TNC Bandwidth Constrained Multipath Routing Protocol for QoS Provision in MANETs

Mobile Ad hoc NETworks (MANETs) are composed of mobile nodes with limited resources and unpredictable node movement. With the on-going evolution in MANETs, provision of Quality of Service (QoS) has become a challenging task. In this paper, a node disjoint Bandwidth Constrained Multipath Routing (BCMR) protocol is proposed which considers bandwidth as the key factor to discover multiple paths for QoS provisioning. BCMR accommodates required bandwidth function in flooding route request packets. Extensive simulation study is carried out to investigate the performance of BCMR. Simulation results reveal that BCMR significantly reduces overheads, minimizes overall end to end delay and significantly improves packet delivery ratio.


Introduction
MANETs consist of several mobile nodes connected by wireless links. MANETs are self-configuring network without having any fixed infrastructure or central administration [1]. Network topology changes continuously due to random mobility of nodes comprising the network [2]. In wired networks, network topology does not frequently change as in MANETs. Due to unpredictable nature of MANETs, mobile nodes may join or leave the network dynamically. Each mobile node in such networks has the capability to communicate directly with other nodes either directly or through multihop communication. MANETs mostly has applications in situations like disaster relief, military operations, rescue operations or conference networking. In these networks, source destination connection may break any time and required to be updated regularly. Quality of Service (QoS) in MANETs is a big challenge. With [95] the evolution of advance technologies in wireless communications, high bandwidth applications such as video or audio streaming, video conferencing and real time data transmission become an important part of application for such networks [3]. These kinds of application require a high level of service quality, so good QoS support is necessary. QoS is a set of service parameters like throughput, overall end-to-end delay, jitter, packet delivery ratio that are essential for any particular application for data transfer through that network [4]. Due to this unpredictable behaviour of MANETs, on demand routing protocols like Dynamic Source Routing (DSR) [5] and Ad hoc On-demand Distance Vector (AODV) [6] are generally implemented. Every node in the network is supposed to have some finite buffer. To select a path with minimum delay from different available paths between source and destination nodes, an optimal route strategy has been proposed.
In recent years, much more work is carried out by authors on QoS multipath routing. Yi et al proposed a Multipath Optimized Link State Routing (MP-OLSR) for large and dense networks considering heavy network load and high mobility [19]. Modified Dijkstra algorithm is proposed to find multiple paths from source to destination nodes. This algorithm is modified by using two cost functions to discover multiple or node-disjoint paths. Lee et al. proposed a backup path scheme for MANETs taking AODV as the base protocol [20]. One hop search mechanism is used to discover a backup route for destination. Main QoS parameters considered in this paper are packet delivery ratio and end-to-end delay from source to destination. This approach creates a backup route with a maximum of two hops from restoring node to target node. Performance analysis shows much improvement in terms of packet delivery ratio and endto-end delay characteristics. Jiang et al. proposed Dual Path Node-disjoint Routing (DPNR) protocol considering AODV as the base protocol [21]. The proposed approach work efficiently for data salvation in case of link failure and improves the robustness of available paths. For this it maintains two shortest backup routes from source to destination. This routing protocol performs well in lightly loaded networks as in heavy loaded networks; packet processing, packet collision and channel contention are increased. Kim et al. proposed a multipath routing algorithm for QoS and multimedia services in MANETs based on Ant Colony Optimization (ACO) technique [22]. The proposed algorithm discovers multiple paths by using ACO to provide QoS. Routing packets are adaptively distributed through multiple paths available to the destination to ensure proper network resources utilization. Igartua et al. proposed dynamic Contention Window-Multipath Multimedia Dynamic Source Routing (dCW-MMDSR) [23]. A dynamic Contention Window (dCW) is considered to outperform 802.11e by assigning smoother values of contention window for each access category. This multipath routing approach is suitable for video streaming depending on the state of the network. Ant Colony Optimization (ACO) technique is used to allocate bandwidth throughout the route [24,25]. To measure various parameters like delay, bandwidth and next hop availability, ant like agents is used. These are Forward ANT (FANT) and Backward ANT (BANT). Path preference probability is calculated to select a particular path from various available paths. To transmit the data from source to destination, path with highest preference probability is selected. During path discovery, this algorithm finds fault prone nodes and skips them from selected path to satisfy QoS requirements. Performance analysis clearly shows much improvement in packet delivery ratio and throughput. Rajendra Kumar Gupta et al. proposed Node Disjoint Minimum Interference Multipath Routing Protocol (ND-MIM) routing protocol which enables discovery of two node-disjoint minimum interference paths from source to destination [26]. Such paths are identified using HELLO messages. To setup two node-disjoint path, three control messages are used which reduces control overheads. It also helps in setting up backup route faster. Simulation results show significant improvement in overall endto-end delay. Extended AOMDV for Multi-Interface Multi-Channel (EAOMDV-MIMC) and its extended protocol Extended AOMDV for Multi-Interface Multi-Channel with Path Metric (EAOMDV-MIMC-PM) for multi interface multi-channel networks is proposed considering AODV as base protocol [27]. This approach utilizes multiple homogeneous networks interface to improve performance of MANETs. In this, channels are not assigned rather nodes can make use of all available channels. In EAOMDV-MIMC, nodes distribute data packets by estimating channel condition of the network interface queuing. In the extended protocol, estimated channel condition is piggybacked in data packets and propagated further. Ping Dong et al. proposed a Beacon-less Geographic Multipath (BGM) routing protocol is proposed that uses location information to discover maximally node-disjoint paths to avoid control messages caused by route discovery [9]. Data packets are forwarded within a disjoint subzone divided by division algorithm. Beaconless mechanism is used in forwarding strategy to reduce number of useless beacons. Zhang et.al proposed an interference based topology control algorithm for delay constrained mobile ad hoc networks. Firstly, minimized the power consumption and at the same time satisfying the interference constraint and then transmission power is increased to meet the delay constraint [28]. MAC protocols are proposed for dynamic channel allocation and cooperative load balancing to improve bandwidth efficiency under non uniform load conditions [29]. CDCA-TRACE algorithm is incorporated to provide support for non uniform load distribution. An opportunistic routing protocol JOKER is proposed to improve the performance of network supporting multimedia traffic as well as enhancing the nodes energy efficiency [30]. New metrics for selecting candidates is developed as packet delivery and distance progress towards the final destination. Ant based multipath backbone routing is proposed for load balancing in MANETs [31]. Source selects the multiple paths with maximum path preference probability using swarm based ant colony optimization technique. Path preference probability is estimated based on next hop availability, delay and bandwidth.
Many QoS aware routing protocols have been proposed based on multipath routing to improve network parameters like throughput, end-to-end delay, jitter, packet delivery ratio. However, the most effective method should improve these parameters as well as should be compatible with the MANETs.

