Software-Defined Small Cell-Linked Vehicular Networks: Architecture and Evaluation

Abstract

Vehicle-to-everything services are in the implementation phase, and automakers agree that V2X would improve the safety-critical applications already deployed. 3GPP Release 12 introduces LTE-V for V2V and V2I services. The LTE-V is extended to C-V2X to support V2N. Because of the challenge of high mobility in the V2X system, cutting-edge technologies, such as SDN and small cell in 5G networks, pave the way to the next generation of vehicular networks. SDN is a network technology concept that divides the data and control planes. The OpenFlow protocol is used for communication between the control layer and the network layer in SDN. Different from wireless traditional cellular base stations, small cells are lower-power cell sites that are deployed every few blocks. Small cells can transmit data using mid- and high-band spectrums. Small cell-linked road side unit (RSU) is considered a key enabling technology because it has the capability to create a logical cluster platform residing at the edge of the network, which provides high computation performance. Accordingly, we consider a novel distributed software-defined small cell-linked road side unit vehicular network architecture (diSRsVN). Based on diSRsVN, logical software-defined on-board wireless vehicle, and topology discovery over diSRsVN are presented. The proposed architecture is evaluated under an omnet++ network simulator. The simulation results show the effectiveness of the proposed architecture, which improves the packet delivery ratio and minimizes end-to-end delay.

1. Introduction

The deployment of vehicular applications based on the 5G-vehicle-to-everything (5G-V2X) is the hottest approach within recent years [1]. 5G-V2X is an efficient technology that manages comprehensively solutions for road safety and traffic efficiency by allowing vehicle-to-vehicle (V2V) communication, vehicle-to-pedestrians (V2P), vehicle-to-vulnerable road users/motorcycles via smartphones, road infrastructure (V2I), and mobile network operators (V2N, P2N, I2N) to provide full coverage and continuity of vehicular services [2]. 5G-V2X performs greatly, as the vehicle is out of the road side units (RSUs) coverage but the placement of the RSUs scales the quality of reception of safety messages exchanged between vehicles [3]. In addition, V2X applications use RSUs as relay nodes for V2V, V2P, and V2I [4]. The V2X architecture integrated with vehicular ad hoc network (VANET)-based 5G and RSUs enables a multi-protocol access communication.

5G represents the next major phase of V2X communication. This is due to the fact that such a system will connect everything seamlessly and ubiquitously. Aside from addressing massive internet of things (IoT) deployment, 5G is also expected to meet the requirements of previously unavailable V2X applications that are based on ultra-reliable and low-latency communications [5]. To accomplish this, the 5G research community [6] discussed new technologies, including, among others, the densification of the network with small cells (SCs) [7]. The use of SCs improves spectral efficiency. The LTE-A with SCs densification is positioned as a potential candidate technology for V2I communications. However, frequent handoffs and the increased signaling overhead of LTE-A small cells issues were addressed. For instance, Ahmed et al. [8] proposed a novel selection scheme-based mobile gateway with one-hop clustering to efficiently relay traffic from neighboring vehicles toward the serving SC. Exploring spectrum and energy efficiency is a great challenge in 5G vehicular small-cell networking (V-SCN) due to the fast vehicle mobility and varying communications environment. Refs. [9,10] designed the non-orthogonal multiple access technology to address the interference management and handover in 5G V-SCNs. Software-defined network (SDN) is another potential technology that has been available for decades and needs to be reconsidered for the design of new concepts suitable for 5G use cases. Most visions [11,12] agree that 5G-V2X networks will be extremely densely populated by a heterogeneous combination of cells from different generations of 3G, 4G, and 5G.

As illustrated in Figure 1, the 5G V2X communication technologies enabled an advanced ITS road system (safety message and weather service solution), which includes a suite of sophisticated facilities and services designed for public convenience and road traffic safety. In the 5G vehicular network, the RSUs and 5G V-SCNs must provide adequate information about the network status and geolocation services. The access networks with a diverse set of fields related to V2V, V2N, and V2I networking uses are deployable in smart cities or even for internet-of-vehicles (IoV) applications Moreover, selecting the appropriate access point (AP) or small cell base station (BS) from the accessible heterogeneous wireless access network that meets the minimum necessary requirements for quality of service (QoS) is a great challenge. In this paper, we propose the use of the SDN concept, where the controller plane performed a decision-making procedure that took into account a variety of situational factors, as well as the network performance. To maintain a reliable connection and QoS, we opted for the topology discovery method to accurately localize the small cell-linked RSU (co-located) infrastructure. The deployment of the SDN for 5G V-SCNs can play an essential role in providing V2X services to the general public.

