A software defined network framework in 5G wireless systems

According to Cisco's Visual Networking Index Global Mobile Data Traffic Forecast Report, it is estimated that the global mobile data traffic will increase sevenfold between 2016 and 2021. While, wireless data traffic will grow faster than wired/fixed data traffic (connected through Ethernet) and reach to 67 percent of overall IP traffic by 2020. This aggressive trend from wired to wireless communication technologies also leads a huge increase in the operational, administrative and maintenance activities costs in the current core network deployments. In order to meet these challenges of the current wireless networks deployments (WiMax, LTE), the technologies such as Self-Organizing Networks, Heterogeneous Networks (HetNets) have been proposed and investigated thoroughly in the recent years by academia and industry. However, none of these solutions is solely sufficient to meet the huge upcoming data traffic demand by users globally. Thus, the 3rd Generation Partnership Project (3GPP) focuses on completing standardization of 5G technologies. To this end, 5G wireless systems aim to satisfy bandwidth-thirst of users and industry by improving data-rates, network coverage and latency. In this manner, it is foreseen that 5G systems should rather combine and utilize currently deployed wireless technologies than target for a particular Radio Access Technology (RAT) to accomplish this crucial goal. However, this proposal brings up high heterogeneity in traffic demands and various mobility patterns that further complicates the operation and the management of the network. Meanwhile, SDN is proposed to ease the management of ossified networks by separating the control mechanisms from the actual network hardware where communication data is processed and forwarded. It also boosts the opportunity to implement and test new protocols without vendor intervention. However, since SDN is initially developed for wired networks, integration of SDN with various wireless domains brings out unique challenges that should be handled individually. With the consideration above in mind, this thesis aims to solve the unique challenges of wireless technologies consisted by 5G systems and evaluate the benefits of integrating SDN approach will be to these systems by emphasizing the basic features of SDN. The problems meet in the thesis are defined under two categories such as flow and energy management. First, the proposed flow management solutions are conducted by using manipulating admission control and routing protocols. As mentioned earlier, towards 5G system, it is planned to utilize all possible radio access technologies in order to support reliable services and applications.As the initial part of the thesis, a SDN test-bed with an access management protocol is proposed for Vehicular Ad-hoc Networks. Also, the inclusion of already-deployed Wi-Fi networks in VANET topology is seen as a crucial step for the next generation vehicular networks. However, the VANET topology also requires to preserve the features already offered by dedicated short-range communication (DSCRC) and the core cellular network. To this end, Software Defined Network (SDN) provides a promising opportunity to simplify the management and the control of clumsy network infrastructures by decoupling data and control plane in order to provide elasticity for current networks. Thus, we propose an architectural model which exposes this opportunity in order to enhance VANET with Wi-Fi access capability. Moreover, we offer a novel Software Defined (SD) VANET architecture which consists of soft OpenFlow switches with WiFi capabilities both as Road Side Units and Vehicles. In particular, we first investigate existing test tools and environments for software defined wireless networks and also supply a novel test-bed architecture in order to provide feasible test environment for evaluating the proposed architecture. Lastly, a Wireless Access Management (WAM) protocol is proposed to provide wireless host management and basic flow admission with respect to the available bandwidth to validate the capability of the offered architecture. The basic observation results of the deployed testbed prove the conformity of the offered 802.11 architecture to the VANET. Additionally, we develop a Mininet-based Software-Defined Network (SDN) simulation environment that improves the total flow throughput and scalability of the overall network. Mininet is chosen as a suitable and manageable tool to implement the SDN based flow admission control module in order to configure the entire topology since it has already built in OpenFlow switches and virtual controllers. This open-source platform is also easily configurable via its drag and drop capabilities. Furthermore, for the Control Plane, OpenDaylight controller is used in order to simulate flow admission control module that fairly admits flows into the OpenFlow switches. OpenFlow version 1:3: for communication between separated Data and Control plane and Linux based operating system to build Mininet 2:1:0 are deployed in the simulator environment. From routing protocol point of view to flow management