repository
Performance Enhancement and Quality of Service Improvement in Software Defined Networks
- Author(s)
- ALI JEHAD
- Advisor
- Byeong-Hee Roh
- Department
- 일반대학원 컴퓨터공학과
- Publisher
- The Graduate School, Ajou University
- Publication Year
- 2020-02
- Language
- eng
- Alternative Abstract
- The software-defined networking (SDN) paradigm is predicted to be an exquisite choice for future communication and Internet technologies. However, the quality of service (QoS) and performance is affected through various factors in SDN. Therefore, in this dissertation, we investigate three factors that influence the QoS in SDN and we provide solutions for improving the QoS. In SDN the controller plays a prominent role because it intelligently manages the underlying network. However, there are multiple controllers available with their distinct features and each controller offers a different QoS. Further, one controller is not enough for end-to-end (E2E) communication in the presence of multiple providers because each provider offers different QoS metrics therefore E2E QoS provisioning in such a scenario is a challenging task. Moreover, another factor influencing the QoS is the frequent transfer of messages among the switches and the controllers, consequently, the assignment of the switches to the controller also influences the QoS in SDN. Keeping in view these factors, in this dissertation, we focus on the QoS improvement and performance enhancement in SDN. Firstly, we explore the SDN controller's features which influence the performance of the controller. Further, we categorize these features for each controller. Then, we formulate the problem of controller’s selection with a multi-criteria decision-making (MCDM) scheme i.e. analytical network process (ANP). Finally, we provide a novel framework [1] for checking the QoS of the SDN controllers in real-Internet and brite created topologies. The high ranked controller computed with ANP is evaluated with respect to various QoS metrics such as topology discovery, delay, throughput, delay and throughput with high traffic load and central processing unit (CPU) utilization in the real-Internet and brite generated topologies. The controller through the feature-based and performance-based study surpasses other controllers in performance. Secondly, we examine the applicability of the hierarchal SDN architecture to the E2E service class mapping in the presence of heterogeneous networks using a multi-objective decision-making scheme technique for order of preference by similarity to ideal solution (TOPSIS). As the traditional Internet and prevalent SDN schemes lack an effective QoS class mapping mechanism for the E2E QoS guarantee. Therefore, we first explore different SDN architectures for E2E collaboration among the domains. Moreover, we propose a packet processing algorithm and make a network model in the hierarchal control plane for service class mapping. Then, we formulate the problem of service class selection using TOPSIS based ideal service class mapping scheme in SDN. Finally, we demonstrate the proposed approach with an example. Our proposed approach demonstrated with simulations suffices the E2E QoS of various application service requests. Thirdly, we analyze the controller placement problem in SDN with emphasizing on the assignment of switches to ultimately reduce the E2E delay. The frequent communication among the switches and the SDN controllers influence the performance in SDN. We explore different factors that contribute to E2E delay such as path discovery and E2E delay. Moreover, we calculate the delay in the real-Internet topologies from the emulated Mininet environment. Then, we adopt an efficient clustering algorithm to reduce the E2E delay among the controllers and their associated switches. Finally, we make a comparison of the E2E delay for the controller locations with standard K-mean controller placement.
- Fulltext
- Appears in Collections:
- Graduate School of Ajou University > Department of Computer Engineering > 4. Theses(Ph.D)
- Files in This Item:
- There are no files associated with this item.
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.
-
- License
- STATISTICS
- Total Visit :3,729,606
- Total Download :1,818
- Today View :7,264
