Load balancing ad hoc on-demand multipath distance vector (LBAOMDV) routing protocol

A mobile ad hoc network (MANET) is a collection of wireless mobile nodes which dynamically form a wireless network without a fixed infrastructure or a wired backbone network, as shown in Fig. 1. A mobile ad hoc network is a multi-hop network without any predetermined topology or central control, hence making routing a crucial design issue for these networks [1]. One of the most important and thoroughly discussed areas in wireless communication has been channel assignment and usage. As ad hoc networks are decentralized, they require advanced bandwidth allocation schemes [2]. Various issues ranging from topology control and network coverage to channel assignment have been addressed recently, and viable solutions have come forth. The issue of topology control and streamlined data communication has been thoroughly discussed [3, 4]. The advent of wireless technology in the public domain has raised issues related to network coverage for reliable communication [5‐7]. Recently, a new area of concern for researchers has emerged known as green mobile networks, which are wireless networks sensitive to energy usage and dissipations [8].

The rapid node addition or departure frequency of wireless nodes adds to the complexity of routing in MANETs. The inherent issue of rapid mobility complicates the issue further, by making it difficult to address a particular topology for a long [9‐11]. The fundamental concern related to the complexity of routing has been addressed, but still some inherent bottlenecks are unresolved. The issue related to delay in route selection has been discussed, and many viable solutions have been proposed in the past [12, 13]. The issues related to congestion and its impact on data delivery have also been addressed [14]. In the past, the basic principle of routing protocols generally was limited to identifying a single path between source and destination node, with a constraint of minimum hop counts. This genre of routing protocols had limitations of reliability and bandwidth usage owing to path breakdown. Nevertheless, in a reasonably well-connected network, there exists several paths between a source node and a destination node. Multipath routing is one of the most effective trends in the area of wireless routing [15]. The concept of multipath routing involves discovering all available paths from the source node to the destination by taking advantage of path redundancy of the underlying network. These multiple available paths may be used alternately or in certain scenarios concurrently for data transfer [16, 17].

The process of routing both in general computer networks, as well as in wireless networks, can be reactive or proactive. The category of reactive routing protocols has been the most effective for wireless scenarios, and the introduction of reactive multipath on demand protocols has steered to a new generation of MANET routing. The most rigorously discussed reactive wireless (on-demand) routing protocols to date have been Dynamic Source Routing (DSR), Ad hoc On-Demand Distance Vector Routing (AODV) and Temporally Ordered Routing Algorithm (TORA) [18‐20]. All these reactive routing protocols flood a network with route request packets (RREQ) to discover and maintain routes, only after receiving a request for data transfer. Proactive routing protocols such as Destination Sequenced Distance Vector Routing (DSDV) or other DSDV variants tend to discover and maintain routes between all pairs of nodes in advance to any data requests [21]. The application of proactive protocols seems unrealistic for mobile wireless scenarios, as the frequent topology changes, that may require frequent updating of routing tables which will incur huge routing overhead [22].

The transfer of data over a single path among the various paths has been the principle of the most on-demand multiple path routing schemes. If the route fails again, a new route is evaluated from source to destination, consuming time and resources. So it is obvious that distributing data packets among the discovered multipaths is the key solution. The research in the recent past has resulted in some advanced techniques for routing in wireless communication networks [23]. Advanced areas of data communication in wireless sensor networks have been taken both by the academia and industry and have yielded significant results [24, 25]. Techniques addressing the issues related to sensor networks for data processing and delivery have also been part of recent research developments in wireless technology [26‐28]. The concept of information-centric networks (ICN) has also come into light during the last half decade. These information-centric networks pose quite a different set of challenges [29]. Advances in making IoT a successful technology in future requires greater integration between existing network platforms and the internet, thus enabling their ability to provide services across the globe [30].

The work through this paper mainly concentrates on effectively balancing the data load among the discovered multipaths in order to maintain nodes resources and reduce the traffic to avoid creating traffic load or bottlenecks.

This paper is organized as follows. Section 2 covers related work in the area of on-demand multipath routing protocols. Section 3 proposes the LBAOMDV load-balanced ad hoc on-demand multipath distance vector routing protocol. Section 4 provides the simulation details of the environment set for the NS2 network simulator. Section 5 of this paper discusses in detail the results generated from simulation of the proposed LBAOMDV protocol.

Many conventional routing protocols developed in the past have least emphasized the optimum utilization of network bandwidth and node energy [31]. But lately, several multipath routing protocols have targeted the energy consumption issues while devising routing schemes for wireless mobile networks [32]. A detailed discussion of the advances in mobile ad hoc routing is provided in this section.

