorithm, researchers are also making continuous efforts to obtain better performance by modifying some control parameters of BBR [19][20][21][22][23][24].…”