Parallel TCP is one possible approach to increasing throughput of data transfer in Long Fat Networks (LFNs). Using parallel TCP is something of black art. As high-speed transport-layer protocols appear, e.g. HSTCP, it is necessary to reinvestigate the performance of parallel TCP, because a choice has to be make among them for the system. In this paper, the performance of parallel TCP is evaluated by mathematical analysis based on a simple dumbbell topology. Packet drop rate and aggregate goodput are used as two metrics to characterize the performance of parallel TCP. Two cases, namely synchronization and non-synchronization, are analyzed in detail when DropTail is deployed on routers. The synchronization case is common in using parallel TCP, but the goodput deteriorates seriously. The non-synchronization case may benefit parallel TCP, but extra mechanisms are required, and it is not easy to implement in the real world. The problem also remains even if Random Early Detection (RED) queue management is employed on routers. The analysis results show the difficulty in using parallel TCP in practice.

As computer hardware components are achieving greater speeds, network link bandwidths are becoming wider. A number of enhancements to TCP have been developed in order to fully exploit these improvements in network infrastructures, including TCP window scale option, SACK option, and HighSpeed TCP (HSTCP) modifications. However, even with these enhancements, TCP cannot provide satisfactory performance in high-speed long-delay networks. As a means addressing this problem, gentle HighSpeed TCP (gHSTCP) has been proposed in [1]. However, its effectiveness has only been demonstrated in simulation experiments. In the present paper, a refined gHSTCP algorithm is proposed for application to real networks. The performance of the refined gHSTCP algorithm is then assessed experimentally. The refined gHSTCP algorithm is based on the original algorithm, which uses two modes (Reno mode and HSTCP mode) in the congestion avoidance phase and switches modes based on RTT increasing trends. The refined gHSTCP algorithm compares two RTT thresholds and judges which mode will be used. The performance of gHSTCP is compared with TCP Reno/HSTCP and parallel TCP mechanisms. The experimental results demonstrate that gHSTCP can provide a better tradeoff in terms of utilization and fairness against co-existing traditional TCP Reno connections, whereas HSTCP and parallel TCP suffer from the trade-off problem.

Continuous and explosive growth of the Internet has shown that current TCP mechanisms can obstruct efficient use of high-speed, long-delay networks. To address this problem we propose an enhanced transport-layer protocol called gHSTCP, based on HighSpeed TCP proposed by Sally Floyd. It uses two modes in the congestion avoidance phase based on the changing trend of RTT. Simulation results show gHSTCP can significantly improve performance in mixed environments, in terms of throughput and fairness against the traditional TCP Reno flows. However, the performance improvement is limited due to the nature of TailDrop router, and the RED/ARED routers can not alleviate the problem completely. Therefore, we present a modified version of Adaptive RED, called gARED, directed at the problem of simultaneous packet drops by multiple flows in high speed networks. gARED can eliminate weaknesses found in Adaptive RED by monitoring the trend in variation of the average queue length of the router buffer. Our approach, combining gARED and gHSTCP, is quite effective and fair to competing traffic than Adaptive RED with HighSpeed TCP.