DMPFS: Delay-efficient multicasting based on parked vehicles, fog computing and SDN in vehicular networks

Abstract

The vehicular ad hoc networks (VANETs) can be exploited in many safety applications like emergency, firefighting, etc. to multicast safety information quickly to several destinations. Thus, the safety applications need to design multicasting schemes to select the optimal path with minimum delay. On another side, Software Defined Network (SDN) is a modern concept that presents availability, reliability, and flexibility and enhances vehicular network architecture. Moreover, the Fog Computing (FC) framework is appropriate to perform the multicast routing process in VANET because it supports the continuous mobility of vehicles and location awareness. The parked vehicles can be used as fixed FC nodes to transfer data with high link stability. This paper aims to produce a Delay efficient Multicasting approach depending on Parked vehicles, FC and SDN for VANET (DMPFS). In DMPFS, the bandwidth constraint is taken into consideration in discovering the optimal multicast route. Moreover, a scheduling approach based on the priority of application type is proposed to classify and schedule the multicast demands. The partitioning technique is exploited to reduce the overhead and time complexity of DMPFS. The simulation results illustrate that DMPFS is superior than adaptive Distributed Tree-based Multicast Routing protocol (DTMR) and Micro Artificial Bee Colony based Multicasting (MABC). In terms of the packet delivery ratio (PDR), DMPFS results in about 9% and 14% improvement compared with DTMR and MABC, respectively. Average end to end delay with DMPFS is approximately 33% and 41% less than DTMR and MABC, respectively. In terms of the communication overhead, DMPFS leads to about 22% and 28% reduction compared with DTMR and MABC, respectively.

Introduction

Vehicular Ad hoc Networks (VANETs) have been investigated in the last years for delivering safety information about the road, accidents, traffic jams, etc., to the drivers, police, ambulance, and passengers [1], [2]. These networks consist of a large number of moving and stopped vehicles, road side units (RSUs), and base stations (BSs) [3]. The mobility of vehicles poses challenges associated with high broken links and dynamic topology. Therefore, routing is one of the most significant challenges in VANETs and is accompanied by many negative things for example but not limited bandwidth consumption, packet loss, and overhead due to the exchanging of control packets required to obtain the optimal paths [4], [5]. Different connection approaches in a VANET i.e., Infrastructure/Vehicle to Infrastructure/Vehicle (I2I, I2V, V2I and V2V) may be used to send information in multicast, broadcast, and unicast transmission modes [2], [4].

Multicasting is one of the essential fields in VANETs that helps to deliver data packets to the target vehicles with low energy, traffic, and delay. In addition, it prevents the loops and enhances the network throughput and channel utilization [6], [7]. It can be used in some events, e.g., accidents, to send safety messages [8]. After receiving these messages, the target vehicles should take actions quickly to avoid a new accident. Therefore, delay and bandwidth have significant impacts in these circumstances [4]. The authors in [9], [10] proved that to do packet multicasting with multiple quality of service (QoS) parameters is an NP-complete problem.

Fog Computing (FC) is one of the newest technologies that exploits network edge resources to perform the processing, analysis, and computation quickly [11]. Moreover, it can be used in dynamic networks such as VANETs since it provides great support for movement and location awareness [12], [13]. Thus, it can help in addressing the challenges of the vehicles' mobility and routing process. The fog nodes such as access nodes, routers, switches, and servers that are found at the edge of the networks are the closest devices to the end users or Internet of Things (IoT) devices [14]. In the VANET, the parked vehicles can also be used as fixed FC nodes which can increase the link stability [15].