Effective load balancing of data across multiple available paths has been utilized for fair bandwidth usage in mobile ad hoc networks [33]. Pearlman et al. proposed an alternate path routing (APR) that provided load balancing by distributing traffic among a set of diverse paths. APR is an ideal candidate for mobile ad hoc networks with limited channel bandwidth [34]. Yin et al. introduced a load balancing technique called multipath adaptive load balancing (MALB). MALB dynamically distributes the traffic among multiple paths, based on path statistics measurement. MALB is an inclusive framework and can collaborate with any kind of multipath source routing protocols [35]. Mérindol et al. had proposed a scheme that offers the possibility to use temporarily alternative routes in order to reduce packet loss and degraded throughput [36]. Nagarjun et al. proposed a Packet Count-Based Routing Mechanism (PCRM) protocol. PCRM is based on Dynamic Source Routing (DSR) protocol. In contrary of all other regular load balancing techniques, PCRM selects the least used path for sending data packets rather than the mostly used one that regularly includes the minimum hop count [37]. Sharma et al. proposed a similar solution to utilize available bandwidth of the channel multiple disjoint paths. The approximation of bandwidth of a given path is done by sending detector packets across a network. The source node chooses the path with maximum bandwidth as the primary route for forwarding data [38]. Recently, Qi et al. proposed a multipath routing protocol (EM-AODV), based on AODV. EM-AODV evaluates the paths using the parameter values of path energy and hop counts [39]. The advancements in AODV have resulted in more efficient protocols like ad hoc on-demand multipath distance vector (AOMDV) that is more suitable for MANET scenarios [40]. Advancements in the routing techniques in wireless communication have also been based on various intelligent techniques. Athanasios et al. introduced an intelligent approach to improve quality of service (QoS) routing in ATM networks [41]. Thrasyvoulos et al. in their work provided insights into design of routing algorithms for vehicular ad hoc networks [42]. Moustafa et al. described the routing metrics for routing in cognitive radio networks in detail [43]. The designing of vehicular networks is also quite challenging owing to rapid mobility. Few techniques to enhance the functionality of vehicular networks are being floated for discussion in the industry and academia [44, 45]. The latest advancements in improving the quality of routing has been studied by researchers, and their findings have uncovered new challenges [46]. The most interesting solution for these kinds of challenges in dynamic routing has been the efficient multipath routing schemes [47]. Some recent work carried out in the direction of QoS, the quality of service constrained efficient path selection has been noteworthy [48‐50]. The applications of basic network concepts have been realized in other spheres of science and technology as well. These realizations have also been incorporated into wireless networks to achieve significant gains in efficient routing [51, 52]. The current work is in continuation of the previous established attempts and has shown significant improvement in the same direction. The current work has utilized the popular NS2 network simulator for deriving and showcasing the results [53]. In the next section, we propose the LBAOMDV protocol that takes QoS to a higher level by enhancing both performance and reliability of MANETs.

A MANET can be realized by an undirected graph G(V,E), where V is the number of nodes and E is the total possible edges. Here, we associate a definite weight representing the amount of energy that gets consumed for the transfer of a single fixed length data packet between a pair of nodes. So, each edge in the given network graph will have a weight WE _ij representing the amount of energy consumed for a single hop transfer from node i to node j. The amount of energy for a single hop transfer between the given two network nodes can be calculated by a trivial function given in Equation (1):

The set of WE _ij values for all the possibilities of edges of the given network graph G(V,E) can be represented by a weighted adjacency matrix depicted in Fig. 2.

The value of the weight WE _ij of a given edge may be zero, in case a direct link is not available between a node pair V _ij . A given path “P _ij ” from a source node V _i to a destination node V _j can be comprised of n nodes and n−1 multiple hops where 1 ≤ n ≤ V.

The amount of energy required for the transfer of a single data packet from V _i to V _j can be calculated by the summation of weights WE _ij of all participating hops and can be given by Equation (2).

The value PE _ij represents the amount of energy consumed in the transfer of a single packet of data along the given path P _ij . A message “M” to be transferred from a source to destination node can have “Q” number of packets. So, the amount of energy “£” required for transferring a message “M” from source to destination node along a given path P _ij can be calculated as given in Equation (3).

A message M comprising of Q packets of data can be divided into fixed C-sized chunks of data to be distributed among K available multiple channels as given in Equation (4).

The load balancing of data across multiple channels ensures optimum utilization of the available bandwidth for a given wireless network. A path P _ij is a sequence of nodes and links from the intended sender to the intended receiver. The efficiency of data transfer among a node pair primarily depends on node capacity L and link traffic T of a given data channel. A protocol is said to have a fair bandwidth usage if it allows balanced traffic across the available paths. The amount of unused bandwidth (UB) is a firm indicator for ensuring equal usage of bandwidth across different paths. The UB parameter for a node pair or hop provides us the value of unused bandwidth at the particular hop and can be derived as given in Equation (5).

To ensure that the node with minimum unused bandwidth (MUB) over a path P _ij can handle the allocated data chunk, it can be derived from Equation (6).

A multipath routing protocol needs to be governed by a load balancing scheme in order to escape from devouring the energy of participating nodes. Unbalanced battery energy consumption may lead to node breakdown consequently affecting the reliability of the given network in general. The objective of this research work was to develop a bandwidth and energy-aware multipath on-demand routing protocol for MANETs. The proposed LBAOMDV protocol achieves the objective by load balancing of data across multiple available paths. The LBAOMDV protocol further ensures balanced retransmission of the data, in case of any path or node breakdown during a data transmission. The results generated from the simulation for node breakdown and average energy consumption parameters showed the improvements achieved by LBAOMDV in comparison to AOMDV, PCRM, and AntHocNet. These results conclude to the fact that the LBAOMDV protocol enhances the reliability of a given MANET and ensures better network performance at peak data transfer times. This work requires further enhancements in future so as to improve the computational efficiency of load balancing.