challenge, first a simple test-bed is implemented for Software Defined Wireless Networks as follows. After that, an environment-specific routing protocol is developed for SD-UAV networks. Raspberry Pi's are used as OpenFlow (OF) Switches in the implemented real-time testbed for SDWN. The testbed provides an opportunity to practical development and testing environment for SDWNs. Moreover, we use OpenDayLight to observe the flows and events in the network. With OpenDayLight integration, we easily provide detailed analysis results for any testing process. Also, we develop a traffic aware routing algorithm add-on which manages the flows w.r.t. their QoS requirements. Finally, we physically implement the proposed routing algorithm to our testbed and validate the improvements of traffic aware routing such as controller response time and network utilization. After SDWN test-bed, next part of the thesis focuses on the design of a resilient end-to-end connectivity paradigm under unique architectural and scenario assumptions for Unmanned Aerial Vehicular networks which have capabilities to extend wireless access for devices without infrastructure coverage, and also help establishing a connectivity backbone during military reconnaissance and disaster events. In this manner, first, we assume that the UAVs themselves are equipped with multiple interfaces that use standardized protocols, with associated variation in data throughout, range, and bit error rates. Second, there may be adversarial agents seeking to disrupt connectivity through targeted jamming in 3D spaces. Third, we assume an overlay software defined control plane, where the UAVs function as software switches, able to execute forwarding commands and determine preferred routes under controller directives. Our proposed solution devises metrics that influence the choice of the wireless interface and weights edges formed between UAV pairs. Furthermore, it also uses a multi-layer graph model and creates maximally separated paths in 3D space to ensure resiliency to jamming. Simulation results conducted for urban scenarios reveal 34% improvement in enhanced resiliency for end-to-end outages by trading off 12% increase in latency over competing approaches. Lastly, energy management problem is investigated in the thesis for cellular networks. As mentioned earlier the wireless user demands dramatically increase everyday, and in order to meet increased traffic demands of indoor users, small cell access point (eNodeB) deployment has been introduced in recent 3GPP LTE-A releases (v.8 and further). However, plug and play nature of the eNodeBs lets a dense deployment of eNodeBs and easily cause over-provisioned network infrastructure. Therefore, the consumed power of network topology is extremely increased. While most of the studies aim to optimize power consumption on single traffic types, in this part of the thesis we provide a novel framework which consists of multi-traffic types approach corresponding power adjustment as well as Quality of Service (QoS) requirements of end-users. In order to gain more power saving by considering QoS of end-users, we focus on the energy efficiency vs. spectrum efficiency trade-off. In this manner, we propose a novel Energy Efficiency Coefficient (EEC) in order to consider QoS of end-users and channel utilization parameters while degrading power consumption. This parameter covers conventional bits= joule energy efficiency metric (power part) and additionally contains parameters such as channel utilization and Carrier-to-Interference Ratio (CIR) (QoS part). Moreover, a novel Resource Auction module is proposed that simply assigns resource blocks to low-utilized cellular networks ignoring interference, because of already taking care in proposed EEC formula as CIR. Furthermore, the p2p delay is defined in terms of different traffic types. The traffic types are modelled by using Queuing Theory that enables parametrize p2p delay in terms of utilization effect in proposed EEC formula. Therefore, p2p delay is defined in closed-form expression that provides transition from empirical p2p delay to analytic p2p delay. We also define an optimization problem which employs novel EEC parameters to determine new power levels (8-levels) for eNodeBs configurations with less power consumption. To solve this optimization problem, we introduce a novel heuristic called "Multi Level Branch and Bound Algorithm" (MLBB). MLBB algorithm contains two nested branch and bound algorithms which manipulates end-users differently. In order to measure performance of proposed EEC, the comparison with Sleep mode (simply power-off non-utilized eNodeBs without considering QoS of end-users), and conventional EEC (bits/joule formula without considering CIR that affects QoS of end-users) is preferred. Consequently, the proposed EEC serves end-users with 67% better QoS than the Sleep Mode and 6,2% than Conventional EEC by waiving from power consumption in acceptable rate (loss about 2,5% compared with conventional EEC) because of including utilization effect in proposed EEC formula and taking care QoS of end-user in optimization process (especially two functions between lines 5-28 in MLBB algorithm).