Analysis of stability of multipath with single path approach
Most of the earlier approach as stated uses single path discovery. In such cases, if a source has data to send and route break occurs in between transmission, then source must reinitiate a route discovery process. This results in significant data loss and delay. Therefore, in many applications where delay and data loss is intolerable, multiple paths are desirable. Multiple paths can be used as an alternate to transfer data without any delay and data loss. Multiple paths also provide fault tolerance. Let us consider a single route R (x1, x2, x3........ xn). 'x1' is source node and 'xn' is destination node. Let the probability of route break be 'Pbr' caused by random node movement. Therefore, stability of route 'Pst' can be calculated as So, unstability of the route will be Here 0 ≤ 1-Pbr ≤ 1 and 1 ≤ a ≤ n Therefore 0 ≤ (1-Pbr)n ≤ (1-Pbr)a ≤ 1 So, it is clear that more the intermediate nodes in a route, route will be more unstable.
Punst (1) Let us compare route unstability between Punst(k) and Punst(k-1) as This clearly shows that Equation (8) clearly shows that more the common intermediate nodes in a multiple path scenario, worst will be the stability.

QoS Aware Routing
QoS is a set of mechanisms which tries to utilize various resources provided by the network efficiently to accomplish any particular application. QoS routing provides the required metrics as per the requirement of the application. QoS can be categorized by certain number of network metrics like throughput, end-toend delay, jitter and packet delivery ratio etc. QoS constraint model provides the best possible service as required by the application. In this paper, bandwidth constrained multipath routing has been proposed. A QoS based routing has been proposed that provides feedback about the available bandwidth throughout the route discovery. Multiple routes are selected on the basis of minimum bandwidth requirement as desired by that particular application.