New radio access and advanced technologies, such as SDN and small cells in 5G network, pave the way to the new next generation of vehicular networks. SDN separates the entities of the control plane and the data plane. The control plane software is run on general-purpose hardware. SDN enables the self-service deployment of control and traffic routing [13]. The SDN controller has some default flow values pre-configured. As soon as the switch is turned on, the traffic flow is pre-programmed. SDN controllers and switches exchange information flow over the network via a secure channel, such as the secure socket layer (SSL) or the transport layer security (TLS) protocols, whereas the OpenFlow manages communication between the network and control layers [13]. In addition, multi-access edge computing is deployed through small cell linked to RSUs. With the design of the next vehicular network enabled by advanced technology, such as SDN, small-cells-assisted RSU to preserve latency requirements for data dissemination in 5G network for V2X is proposed. We propose a novel distributed vehicular network architecture-based SDN and small cells linked-up RSUs (small cells linked-up RSU that abstracts the multi-access edge computing approach). This article provides the following contributions:

• We propose a distributed based software-defined small cells-linked RSU vehicular network (diSRsVN) architecture as the new, next 5G V2X architecture.
• We describe the logical structure of the control plane of the small cell linked-up RSU as well as the software-defined wireless vehicle.
• We provide the discovery topology for reliable routing over the proposed architecture.
• We provide a simulation of the proposed diSRsVN architecture for data dissemination of V2X messages and discuss the relevant improvements of using SDN and small cells-linked RSU to fit the latency requirements of V2X services in a 5G-enabled vehicular network. The results of the proposed scheme are analyzed through network simulation tools that integrate frameworks capable of supporting cellular networks.

We compose the remainder of the paper as follows. Section 2 presents the related works. Section 3 presents the system architecture of the distribution of software-defined small cell-linked RSU as well as the description of components of the control plane layer. Section 4 presents the simulation results of the proposed architecture. Section 5 concludes with some conclusions.

2. Related Works

We highlight relevant proposals that have already been proposed in the literature in order to develop new reliable frameworks for prospective 5G-VANETs. The first methods take advantage of the benefits of the SDN paradigm to propose a more adaptable vehicular communication architecture. In [14], a hybrid SDN approach is proposed, in which network control is shared between the RSU controller (RSUC) and local SDN controllers installed in a subset of RSUs. The goal is for the new architecture to be more reconfigurable due to vehicle mobility. The new adaptable architecture solves the issue of connectivity loss between RSUC and the data plane. M.A. Salahuddin et al. [15] took another step by integrating the SDN into the RSU cloud. Their proposed architecture consists of traditional and specialized RSUs that dynamically instantiate, replicate, and/or migrate services via software-defined networking. The micro datacenter RSUs, such as RSU clouds, are reconfigured to meet more service demands. They consider a heuristic algorithm to reduce the cost of the reconfiguration overhead.

Having the same vision, the authors of [16] were intrigued by the increasing demand for computational resources in the next vehicular networks. They proposed scaling up the computing infrastructure by considering vehicles as potential candidates. Vehicle owners are offered a number of attractive incentives in exchange for adding their vehicles’ computing resources to the fog computing (FC) service provider. SDN and FC provide great computing frameworks for vehicular networks. Since SDN divides the software plane from the data plane, FC alleviates the pressure of the core network. Yaomin et al. [17] proposed the SDFC-VeNET architecture. They presented mobility management and resource allocation using the SDFC-VeNET. Even though these approaches have made significant advances in the design of future 5G vehicular networks, we consider 5G-SCN linked to RSUs as a potential candidate for the new generation of 5G and beyond vehicular networks. Furthermore, proposals that considered FC as small cell-linked RSU infrastructures failed to address how these SCs linked to RSUs will be efficiently controlled and supervised. To address this gap, our proposed architecture employs a distributed SDN approach to manage small cell-linked RSU infrastructures and address the issue of heterogeneous wireless access networks in 5G-V2X networks.

The requirements of various vehicular applications impose different wireless communications in vehicular networks. The bandwidth and coverage range of cellular networks for vehicular applications are costly. The work in [18] proposed the SDVN communication using heterogeneous wireless interfaces for network selection, which utilizes the concept of controller layering and link duration. In addition, Ref. [19] addressed the issue of loss of connectivity with the said controller entity by proposing a hierarchical SDN-based vehicular architecture with the goal of improving performance in the event of a loss of connection with the central SDN controller.

The majority of the works are built on an RSU and fog computing architecture, which may have scalability issues in network selection for vehicular applications with varying QoS requirements. Furthermore, most of these works do not deal with multiple RSUs linked to SCs to keep communication alive when no communication infrastructure is managing the computing resources of deployed RSUs. The following subsection discusses the motivation of small cells-linked RSU paradigm for V2X services.

