Akademik Veri Yönetim Sistemi
Tezin Türü: Bütünleşik Doktora
Tezin Yürütüldüğü Kurum: Doğu Akdeniz Üniversitesi, Faculty of Engineering, Department of Electrical & Electronic Engineering, Kıbrıs (Kktc)
Tezin Onay Tarihi: 2025
Tezin Dili: İngilizce
Öğrenci: Atefeh Ahmadniai Khajekini
Asıl Danışman (Eş Danışmanlı Tezler İçin): Ali Hakan Ulusoy
Eş Danışman: Enver Ever
Özet:
The demand for wireless networks, in particular mobile communication systems, is increasing considerably day by day. The key goals of wireless networks is to provide enhanced performance in data transmission with high speed and low delay. To satisfy these goals, multi-homed devices are able to access heterogeneous wireless systems simultaneously by using multipath mechanisms. To achieve this, the concept of Multipath TCP (MPTCP) is introduced, enabling the utilization of multiple interfaces for communication. Nowadays, smart devices like mobiles and tablets can simultaneously use multiple interfaces, allowing MPTCP to transmit packets over several subflows at once. This approach results in lower bandwidth consumption and higher throughput compared to traditional single-path TCP. Additionally, MPTCP improves load balancing and enhances resistance to link failure in wireless networks. The path scheduler plays a crucial role in MPTCP by determining which available subflow to use for packet transmission. A well-informed scheduling decision significantly improves MPTCP throughput, while an incorrect one can impact system performance and prevent MPTCP devices from effectively utilizing multiple paths. Furthermore, the diversity of subflows, especially in wireless paths, can cause packet loss issue in the MPTCP, causing an increase in Out-of-Order (OfO) packets occurred in the receiver node. To solve these performance-related challenges, numerous schedulers have been presented to mitigate the issues of OfO packets. However, most existing methods have primarily focused on improving path delay without adequately considering the effect of packet loss problems. This research provides a comprehensive overview of MPTCP schedulers, offering insights into the existing solutions and conducting a detailed evaluation of widely used path schedulers. Subsequently, we propose, implement, and evaluate three novel scheduling models, PRS-MPTCP, EMPTCP, and DynPerf-Scheduler with the specific aim of minimizing OfO packets to improve MPTCP performance in wireless systems. The proposed schedulers take into account various characteristics of each subflow, such as RTT, CWND, and the number of packet losses, to effectively sort available paths in MPTCP at any given time and decide which packets should be assigned to which ordered paths. The first proposed scheduling method, PRS-MPTCP, improves throughput in various scenarios with varying loss rates, reduces the number of OfO packets, and significantly reduces the retransmission rate. The second proposed method, EMPTCP, dynamically adjusts the path scheduling strategy for each available path, reducing the influence of network heterogeneity on MPTCP performance. The third proposed technique, DynPerf-Scheduler, dynamically chooses the best path and performs well in wireless systems with high link failure rates. All suggested schedulers have been tested in several network scenarios, varying the loss rate and RTT ratio. We conducted a comparison with the state-of-the-art schedulers, considering different performance metrics, including throughput, and the number of OfO packets for heterogeneous subflows. These proposed methods overcome the shortcomings of existing path schedulers and increase the performance of MPTCP connections. In lossy networks, our proposed approaches achieve significantly lower number of OfO packets and higher throughput than traditional schedulers.