Available Bandwidth Estimation
In bandwidth constrained QoS routing, path is discovered which fulfils the requirement of minimum available bandwidth throughout the route. There are several approaches by which end to end available bandwidth can be calculated. In our approach, end to end available bandwidth is calculated by minimum residual bandwidth among the intermediate nodes throughout the route. As each node shares its available bandwidth between its neighboring nodes, so it is difficult for individual host to calculate residual bandwidth throughout the path. A host will offer bandwidth guaranteed route only if residual bandwidth at a given host is known. However, calculation of residual bandwidth through 802.11MAC is still a difficult problem as the bandwidth is shared among neighbors. Even neighboring hosts are not aware of the traffic status of each other. QoS constrained routing protocol to calculate available bandwidth throughout the route has been proposed by Chen and Heinzelman [32]. Author estimated the residual bandwidth of the host by listening to the channel, the amount of idle time.
Let us consider a network consisting of 'n' number of mobile nodes i.e. N1, N2, N3,…… Nn. Each node in the network will have to maintain two hop neighbor routing tables. Firstly, the one hop neighbor table and then the two-hop neighbor table. Let us consider a mobile node Nx with one hop neighbor as Ny and two hop neighbor as Nz as shown in Figure1.

Figure1. Nx and its two hop neighbors Ny, Nz
Let consumed bandwidth of Ny be Bycons and consumed bandwidth Nz be Bzcons for all inflows and outflows of the data transmission by Ny and Nz. As Ny is considered as one hop from Nx, so there is a direct link from Nx to Ny. Similarly, there is a link directed from Ny to Nz as Nz is two hop neighbor from Nx. In order to maintain the two-hop neighbor table, these two neighboring host should exchange their one hop table along with their consumed bandwidth periodically via control packet Bicons. Bicons represents consumed bandwidth by node Ni i.e. ith node. In this approach, Ny and Nz have to exchange their routing table along with their consumed bandwidth Bycons and Bzcons respectively.

Figure2. Two hop neighbor table of node Nx
When Ny and Nz receives Bicons from Nx node, these nodes will reply with their consumed bandwidth Bycons and Bzcons respectively. Two hop routing table of node Nx is shown in Figure 2. When Nx receives the reply about currently consumed bandwidth Bycons and Bzcons from both its two hop neighbors Ny and Nz respectively, the residual bandwidth is easily calculated by subtracting consumed bandwidth of the two neighboring hops from maximum available bandwidth.
n Bresidual= (Bmax -∑Bicons) / Wf (9) i=0 Bresidual represents the available residual bandwidth, Bmax is maximum available bandwidth across a path and Bicons is bandwidth used by Ni. Wf denotes the weight factor as defined below in equation (2). The division of the residual bandwidth by the weight factor Wf is done due to 802.11MAC. The control messages like RTS, CTS and ACK are induced by MAC. These control messages also consume some T r a n s a c t i o n s o n N e t w o r k s a n d C o m m u n i c a t i o n s ; V o l u m e 5 , N o . 3 , J u n e 2 0 1 7   C o p y r i g h t © S o c i e t y f o r S c i e n c e a n d E d u c a t i o n U n i t e d K i n g d o m   7 bandwidth, that's why back off scheme is not successful for use of the entire bandwidth and also, collision of packets can be there.
The weight factor Wf can be calculated as Wf ={RTS+ CTS+ (Data+ MAChdr+ IPhdr)}/Data (10) All the terms like RTS, CTS…. are used to represent the size of these packets respectively. The value of the weight factor calculated above is more as fading errors can cause the retransmission of control or data packets.

Proposed Route Discovery
Let us consider a dense network of mobile nodes having same transmission range. The base protocol considered for our proposed route discovery is AODV. In this protocol, each mobile node has its own routing table based on reactive routing. Each node adds or updates a new route in a reactive manner. If a node has data to transmit to a destination, firstly it will check whether there is any route available in its routing table. If the source finds a route to destination, it can start transmission of data directly. If there is no route for that destination, the source node has to initiate a route discovery process. A Route Request (RREQ) packet is generated at source node and broadcasted in the network to find route to the destination. Each neighbor node that gets RREQ adds or updates its RREQ table and broadcasts the packet again. Intermediate nodes discard late RREQ packets that have same source node and destination sequence number. After receiving the RREQ packet, destination node replies with a Route Reply (RREP) packet to the source node through symmetric links. As source node receives RREP, the route is set up and source node starts transmission of data to the destination.