2.1. Small Cells-Linked RSU Paradigm for V2X Services

Road side units (RSUs) are infrastructure communication nodes with vehicular networks. This implies that the vehicle must have access to road infrastructures via RSUs via V2I [20]. Furthermore, RSUs use wide area networks to forward received messages to ITS application servers. The number of RSUs deployed and their coverage has a large impact on vehicle communication. However, RSUs are unquestionably expensive to install and maintain; thus, there is a trade-off between full connectivity via RSUs and the deployment cost. To cover the increased cost of RSU deployment, road operators (ROs) can also use small cell architecture owned by mobile network operators (MNOs) to best fit the network densification or dedicated coverage that the control traffic management services hope to deploy. This may also be an economic consideration. For multi-operator coverage within a building, virtualized RSU and open radio access network (RAN) architecture will work best, enabling the flexibility of more capacity for ROs. In this situation, ROs are almost certainly expected to deploy and manage RSUs in the public sector [21]. Following that, the ROs can enter into commercial agreements with MNOs, who will undoubtedly deploy RSUs and run V2X services provided by ITS authorities [21]. Therefore, MNOs should make use of the existing 5G small cells infrastructure to support V2X applications among ultra-reliable low latency communication services, then motivate the efficient deployment of V2X services. Having a radio platform that can adapt and scale to support new V2X services is critical for ROs.

2.1.1. Typical Small Cell Linked RSU Deployment Consideration

The typical small cell-linked RSU scenario is considered for transport corridors. The deployment is envisioned for the following reasons: (1) to ensure service continuity for transport corridors (highways, public transit); (2) to boost transportation system efficiency and safety; and (3) to enhance the overall travel experience. The small cell-linked RSUs are located everywhere the RSU is deployed (e.g., traffic signs, tunnels, and bridges) in order to exchange information from and to the RO data processing centers. It is envisioned that the deployment density is medium, which means linearly along transport corridors. The small cell-linked RSU application server ends up saving road traffic conditions or caches data from passing vehicles. This small cell-linked RSU allows the prediction of the trajectory path and destination of by-passing vehicles.

2.1.2. Difficult Route forwarding for Small Cell Linked RSU in Vehicular Network

The routing and forwarding of traditional VNs depend on the vehicular node’s address. The software-defined vehicular network (SDVN) packet forwarding is based on node address in the shape of flow table [17]. However, in a small cell-linked RSUs network, there may be such a situation when other small cell-linked RSU nodes in the same site density may have the most recent road traffic data of the small cell-linked RSU area, and the ROs consumers can disregard information from the nodes themselves. Thus, the small cell-linked RSU node aggregates the safety traffic data. In this scenario, the data name can be used to request the data, regardless of which small cell-linked RSU node fulfills the data request. The requesting small cell-linked RSU node sends a message containing the requested content name to request data or content. When any small cell-linked RSU node with the required data receives the message of interest, it only sends the required data message. Recently, the research community proposed the vehicular named data network (VNDN) architecture [22] for this use case, but these architectures are only intended for content-based routing and are incapable of supporting other routing schemes such as geographic-location-based routing [23]. To meet the requirements of distributed small cell-linked RSU methods, a small cell-linked RSU framework supporting multiple 5G V2X communication modes is needed. With the development of V2X applications, 5G-SCN requires the scalability and capabilities of heterogeneous wireless communications which are challenging to manage.

The current 5G-V2X systems are normally vehicular network communication-oriented [24] and the system design for one V2X network communication cannot be used to other V2X network communications. With the development of 5G-V2X applications, traditional V2X onboard (communication equipped with the vehicle) network communication-oriented unit architecture has struggled to meet the increasing complexity of V2X service requirements. The existing onboard vehicle network unit cannot meet the requirements of all communication traffic in a unified way. With the assistance of the same network architecture, multiple V2X communication configurations interrupt each other, and cannot offer personalized network forwarding for each V2X network communication. This affects the normal V2X routing of the entire 5G V2X architecture communication. Therefore, we propose an onboard V2X unit framework that can enable multiple V2X network communications to use simultaneously on the same onboard vehicle network unit equipped with a 5G V2X system without interrupting each other.

2.1.3. Vehicular Network Management for Distributed Small Cell-Linked RSU for 5G V2X Services Requirements

The current 5G V2X system based on small cell-linked RSUs are V2X network communication oriented. Even the emerging 5G V2X architecture only allocates the V2X network resources to the vehicle’s onboard unit, and cannot realize customized vehicular network traffic routing based on the requirements of small cell-linked RSU. Because of the multi-mode V2X network and small-cell network characteristics, the traditional VN communication resource (routing and forwarding) allocation model cannot fully exploit the benefits of a multi-mode 5G V2X network, but also cannot meet the customized requirements of 5G V2X network communication resources (routing). Therefore, it is necessary to develop a distributed software-defined small cell vehicular network architecture that can adapt to multiple modes and diversified V2X networks and provide differentiated V2X network resource services.

3. System Architecture of Distributed Software Defined Small Cell Linked-Up Road Side Unit Vehicular Network

Next-generation vehicular networks must support heterogeneous radio access technology while also implementing a dependable data dissemination routing protocol. As a result, practitioners and researchers are thinking about focusing on communication technologies and programmable mobile cellular network elements. The framework is expected to incorporate flexibility in extending network elements, centralizing network management, and partitioning radio access coverage area to provide reliable data dissemination routing and control.

