الفهرس 
IntroductionOnly 14 pages are availabe for public view
 |  | from | 86 |  |  |
| | | from | 86 | | |
Abstract
Networks nowadays are playing an integral role in how people interact and communicate,
as well as how businesses operate and offer services to end customers. With the growth
of user demands and high data rates, the ability to manage this data traffic and utilize the
network resources has become a pivotal aspect for future networks (e.g., Fifth Generation
(5G)). The evolution to 5G system and services is a key enabler of the networked
society paradigm where all persons and things will be connected everywhere, every time,
using any type of devices and technology. The 5G technology is presently in its early
research stages and expected to merge various types of advanced paradigms which make
5G technology more powerful, one of these paradigms is Software-Defined Networking
(SDN).
SDN is a recent technology that established a new method for creating and administering
networks. It is a new architecture which has been designed to enable more agile
and cost effective networks. It has a significant role in network optimization and performance
improvement and has been seen as a promising approach for the vision of 5G
networks, which will likely play a crucial role in the 5G networks design. A critical
understanding of this emerging paradigm is necessary to address many challenges of
the future SDN-enabled 5G technology such as network security, controller design, resource
management, network monitoring and measuring, etc. In this thesis, we propose
SDN-based resource management algorithms for future 5G networks specially server
load balancing. Since the traditional load balancers suffer from inflexibility and difficulty
to manage network flows and they are very expensive hardware dedicated devices,
SDN offers an inexpensive, a flexible, and a promising solution for traditional load balancer
limitations. The main SDN concept is separating both the data and control planes
allowing network programmability and centralized control for the network. Since the
current traditional network cannot endure more user needs, we apply different load balancing
strategies to distribute the user requests among a number of servers.
iii
This thesis aims to develop OpenFlow-based approaches for dynamic server load balancing
in SDN environment. Hence, we propose “Bandwidth-based”, “CPU-based”, and
“CPU-Memory-based” load balancing techniques using Ryu OpenFlow controller that
distributes the network requests among multiple servers based on the servers’ state. The
performance of the proposed schemes is evaluated and compared with different traditional
load balancing strategies (Random-based, Round-robin, and Connections-based)
under a mininet emulation and a Raspberry Pi-based implementation. Emulation and
hardware implementation results show that the proposed schemes achieve more reliability
and higher resource utilization compared to the traditional load balancing strategies; in
addition, they have a lower response time and higher throughput. The proposed schemes
exhibit more scalability and low-cost characteristics.