Implementing QoS in Route Discovery
Our approach is based on bandwidth constrained multipath node disjoint route discovery and fast route recovery. Node disjoint paths are those paths, which have only source and destination node common. No intermediate node is common in node disjoint paths as shown in Figure 3. The main advantage of node disjoint paths is that they are independent of each other. If any route break is there, data session can be resumed through one of the alternate multi paths available. To discover such paths, AODV RREQ header is modified. AODV control packet for route discovery process consist of request identifier, source address, destination address, destination sequence number, prefix size, hop count and lifetime of the route [6]. To initiate the QoS route discovery RREQ header has been changed to (REQ BW, AODV RREQ Header). REQ BW indicates required bandwidth to satisfy service requirements desired by the application. If the residual bandwidth on that link is greater than required bandwidth, it will forward the RREQ. Otherwise RREQ will be discarded by that node. After receiving RREQ, the whole procedure of host is shown in Figure 4. When any intermediate node receives RREQ packet, it first examines to see duplicate RREQ with same source node and destination sequence number to discover node disjoint paths. If duplicate RREQ is received by the intermediate node, RREQ will be discarded. Only the first received RREQ packet will be forwarded When RREQ packed is received by the destination node, the node sends a RREP. Each intermediate node forwards RREP to source node through symmetric links. As source node receives the RREP packet, it adds route in its routing table.

Proposed Route Reply
When a RREQ packet is received by the destination node, the node sends a RREP to the source. RREP follows the reverse path as stored in the RREQ received table in intermediate nodes. RREP packet in BCMR is modified with three more fields like (Node seq, Reply source, Path no., AODV RREP Packet). Node seq stores the list of all intermediate nodes through which RREP has traversed so far. Reply source field contains the address of the node from which RREP packet has been generated. Node seq and Reply source are added to avoid routing loops and to discover node disjoint paths. Path no. gives the number of path in sequence as received by the source node. When first RREP arrives at source node, source node stores the route in its routing table with Path no.1 and so on.

Proposed Route Maintenance
In MANETs, route link breaks frequently due to uncertain node mobility and packet collision. In conventional AODV whenever there is a route break during data transmission, it simply sends a Route Error (RERR) packet to its upstream node. After arrival of RERR packet at source node it initiates a new route discovery. In BCMR, when a node detects a route break, it will perform one hop route recovery process for smooth transmission of data. In this process, when a node detects a route break, it will send Immediate Route Request (IRREQ) to its one hop neighbours to transmit the data through alternate route.
If any node in one hop neighbour have route to destination, it will send Immediate Route Reply (IRREP) to the node and data will be transmitted through this alternate route. If there is no such route available, immediate upstream node will send a RERR to the upstream nodes. When RERR packet arrives at source node, source node invalidates this route related to that broken link and relays the data to alternate route available to the destination.

Performance Evaluation
Initially mobility of the nodes is varied as 0 m/s, 5m/s, 10m/s, 15m/s, 20m/s and 25 m/s. Number of nodes taken for simulation are 30 nodes and 50 nodes with different mobility scenario. Weight factor considered for BCMR and CAAODV (L) [32] is 1.2.

Performance Metrics
To compare the performance of our proposed approach, following performance metrics are considered.
Average end-to-end delay gives the average time delay that a data packet has consumed from the time it was sent by the source to the time delivered at the destination.
Average packet delivery ratio is the ratio of number of data packets received by the destination to the number of data packets actually sent by the source.
Normalized control overhead is the number of control packets sent to the network hop-wise to the number of data packets delivered at the destination.

Simulation Environment
The covered area for simulation environment is 1000m x 1000m. Different nodes communicate via radio signals having transmission range of 250m. Channel bandwidth taken is 2 Mbps. In our simulation, IEEE 802.11 is used as MAC layer protocol. Initially nodes are randomly distributed in the simulation area. Random waypoint mobility model determines the mobility of the nodes. Path loss model is Two Ray Ground model. For Constant Bit Rate (CBR) data sessions, node pairs are randomly selected with each CBR session generating 5 packets per second with 512 bytes as each data packet size. Table 1 gives the list of simulation parameters used for analysis of our proposed approach. To analyze the performance of our proposed routing protocol with different weight factors and compare with conventional AODV, CAAODV (H) [32] and CAAODV (L) [32]. Basically, two parameters are varied in our simulation scenario.