The proposed architecture is a distributed software-defined small cells-linked up RSU vehicular network (diSRsVN) is given in Figure 2. The whole diSRsVN uses the recent network technology of SDN to support flexibility, programmability, and scalability to the network [25]. Thus, diSRsVN combines SDN with small cells linked up RSU and clustering approach to partition the network. Hence, each partition (small cells linked-up RSU) is managed by a small cell RSU controller, which in turn is controlled by an SDN controller V2X management. Therefore, diSRsVN is distributed multi-hop control plane but logically centralized using a central SDN controller to enhance network management. diSRsVN uses the multiple small cells linked-up RSU controller called the small cell-type RSU controller (smallRSUC). These smallRSUCs interact with each other and react together to obtain global network status. This architecture differs from the one in [26] in the aspect of forming the distributed cluster. The proposal is based on using two levels of clustering. The clustering is based on small cells linked RSU (small cell-type RSU) for scalability and reliability of the whole management of the network. We use the term small cell-linked up RSU and small cell-type RSU interchangeably. The description of other components of the diSRsVN architecture is detailed in

Table 1. Definition of functional components of the diSRsVN.

Components	Description
$smallcell-typeRSU$	RSUs can operate as it is set up with 5G small cells.
$SDNOpenFlowController$	SDN openflow holds overall (global view) network of the proposed architecture
$OpenFlowSwitch$	delivering V2X rules to SD-onboard vehicle module
$smallcelled-typeRSUEnabledSDN$	delivering V2X rules to small cell RSU node
$Small-celltypeRSUController$	SDNCs control and forward routing rules into the network related to small cell-type RSU of the proposed architecture
$SDNV2XnetworkController$	configured to update flow rules for SD-onboard vehicle

.

3.1. diSRsVN

diSRsVN is based on an integration of two pertinent network concepts: SDN and small cells [25]. Thus, the small cell centralizes the intelligence of the small cells cluster and vehicles. The diSRsVN allows (i) partitioning the network in a small cells cluster according to the criteria that consist of linking up cellar base station with RSU to form a small cell; (ii) selecting the head cluster vehicle in each small cell linked-up RSU; and (iii) centralizing the global view of the network by connecting adjacent small cell- linked RSU-enabled SDN deployed at each small cell-linked RSU device. Therefore, deploying a small cell-linked RSU device and using a distributed small cell-linked RSU controller provides better performance in handling the traffic load, and enhances scalability with efficient data dissemination in the 5G V2X system.

Vehicles and road users are included in the data plane. V2X services also include V2N, V2I, V2V, and V2P links. Furthermore, the V2X onboard access unit is regarded as an SDN device, much like the data plane in the SDN approach. Functions of the SD onboard wireless communication are discussed later in this section (Section 3.2). In the V2X access layer of the data plane, the SD-onboard wireless vehicle primarily collects data, forwards it, queues it, and sends it into the small cell-linked RSU node. Furthermore, the forwarding devices, such as the OpenFlow V2X network switch and a small cell-linked RSU-enabled SDN, are included in the forwarding layer.

The control plane decisions are not entirely made by a single SDN controller centralized element; adjacent SDN controllers in the control layer could collaborate to make decisions. In fact, this collaborative decision is provided through mobility management, and quality of service (QoS) support provided at the SDN OpenFlow controller. The SDN OpenFlow controller should achieve a fast recovery after failures between adjacent controllers. A logically distributed control plane, distributed small cell RSU controller and SDN V2X network controller(s), achieve optimal performance and robustness against failures. The SDN V2X network controller has the role to select an in-vehicle wireless access communication radio to connect the small cell-linked RSU.

As shown in Figure 2, diSRsVN adapts multi-cell device-to-device (D2D) communication assisted with densely deployed small cells [27] to support efficient communication between distributed small cell linked-up RSU. Each small cell linked-up RSU represents the first level of cluster called software-defined small cells (SD-small cells), where the distributed small cell RSU controller centralizes the topology state of vehicles driving in the same direction. It forms a cluster head called SD-small cell head (SD-SmCH). We assume that SD-smCH initiates a transient software controller from the small cell RSU controller as long as it has the longest travel time until it leaves the coverage of small cell RSU. The transient software controller of the SD-SmCH includes flow entries of the routing paths in case SD-smCH fails to remain in the small cells. The flow entries stored at SD-SmCH are sent periodically and stored on flow tables of other vehicles grouped in the same small cell as the SD-smCH. The control backhaul between the adjacent small cell RSU controller in diSRsVN is based on multi-cell device-to-device communication or multi-cell cellular communication.