The complex networks need to perform several tasks such as traffic monitoring, routing, access control, and load balancing. On the other hand, the traditional network devices such as routers and switches are vendor-specific (not open devices), making it difficult to modify these devices. Each of these devices performs the functions of controlling and forwarding data. Therefore, in order to define a new policy in the network, the network administrator must individually configure each network device. A high dynamic network like a vehicular network will make this even more challenging [16]. Also, the growth in real-time services has led to the development of inflexible and complicated configurations at the control plane of network devices. Moreover, increasing the service requests that are sent to an infrastructure will increase the load on that infrastructure and may saturate it. Therefore, it will be out the service and not available for the new incoming demands. It indicates that the large number of requests for these services affect the availability of network infrastructures [17]. In this regard, Software Defined Networking (SDN) simplifies the traditional networks' functions by dividing the architecture of the network into two planes: data and control. This separation makes the network devices simple and flexible so that they are used solely to forward packets depending on decisions made by the central controller [18]. Also, the central controlling improves security and robustness while preventing the manual reconfiguration of network devices. SDN abstracts and manages the resources using a centralized intelligent authority to increase the availability and enhance traffic flows [17]. SDN concept can be utilized to simplify the routing process in different networks, especially in VANETs [18]. In the SDN-based VANET, the controller has a global view of the network. It has information about the network devices including fixed nodes and vehicles). Therefore, to compute the route from one point to another, the controller can compute it quickly and efficiently with minimum resources to guarantee quality of service factors [1]. Moreover, it enhances management and flexibility, utilizes network resources and the VANET connections [19], [20].

Safety applications are among the most important applications in the VANETs. In these applications, in some situations such as accidents, there is a high need to transmit some of safety messages with a certain level of QoS from a source vehicle to a set of endangered, police, or/and ambulance vehicles in different locations. This transmission mode is known as multicast mode. As a result, the endangered vehicles must perform some actions quickly before a certain critical time to avoid a road accident or collision. Moreover, the police and ambulance vehicles must reach quickly to the accident location. Thus, the average end-to-end (E2E) delay must be decreased and the system bandwidth must be increased in order to deliver such safety messages with certain QoS. To do this, there must be a good multicast routing protocol which focuses on reducing the delay and takes the network conditions into account. The multicast routing problem has been studied extensively in traditional vehicular networks and several multicast routing algorithms have been suggested. But these studies did not provide an efficient solution to multicast data routing with minimum delay in SDN-based VANETs, especially in safety applications. Therefore, there is a need to devise a novel multicast routing protocol for safety applications to compute the best multicast route with the lowest delay in the SDN-based VANETs as well.

To our best knowledge, this paper is the first research work to provide a multicast routing protocol by exploiting the parked vehicles, FC and SDN together, for vehicular networks, to reduce the transmission time of safety information. Our contributions are summarized as follows:

• We propose a new architecture by adding the parked vehicles to the vehicles layer physically and to the fog computing layer logically. We exploit them to work as fog computing nodes to participate in the routing process and increase the links' stability.

• We propose a new multicast routing protocol called DMPFS for VANETs' safety applications by exploiting the FC, parked vehicles and SDN. We formulate an optimization problem to minimize the overall delay in multicast routing process for safety applications of VANETs by selecting optimal available routes while taking into account the bandwidth constraints. The formulated problem is a mixed integer programming problem that is solved using CPLEX software at the central and the local SDN controllers to get the optimal routes.

• We present a new classification algorithm to classify the incoming multicast requests into ambulance, police, firefighting and regular according to the vehicles' types that launched those requests.

• we propose a new scheduling scheme to prioritize the incoming multicast requests; it gives high priority to critical and essential multicast requests to be served before the other requests and, as a result, it helps in avoiding catastrophe in many safety applications.

• In order to minimize the communication overhead and computation time of discovering the multicast routes, we apply a partitioning method to divide the whole network into smaller parts, which also helps in reducing the load in the central controller.

• Moreover, we propose a new joining/leaving technique which is executed by the central and local SDN controllers to decide about joining/deleting a vehicle(s) to/from the active multicast session.

The behavior of DMPFS with various numbers of parked vehicles and multicast sessions, and vehicles' velocity has been studied using the simulators OMNeT++ 4.6, veins 4.6 and sumo 0.19.0 installed on Windows-7. From the various simulation scenarios, we found that DMPFS was effective and outperforms DTMR and MABC in terms of the packet delivery ratio, average end to end delay, and the communication overhead.

The rest of the article is organized as follows. Section 2 presents a literature review. Section 3 explains the definition of the problem and the proposed system model. Section 4 shows the suggested routing strategy. In Section 5, the simulation scenarios and results are explained. Finally, the conclusion and future directions are provided in Section 6.