Performance evaluation of parallel TCP, and its impact on bandwidth utilization and fairness in high-BDP networks based on test-bed

TL;DR

This study evaluates the performance of various TCP variants in high-BDP networks using a new parallel scheme, demonstrating improved bandwidth utilization and fairness compared to traditional single TCP schemes.

Contribution

It introduces a parallel TCP scheme for high-BDP networks and compares multiple TCP variants, highlighting its advantages over single TCP in bandwidth utilization and fairness.

Findings

Parallel scheme outperforms single TCP in bandwidth utilization and fairness.

CUBIC achieves the best performance among tested variants.

Parallel TCP scheme enhances network resource utilization.

Abstract

After the presence of high Bandwidth-Delay Product (high-BDP) networks, many researches have been conducted to prove either the existing TCP variants can achieve an excellent performance without wasting the bandwidth of these networks or not. In this paper, a comparative test-bed experiment on a set of high speed TCP variants has been conducted to show their differences in bandwidth utilization, loss ratio and TCP-Fairness. The involved TCP Variants in this experiment are: NewReno, STCP, HS-TCP, H-TCP and CUBIC. These TCP variants have been examined in both cases of single and parallel schemes. The core of this work is how to evaluate these TCP variants over a single bottleneck network using a new parallel scheme to fully utilize the bandwidth of this network, and to show the impact of accelerating these variants on bandwidth utilization, loss-ratio and fairness. The results of this work reveal that, first: the proposed parallel scheme strongly outperforms the single based TCP in terms of bandwidth utilization and fairness. Second: CUBIC achieved better performance than NewReno, STCP, H-TCP and HS-TCP in both cases of single and parallel schemes. Briefly, parallel TCP scheme increases the utilization of network resources, and it is relatively good in fairness.