3.2. Software Defined On-Board Wireless Vehicle

Software defined on-board wireless vehicle (SD-OBVe) is a traditional vehicle capable software defined function in the future prospective generation of vehicular networks. Thus, apart the hardware resource that is available from the traditional onboard vehicle unit (OBU), the software resources of OBU need to support the SDN module that allows the integration of SD-OBVe to act as a forwarding device, while the logic control for routing is programmable and deployed elsewhere than on the vehicle. The hardware side includes the computing and storage unit as proposed in [26]. The difference of the software resource of the SDN module in [26] is based on the transient software controller service that keeps for a while the recovery controller (flow rules entries ) for when the SD-OBVe is elected as the SD-smCH. Furthermore, diSRsVN defines two software components deployed as virtual machines on the hypervisor as illustrated in Figure 3:

• The onboard vehicle controller (onboard controller): it is configured in standby mode and it is activated in two cases, first when the smallRSUC that control the small cell-type RSU partion is down, and the second case is when SD-OBVe is selected as the SD-SmCH in the small cell-type RSU cluster for routing data dissemination. In case the smallRSUC is up, the OpenFlow enabled V2X user equipment (EU) switch updates the flow tables accordingly to flow entries rules configured on smallRSUC. As the SD-OBVe moves to the next small cell-type RSU cluster, the onboard controller is initiated under the control of the current small cell-type RSU controller.
• transient software controller. It is a backup and recovery controller of the on-board controller. It keeps for a while the recovery controller (flow rules entries) for when the SD-OBVe is elected as the SD-smCH.

3.3. Topology Discovery over diSRsVN

We define four type of message in this topology:

• The Hello message: It is the message that carries the diSRsVN node interface information. This message broadcasts between the diSRsVN node in the one-hop neighboring term.
• Cluster topology control message (CTC) message: It is a unicast packet carries the cluster node information set (IS). It (IS) includes node information, link information, neighboring information, velocity information). The cluster head node’s packets are carried to the smallRSUC and are collected by the small cell linked RSU in its area of coverage. The smallRSUC will send the cluster control request (CCREQ) to the SD-smCH node. Then, the SD-smCH conveys the node IS of each SD-sm vehicle to smallRSUC as a response to the CCREQ request. Thus, the smallRSUC requests the CTC message when SD-smCH moves from one small cell to another.
• Hello-cluster-controller message: It is only exchanged between the SD-smCH controller (software) in adjacent small cell-type RSUs to exchange routing information.
• Hello-Controller message: This type of message is exchanged only between small cell-type RSU controllers. We assume that the routing information and unique identification of the small cell-type RSU include in the packet carried to the central SDN controller.

Each smallRSUC keeps two different topology state databases (TSDBs): Inter-TSDB includes the details topology of small cells topology (small cell-type RSU and SD-smCH nodes), and Intra-TSDB contains the topology information of adjacent small cells. The information on the topology state of smallRSUC (via Hello-Controller message) and is stored in the Inter-TSDB, consequently updating Intra-TSDB.

3.4. A Case Scenario: Disseminating Speed Notification over diSRsVN

We investigate how V2V communication would happen for SD-small cell vehicles equipped with same wireless interface in the same small cells’ coverage. The SD-smCH is elected since its transient software controller for recovery keeps the flow entries. The flow table of OpenFlow switch is updated by new flow entries generated by the smallRSUC located at the small cell-type RSU infrastructure. The OpenFlow switch is periodically updated with new flow entries when a new SD-smCH comes in the small cells linked up RSU. Then, when the SD-smCH receives flow entries, it updates them on the SD-vehicle OpenFlow communication interface so that they in turn broadcast speed notification packets to the one-hop neighboring SD-vehicle according to routing rules in the SD-vehicle OpenFlow communication interface.

The routing policies for speed notification are configured at the diSRsVN application layer and injected into the SDN controller. The SDN controller updates the small cell-type RSU application server that examines the speed notification services and determines its category of service, content, propagation range, target area, timeliness, and priority. The small cell-type RSU application server may coordinate and communicate with the adjacent small cell application server to determine the target area and the distribution of speed notification messages in the target area. This information collected by the small cell application server (AS) is transferred to smallRSUC so that it can update or generate routing rules for the OpenFlow switch responsible for forwarding packets to SD-vehicles. The message is downlinked by the SDN OpenFlow switch to the small cell-type RSU in the target area via 5G-V2X [28]. The small-type RSU broadcasts the speed notification message to nearby vehicles [28].

4. Implementation of the V2x Data Dissemination over diSRsVN and Its Results

4.1. Disseminate V2V Message Based C-V2X over diSRsVN

In this section, we review the simulation setup of the scenario to handle the dissemination of speed notification. The simulation evaluation uses the omnet++ simulator [29]. To integrate the OpenFlow protocol and SDN technology, we import the OpenFlow Module for omnet++ 5.4+ and INET 3.6 [29]. The VeinsLTE [30] and SimuLTE [31] frameworks were also imported into omnet++ to run the simulation of the proposed architecture. We outline the simulation parameters in

Table 2. Main simulation parameters and settings.

