Performance-Adaptive Prediction-Based Transport Control over Dedicated Links

Several research and production networks now provide multiple Gbps dedicated connections to meet the demands of large data transfers over wide-area networks. End users, however, have not been able to see corresponding increase in application goodputs mainly because (i) such rates have pushed the bottleneck from the network to the end system, and (ii) the traditional transport methods are not optimized for handling host dynamics. Due to the sharing with unknown background workloads, the data receiver oftentimes lacks sufficient system resources to process packets arriving from high-speed dedicated links, therefore leading to significant packet drops at the end system. We propose a rigorous design approach for a new class of transport protocols that explicitly account for the dynamics of the running environment to maximize application goodputs over dedicated connections. The control strategy of the proposed transport method combines two aspects: (i) the receiving bottleneck rate is predicted based on performance modeling, and (ii) the sending rate is stabilized at the estimated bottleneck rate based on stochastic approximation. We test the proposed method on a local dedicated connection and the experimental results illustrate its superior performance over existing methods.