Results and Discussion
Initially mobility of the nodes is varied as 0 m/s, 5m/s, 10m/s, 15m/s, 20m/s and 25 m/s. Number of nodes taken for simulation are 30 nodes and 50 nodes with different mobility.
Average end-to-end delay Figure 5 and Figure 6 shows the plot of end-to-end delay vs. node mobility using the simulation parameters given in Table 5.1 with different number of nodes i.e. 30 nodes and 50 nodes respectively. From the comparison, we can observe that average end-to-end delay in our proposed approach is much less as compared with CAAODV (H), CAAODV (L) [31] and AODV with different number of nodes and mobility. Comparative results of end-to-end delay of our proposed node-disjoint scheme and other proposed schemes are shown in Figure 5 and Figure 6. Results show improvement in end-to-end delay with weight factor as 1.2 for BCMR and CAAODV (L) as shown. In proposed approach, end-to-end delay is minimized due to availability of alternate route for the destination. This eliminates the route rediscovery latency due to active route failure. As per the simulation results shown in Figure 5.5 and Figure 5.6, average of 32% and 36% improvement in terms of end-to-end delay is obtained for BCMR respectively as compared to AODV considering 30 and 50 nodes in the network. As compared to CAAODV (H) and CAAODV (L), we have obtained improvement of 9% and 10% respectively in BCMR for 30 nodes and in case of 50 nodes; we have obtained improvement of 11% and 13% respectively in BCMR. Normalized control overhead From Figure 7 and Figure 8 shows the plot of normalized control overhead vs. node mobility using the simulation parameters given in Table 1 with different number of nodes i.e. 30 nodes and 50 nodes respectively. From the comparison, we can observe that control overheads in our proposed approach is much less as compared with CAAODV (H), CAAODV (L) and AODV with different number of nodes and mobility. Results show improvement in control overheads with weight factor as 1.2 for BCMR and CAAODV (L) as shown. In proposed approach, control overheads are minimized due to bandwidth constrained routing for the destination. As per the simulation results shown in Figure 7 and Figure 8, average of 23% and 26% improvement in terms of control message overheads is obtained for BCMR respectively as compared to AODV considering 30 and 50 nodes in the network. As compared to CAAODV (H) and CAAODV (L), we have obtained improvement of 8% and 9% respectively in BCMR for 30 nodes in terms of control message overheads. Taking 50 nodes, we have obtained improvement of 10% and 12% respectively in BCMR as compared to CAAODV (H) and CAAODV (Listen). Average packet delivery ratio From Figure 9 and Figure 10 shows the plot of packet delivery ratio vs. node mobility using the simulation parameters given in Table 5.1 with different number of nodes i.e. 30 nodes and 50 nodes respectively. From the comparison, we can observe that packet delivery ratio in our proposed approach is significantly improved as compared with CAAODV admission scheme (Hello), CAAODV (Listen) and AODV with different number of nodes and mobility. Results show improvement in packet delivery ratio with weight factor as 1.2 for BCMR and CAAODV (Listen) as shown. In proposed approach packet delivery ratio is improved due to availability of alternate route for the destination. This minimizes packet loss during transmission of data. As per the simulation results shown in Figure 9 and Figure 10, average of 12% and 14% improvement in terms of packet deliver ratio is obtained with 1.2 as weight factor for BCMR respectively as compared to AODV considering 30 and 50 nodes in the network. As compared to CAAODV (H) and CAAODV (L), we have obtained improvement of 6% and 5% respectively in BCMR for 30 nodes and improvement of 7% and 5% respectively in BCMR for 50 nodes in terms of packet delivery ratio.

Conclusion
This paper proposes a bandwidth constrained multipath routing protocol for QoS provisioning. The use of node-disjoint multiple paths minimizes frequent route recovery and route rediscovery due to frequent route failures as in conventional AODV. Simulation results have shown significant improvements in terms of certain QoS parameters like end-to-end delay, control overheads and packet delivery ratio for different node mobility.
QoS is the most important issue for latest computer networks. As MANETs follows a distributed and uncertain environment, prioritized QoS is more suitable for such networks. Frequent link failure is major issue in MANETs, therefore alternate route strategies should be implemented as per QoS requirements.