Parameters	Values
small-cell-RSU
application Type	TraCIDemoRSU11p
beacon Interval of application	1s
accident Count on mobility model	1
Topology discovery parameters
Transmission range of clusters	500 m
Processing time for queries	10 ms
Broadcast of channel information	10 ms
IEEE 802.11 parameters
Transmission range	500 m
Carrier frequency	5.890 $\times {10}^3$ Mhz
Data rate	6 Mbps
The number of available channels	1 CCH and 4 SCH
Propagation model
pathloss	two-ray interference
fading	Nakagami m = 1.3
General parameters
Number of small-cell-RSU	10
Number of vehicles	10–100
Number of small-cell controller	2
Number of openFlow Controller	1
Number of openFlow switch	2
Simulation time limit	400 s
Repetition run simulation	5
Mobility Model	Freeway
simulation area	5000 m × 5000 m

.

As the diSRsVN architecture is based on the SDN approach, the INET framework provides modules that apply and implement the SDN concept. This paper examines an initial model presented in [32] and applies two basic SDN elements, the SDN centralized controller and distributed small cell-enabled OpenFlow switches. They manage the basic openFlow messages for flow establishment and component communication [32].

We use the traffic simulators called simulation of urban mobility (SUMO) version 0.30 [33] to simulate real-world traffic conditions. SUMO is a microscopic traffic simulator with features, such as collision-free vehicle movement. The open street map is used to download the real-time traffic scenario (OSM) [34] of a section of San Francisco, California, USA.

To obtain trace files, the osm file is converted and processed using SUMO simulator tools. The latter is fed into omnet++ to run simulations. The vehicle density ranges between 10 and 100 vehicles. The vehicle’s speed ranges from 15 to 40 km/h. Small cell-linked RSUs are available in threes and are equipped with both DSRC and cellular interfaces (LTE-V2X). The transmission range is 500 m to the cluster head. The fixed query transmission time and the information broadcast time are both set to 10 ms. The small cell-linked RSU query processing time is set to 10 ms. For IEEE 802.11p, we set a transmission range of 500 m, a data rate of 6 Mbps, and a carrier frequency of 5.890e9 MHz. The simulation time is approximately 400 s. The results are saved with an average of five realizations. For vehicle movement, we configure the freeway mobility model. Another network simulation framework is Veins [30] which was used to simulate a vehicular application with RSU unit (TraCIDemoRSU11p). During the simulation, every vehicle is sending data to another, selected randomly. There is a pause interval chosen between [0, 1] s, and after that, the accident count mobility traces is about 1 in the chosen area. This accident count is enough to allow the Veins cars to propagate the accident packet for an amount of time selected between [10, 60] s. We considered a low varying traffic density because of the additional computation of small cells and distributed SDN controller to study the effect of increasing the number of vehicles. The clustering of vehicles in their respective small cells-linked RSUs clusters provides stable connectivity between the clusters’ vehicles, which fastens the transmission packets after the distributed controllers compute the global network topology. The input values for radio medium are the default value as defined in the TraCIDemoRSU11p application of the Veins framework. The propagation parameters used were extracted from [35].

4.2. Simulation Scenario

The aim of the simulation scenario is to disseminate vehicle safety messages (query vehicle information from the nearest vehicle) inside the diSRsVN architecture. Two main scenarios are simulated to evaluate the proposed diSRsVN architecture as follows.

We use three performance metrics: average end-to-end (E2E) delay, packet delivery ratio (PDR), and average throughput. They are defined as follows:

• E2E delay measures the average time it takes for transmitted packets to reach their respective destination nodes.
• PDR calculates the ratio of successfully delivered packets to total packets transmitted.
• The average throughput allows you to calculate the average rate at which packets are successfully received by the destination node.

4.2.1. Randomly Source and Destination Vehicle Node

The vehicle placed based on the sumo traffic models the vehicle speed, and the vehicle starting position in the simulation is based on the SUMO configuration file. Randomly, we select the source and destination. The source node sends the hello (beacon) message to all nearest nodes. If the source in the local controller is selected, directly communicate with the local controller, or select the forward nodes by exchanging the topology discovery messages. If the small cell RSU controller is busy, send the request to the SDN controller and make the route selection and packet transmission between the source and destination.

4.2.2. Update Flow Rules Based on Vehicle’s Mobility and Topology Discovering

According to the SUMO model, the vehicle placed based on the SUMO traffic model, the vehicle speed and the vehicle starting position in the simulation are based on the SUMO configuration file. The vehicle node sends the packet with the speed limit to the nearest nodes, based on their mobility speed and also the moving speed and moving direction based on the SUMO configuration. Compare the speed limit with the flow table in the small cell RSU switch. If it does not match with flow table, the small cell RSU switch updates the flow rules into the SDN controller, distributes to all nodes, and makes an upload.

