Modelling and Analysis of Large Buffer Probe for B3R Congestion Control Algorithm

Abstract

Bottleneck Bandwidth and Round-trip time (BBR) congestion control is the first congestion control to claim to operate at Kleinrock’s optimal operating point without filling up queues and high packet loss. Recent research on BBR revealed that inter protocol unfairness was observed in terms of bandwidth share due to the insignificant queue share occupied by loss based Congestion Control Algorithm (CCA) flows when bottleneck buffer size shared by the flows were less than two times the product of bandwidth and delay (BDP) for the bottleneck link. Also, when buffer sizes exceeded twice its BDP, the longer Round Trip Time (RTT) BBR flows dominated over shorter RTT BBR flows by grasping a greater portion of bottleneck link bandwidth (intra protocol unfairness). As inter protocol unfairness can cause other TCP variants to lose throughput and CUBIC is still one of the most widely used CCA’s, it is important to establish fairness. Inter protocol unfairness can currently allow latency cheating which allows individuals to add artificial latency to get a better share of bandwidth. In this paper, we present Bottleneck Bandwidth Buffer and Round-trip propagation time (B3R) which uses a modified BBR Congestion Control Algorithm (CCA) when operating under the above mentioned bottleneck buffer sizes to reduce inter/intra protocol unfairness. The results obtained after testing the algorithm are very encouraging; B3R improved inter protocol fairness by increasing CUBIC flows from 3% of bottleneck link bandwidth up to 31% and ensured complete intra B3R fairness for up to 4x RTT differences. B3R also helped reduce queue formation on the bottleneck by up to 50%