TCP Stack Optimization for Network Performance Improvement in Cloud Infrastructure
Students Name: Sheremeta Oleksandr Vitaliiovych
Qualification Level: master (ESP)
Speciality: Telecommunications and Radio Engineering
Institute: Institute of Telecommunications, Radioelectronics and Electronic Engineering
Mode of Study: full
Academic Year: 2022-2023 н.р.
Language of Defence: ukrainian
Abstract: Cloud infrastructures are very important for businesses and organizations nowadays. They enable storing and processing large volumes of data, providing accessibility and operational efficiency. They must ensure a high level of performance and efficiency to ensure fast and uninterrupted access to data and services. With each passing year, the volume of data processed in cloud networks increases, leading to a rise in resource usage. Additionally, many new applications and services emerge that require minimal data transfer delays. This is why the network stack is one of the key components of a cloud infrastructure, responsible for data transmission across the network [1]. Its optimization is necessary to ensure a high level of performance. Optimizing the TCP stack in cloud infrastructures can reduce the risk of network attacks, such as DDoS attacks, and provide better protection against malicious actors. It can also help reduce the cost of deploying and maintaining cloud infrastructure by reducing network load and improving its performance [2]. There are numerous methods available for optimizing the TCP/IP protocol stack. However, not enough attention is given to the research question of optimizing the stack in a cloud infrastructure with the use of automatic deployment of configuration, which is considered in this work. The object of research is the TCP network stack in cloud infrastructures, with automatic configuration deployment on virtual machines for experimentation. The subject of the research is the optimization of the TCP network stack to increase network performance in cloud infrastructures. Goal and objectives of the research. The goal of this master’s thesis is to analyze the issues that affect network performance in cloud infrastructures and optimize the TCP network stack to improve network productivity. In order to achieve the goal of the work, you need to complete the following tasks: 1. To analyze the TCP network stack in a busy infrastructure, and problems that affect network performance. 2. Explore different options for optimizing the TCP network stack and provide optimal settings to improve performance. 3. Investigate the impact of queue processing methods on the optimization of the TCP network stack 4. Experimentally verify the effectiveness of stack optimization methods on real cloud infrastructures. Research Methods. During the work, the analysis of scientific literature was carried out, experimental studies were conducted on a test bench. In order to determine the optimal parameters of the network stack for cloud infrastructure, experimental tests will be conducted under different conditions. Scientific novelty of the research lies in addressing the pressing issue of optimizing the TCP network stack in cloud infrastructure to improve network performance. Modern research methods, including experimental studies, methods of automatic deployment, and statistical experiments will be used in the work. The obtained research results will allow proposing recommendations for optimal configuration of the TCP network stack in cloud infrastructures, which will ensure increased service delivery performance and efficient use of cloud infrastructure resources. Thus, the research is an important step in the development of cloud technologies and improving their effectiveness in modern conditions. The first chapter of the work examines the history of the TCP/IP protocol stack, which allows understanding the importance of this protocol and its impact on modern networks. It also provides an overview of the TCP/IP protocol stack and its functional features, including an analysis of congestion control algorithms in TCP protocols and transport connection management mechanisms. Additionally, an overview of network technologies and protocols in the Linux kernel-based operating system was conducted. In the second chapter of the thesis, cloud infrastructures [3] and their computing models are discussed, including the advantages of using cloud infrastructures. A comparative analysis of queue processing methods [4] in the network stack of cloud infrastructures is conducted. The role of virtualization in cloud infrastructures is also described, including its peculiarities and ensuring data integrity during transmission. In the third chapter of the thesis, the set of software tools used for the research is discussed. Proxmox is described as a cloud environment, and command-line utilities are introduced to allow for testing network throughput, operating system utilization, and identifying system faults. The WireGuard protocol is also introduced as an implementation of VPN technology that will be used as an optimization example in the study. Additionally, optimization of network resources is described using tools such as traffic control and changing kernel parameters in the operating system, as well as selecting the maximum transmission unit (MTU). In the fourth chapter, a study was conducted on the throughput capacity of the Linux kernel-based operating system (Debian 11) with default settings. Time intervals were established when the most load falls on the cloud infrastructure, which affects the speed of data transfer. The impact of different queue types on packet traffic with different content was also determined. The decision to use network resource optimization using traffic control tools with the HTB queue type and changing the kernel parameters of the operating system, as well as choosing the optimal packet transfer size (MTU), was justified.