4.3. Influence of the SDN Technology

To indicate the SDN approach’s impact on diSRsVN, we evaluate the performance of the proposed architecture by taking into account the presence of a distributed small cell RSU controller, central SDN controller, and without an SDN controller.

In the simulation, as the number of vehicles increases, so does the PDR for both the diSRsVN architecture aided by a single central controller and the no-SDN-based architecture. The use of distributed small cell RSU controllers decreases the PDR of the diSRsVN as the density of the vehicle increases. In the case of using distributed small cell RSU controllers, it seems that a higher number of packets are exchanged between vehicles. The SDN controller may select another small cell RSU controller distinct from the one previously chosen and located in proximity to the SD-SmCH to disseminate the vehicle’s speed information. This reduces the packet delivery ratio. In the case of cooperative warning query dissemination from ROs, the proposed scheme is aided by the central SDN controller, which provides stable route information based on network performance as measured by the number of small cell RSU. It is not necessary to rely on the distributed small cell controller because the V2X controller has the routing rules for the destination vehicles. Therefore, it does not need to forward the dissemination request to the source node, which may influence the V2X controller to support the best small cell RSU. This reduces the number of messages exchanged between the small cell RSU and the vehicle. It also reduces packet collision in the SDN controller. This is due to the computation of the global view of the network incurring a lower packet delivery ratio. A number of packet losses are incurred when the proposed architecture partially relays distributed small cell controllers to reach the small cell RSUs.With a vehicle density of 80 (Figure 4c), the distributed small cell RSU controller has a packet delivery ratio of $95.18$, with the use of the central controller, it measures $96.15$, and with a PDR value of $90.68$ without using an SDN controller.

Figure 4a shows the average throughput. On an expanding number of nodes, the average throughput of the proposed diSRsVN architecture assisted by the distributed small cell RSU controller increases for all scenarios. As the number of vehicles increases, the connectivity is stable between cluster members; consequently, the diSRsVN architecture fastens the transmission packets. However, the use of a central SDN controller provides stable average throughput to generate routes to the small cell RSU nodes than no-SDN-based. The diSRsVN uses the topology discovery message when the distributed small cell RSU controller suggests a list of different small cell positions and according to the selection of the cluster head. In terms of vehicle density of 80, the average throughput of diSRsVN assisted by the distributed small cell RSU controller is 101.57 Kbps, whereas 93.24 Kbps and 89.46 Kbps are the average throughputs of diSRsVN assisted by a central SDN controller and without an SDN controller, respectively.

Figure 4b shows the average E2E delay versus vehicle density. In the proposed architecture, the E2E delay increases as the number of vehicles increases. In general, as the number of vehicles increases, connectivity remains stable. The results indicate that as the number of dissemination requests increases proportionally to the number of vehicles, the SDN controller receives an important number of flow routing updates according to the number of small cell RSUs available to forward dissemination responses to source nodes. In this paper, the topology discovery process increases the E2E delay as the number of nodes increases. Hence, diSRsVN with distributed small cell RSU controllers has less delay because they assist in selecting reliable routing paths to deliver the packets as the knowledge of the topology discovery process is completed and flow tables are populated to update the routing rules. When the vehicle density is 80, the E2E delay experienced by distributed small cell RSU controllers is 24.29 ms, which is less than 37 ms when only the central SDN controller is used and 58.87 ms when no SDN is used.

Varying the three methods for performance metrics allows evaluating the impact of distributed SDN controllers in the topology discovery of the proposed architecture. We observe that the distributed SDN controllers show better performance, which is because the routing rules are updated to small cell-linked RSUs in the cluster area closer to vehicles. The small cell-linked RSUs deliver packets to vehicles that are closer, which in turn implies lower delay. The use of distributed controller achieves a reduction in E2E delay of about 58 percent compared to the traditional one without an SDN controller and a decrease of about 35 percent in relation to the centralized SDN controller. This decrease is due to the distributed SDN controllers’ efforts to keep only valid routes in their flow tables; in other words, it works to remove routes that have become invalid due to changes in the network topology. Although a centralized SDN controller has better PDR performance, the distributed controller solution achieves a higher average throughput.

4.4. Results Analysis in Comparison with Existing Works

The diSRsVN architecture is evaluated in comparison with two SDN-based schemes, the hierarchical software-defined VANET (HSDV) [19] and VSDN-based heterogeneous wireless interface (HetSDVN) [18]. These existing schemes were chosen because both use the concept of local controllers to reduce the controller’s signaling network traffic. The VSDN architecture in [18] includes a network selection and data dissemination protocol. The network selection is carried out using utility functions and a game theory approach. The controller manages dissemination in HSDV by selecting the next hop that is closest to the destination.

Figure 5c shows that the controller manages dissemination in HSDV by selecting the next hop that is closest to the destination. The PDR for HetSDVN and HDSV increases as the number of vehicle density increases (fewer packets are dropped). However, as the vehicle density increases, the proposed architecture’s PDR decreases. To summarize, the proposed scheme maintains the PDR regardless of the density of the vehicle nodes.

