الفهرس 
Chapter 1: IntroductionOnly 14 pages are availabe for public view
 |  | from | 156 |  |  |
| | | from | 156 | | |
Abstract
Cloud computing has grown rapidly over the past few years. Cloud users expect and demand to run their applications with the highest performance, lowest cost, and best Quality of Service (QoS). Traffic engineering techniques are widely deployed in the cloud data center to manage users’ requests and distribute application traffics among available datacenter resources. Software-Defined Networking (SDN) is a highly flexible architecture that automates network configuration using a centralized software controller to overcome the limitations of traditional architecture and manual configuration for every network device.
This research work tackles the problem of cloud datacenter resources management and proposes three different algorithms to enhance resources management using the SDN.
The first proposal is a dynamic load balancing algorithm using SDN. It handles different application types in real time based on traffic type and the required QoS. The proposed method aims to bridge the gap between application requirements and the available resources in the cloud datacenter. It selects the best path for data transmission and the best server to process the users’ requests according to traffic type for efficient resource utilization, minimizing response time, and maximizing throughput. The simulation results show that the proposed method utilizes the datacenter resources more efficiently compared with the current load-balancing techniques and enhances the performance of the running applications.
The second proposal is an SDN-based monitoring algorithm that uses the Support Vector Machine (SVM) algorithm to monitor the performance of the distributed applications in the cloud datacenter with the aim of detecting anomalies. It collects the data from the network devices and calculates performance metrics for the distributed application components, which are used to train the SVM algorithm. The SVM employs binary classification and multi-class classification algorithms in two stages. The proposed algorithm does not require any knowledge about the running applications and does not rely on static threshold values to measure performance. Simulation results show that the proposed method can detect and locate failure occurrences efficiently with high accuracy and low overhead compared to a decision tree, Naive Bayes, and logistic regression machine learning methods.
The third proposal is an Intrusion Detection System (IDS) based on SDN. The proposed approach integrates adaptive boosting and cost-sensitive techniques to optimize the detection rates of rare attacks without compromising the detection accuracy of familiar attacks. Cost matrix values for various attacks are optimized using grid search and genetic algorithms. The integration of AdaBoost and cost-sensitive aims to build a classification model that minimizes the total number of high-cost errors caused by incorrectly classifying attacks. Experiments were conducted on the CSE-CIC-IDS2017 and CSE-CIC-IDS2018 datasets and performance is evaluated using the macro F1-score and the geometric mean of recall (G-mean). Simulation results prove that the proposed algorithm efficiently detects rare intrusions.