The proposed scheme employs the SDN controller approach; VSDN includes a heterogeneous wireless interface. The proposed architecture improves PDR because the distributed small cell topology discovery reduces the connectivity loss through the network’s global view, which is known in advance and updated frequently. The routing paths are predefined in the OpenFlow switch, reducing the flow of messages exchanged between small cell RSU nodes and, as a result, decreasing packet collision. The transmission loads between vehicles and small cell RSU infrastructures are also reduced by the clustering nodes.

Since the cluster head would also update flow tables with routing rules received from the distributed small cell RSU controller, the small cell RSU updates route information for previous query dissemination messages from the cluster head. The PDR for the proposed scheme is 94.32 % at a vehicle density of 80, whereas HetSDVN using a heterogeneous wireless interface and HSDV are 91.08 % and 88.04 %, respectively, at a vehicle density of 80.

Figure 5b shows the average throughput with the vehicle density. When compared to existing SDN-based schemes, the proposed architecture has a higher average throughput. This is because the node vehicles, with the help of the SDN controller’s small cell RSU and topology discovery, distribute packets to their destination nodes. HetSDVN with a heterogeneous wireless interface provides stable average throughput as the vehicle density increases, whereas HSDV is lower than the proposed architecture. The best routes predicted by the SDN controller must be chosen in order to achieve higher average throughput. At a vehicle density of 80, HetSDVN with a heterogeneous wireless interface appears to have 92.51 Kbps, whereas HSDV and the proposed architecture have average throughputs of 89.67 Kbps and 103.45 Kbps, respectively.

Figure 5a shows the average E2E delay. As the number of vehicles increases, the proposed scheme reduces the average E2E delay. This is due to the cluster header having a high probability of selecting a group of vehicles that travel in the same direction when the Hello-cluster-controller message is sent. Because the path of disseminating traffic messages is established from the distributed small cell RSU controller based on the topology discovery process, the delay is gradually reduced. The latter is determined after the control message is computed based on the formation of the cluster head with the available number of small cell RSU nodes. If a traffic query request is generated by the nearest vehicle nodes, the SDN controller configures new path routing rules; otherwise, the distributed small cell RSU controller holds ideal route information for the source vehicles grouped in their respective cluster region, which is under the control of small cell RSU. At the vehicle density of 80, the proposed architecture has an E2E delay of 24.27 ms, which is lower than 30.64 ms in HetSDVN using a heterogeneous wireless interface, 25.27 ms in HDSV. HetSDVN, using a heterogeneous wireless interface, considers calculating the duration of the route before it is selected and injected into the local controller, which would increase the processing latency. The proposed scheme calls for the use of the assigned small cell RSU chosen based on the topology (classified in the cluster). As a result, the topology discovery quickens the location of small cell RSU.

4.5. Routing Overhead of Packet Dissemination over 5G V2X Network Protocol

We perform a process of data dissemination (speed data in RSU) to a cloud application server through the core network; the core network is based on both LTE-V2X and 5G core network.

Figure 6 shows the routing overhead with varying vehicle densities. The proposed 5G V2X networks’ routing overhead increases with traffic density. This is because establishing a cluster head near the source vehicle increases the routing messages while distributing packet responses to the destination node. As a result, the input topology is routed to SD-SmCH flow entries in the distributed small cell RSU controllers, resulting in fewer beacon messages exchanged between the source vehicle and the small cell RSU. However, when compared to the use of the 5G V2X core network, the SDN-based approach, aided by distributed small cell controllers and the central controller, is the foundation of the lower routing overhead. The small cell RSU node communicates with the SDN controller only if the flow entries in the distributed small cell RSU controller do not match the routing rules. The more distributed local controllers use the input topology to construct the route information, the lower the link failure by lowering the routing overhead. The routing overhead of the proposed architecture assisted with the 5G core network is 22.15 at the vehicle of 80, whereas it is 24.44 for LTE-V2X communication.

5. Conclusions

In this article, we present a logical diSRsVN architecture based on SDN and small cell-linked RSU. Consequently, we describe the control plane layer of diSRsVN, which has the small cell-type RSU and vehicle-enabled SDN for reliable data dissemination. The dissemination of speed notification in small cell-linked RSU leads each small cell to elect an SD vehicle cluster head to improve the reliability of successfully disseminating V2X messages with delay optimization through the small-cell concept. Moreover, the topology discovery flow messages for the SDN are presented. The small cell-link RSU model is implemented in the omnet++ network simulator during the simulation evaluation. The simulation results indicate that SDN and small-cell concepts applied to the vehicular network can guarantee latency requirements for deploying V2X applications. We will consider the time synchronization for LTE-V2X communication in a dense deployment of vehicle-to-everything communication. The reliability of the data transmission and interference management will be investigated in future studies.