1 Multipath Routing Protocol for Congestion Control in Mobile Ad-hoc Network Nilima Walde, Assistant Professor, Department of Information Technology, Army Institute of Technology, Pune, India Dhananjay Singh, Student of 2 nd Year, IT Deptt., Army Institute of Technology, Pune,India ABSTRACT Congestion occurs in mobile ad hoc networks (MANETs) with limited resources. Congestion control is a key problem in mobile ad-hoc networks. In such networks, packet transmissions suffer from interference and fading, due to the shared wireless channel and dynamic topology. Mobile ad hoc networks are wireless networks without the support of any fixed network infrastructure, which are typically characterized by no centralized access, selforganizing, rapid development, dynamic topology, multiple hops and so on. Due to such characteristics, MANETs have much potential for many applications. To answer those challenges, many routing algorithms in MANETs were proposed, so here by we are trying to identify the best routing algorithm which will improve the congestion control mechanism. Keywords MANETs, Ad-hoc network, congestion control I. INTRODUCTION A wireless ad hoc network is usually defined as a set of wireless mobile nodes dynamically self-organizing a temporary network without any central administration or existing network infrastructure. Since the nodes in wireless ad hoc networks can serve as routers and hosts, they can forward packets for other nodes if they are on the route from source to the destination. To prepare for this promising future, besides other issues, routing is an important problem in need of a solution that not only works well with a small network, but also sustains efficiency and scalability as the network gets expanded and the application data gets transmitted in larger volume. Though essential, routing in MANETs is a nontrivial matter. MOBILE AD-HOC NETWORK An ad-hoc network is a collection of wireless mobile hosts forming a temporary network without the aid of any stand-alone infrastructure or centralized administration. Mobile Ad-hoc networks are self-organizing and selfconfiguring multihop wireless networks where, the structure of the network changes dynamically. This is mainly due to the mobility of the nodes. Nodes in these networks utilize the same random access wireless channel, cooperating in a friendly manner to engaging themselves in multihop forwarding. The node in the network not only acts as hosts but also as routers that route data to/from other nodes in network. In mobile ad-hoc networks where there is no infrastructure support as is the case with wireless networks, and since a destination node might be out of range of a source node transmitting packets; a routing procedure is always needed to find a path so as to forward the packets appropriately between the source and the destination. Within a cell, a base station can reach all mobile nodes without routing via broadcast in common wireless networks. In the case of ad-hoc networks, each node must be able to forward data for other nodes. This creates additional problems along with the problems of dynamic topology which is unpredictable connectivity changes. When too many packets are present in a part of a subnet, the performance degrades; this situation is called as congestion. Congestion in a network may occur when the load on the network i.e the number of packets sent to the network is greater than the capacity of network. Congestion is a problem that occurs on shared networks when multiple users contend for access to the same resources (bandwidth, buffers, and queues). II. MINIMUM OVERHEAD MULTIPATH ROUTING PROTOCOLS Multipath routing protocols that have been proposed for discovering and using multiple paths with minimum routing overhead fall into this category. Multipath routing protocols inherently need to discover as many paths as possible. To discover as many path as possible, a multipath routing protocol uses an additional type of control messages other than route request, route reply and route error messages. Thus, a lot of control overhead messages are generated in the network to discover and to maintain these paths. That is why discovering multiple paths with low overhead is the main objective of the following multipath routing protocols. 2.1 SPLIT MULTIPATH ROUTING The main objective of SMR is to reduce the frequency of route discovery processes and thereby reduce the control overhead in the network. The protocol uses a per packet allocation scheme to distribute a load into multiple paths. When a destination node receives route request packets from different paths, it chooses multiple disjoint routes
2 and sends replies back to the source. The basic route discovery mechanism of the DSR protocol is used in the SMR protocol. Figure 1 shows the paths taken by RREQs using this technique. We can select more disjoint paths from routes available in Figure 2. Fig1.Overlapped multipath routes, fig 2. multiple route with maximally disjoint path. 2.1.1. Route Selection Method: In our algorithm, the destination selects two routes that are maximally disjoint. More than two routes can be chosen, but we limit the number of routes to two in this study. One of the two routes is the shortest delay route; the path taken by the first RREQ the destination receives. When receiving the first RREQ, the destination records the entire path and sends a RREP to the source via this route. 2.1.2 ROUTE MAINTENANCE: A link of a route can be disconnected because of mobility, congestion, and packet collisions. It is important to recover broken routes immediately to do effective routing. In SMR, when a node fails to deliver the data packet to the next hop of the route or not receiving passive acknowledgments it considers the link to be disconnected and sends a ROUTE E RROR( RERR) packet contains the route to the source, and the immediate upstream and downstream nodes of the broken link. Upon receiving this RERR packet, the source removes every entry in its route table. If only one of the two routes of the session is invalidated, the source uses the remaining valid route to deliver data packets. When the source is informed of a route disconnection and the session is still active, it may use one of the two policies in re-discovering routes: 1) initiates the route recovery process when any route of the session is broken 2) initiates the route recovery process only when both routes of the session are broken. The first scheme reconstructs the routes more often and produces more control overhead than the second scheme, but the former provides multiple routes most of the time and be robust to route breaks. route to the destination. One drawback of this scheme is out of order delivery and re-sequencing burden on the destination. We believe, however, that cost-effective reordering buffers are easily implementable. We decided to use the per-packet allocation approach because it is known to work well in most networks [8], and most of all, it is fairly difficult to obtain the network condition (such as available bandwidth) in ad hoc networks to apply more sophisticated schemes. 2.3 Simulation Environment We evaluate and compare the performance of the following protocols: SMR-I: SMR which performs the route recovery when any route to the destination is invalidated. SMR-2: SMR which performs the route recovery only when both routes to the destination are invalidated. 2.4. Results and Analysis Figure 3 shows the throughput of each protocol in packet delivery fraction. Packet delivery ratio is obtained by dividing the number of data packets correctly received by the destinations by the number of data packets originated by the sources. Many data packets are dropped during this process and more delay is needed to discover correct routes. 2.2 Allocation Granularity : When the source receives a RREP after flooding the RREQ, it uses the first discovered route to send buffered data packets. When the second RREP is received, the source has two routes to the destination, and can split traffic into two routes. We use a simple per-packet allocation scheme when there are more than one available
3 information with regards to the hop count of the route, and the first and last hop relays on that route. As in the AODV protocol, we use sequence numbers to prevent loops. When a source node initiates an RREQ, it increases its sequence number seqsrcsrc (seqij represents node i's latest sequence number known to node j) and the destination's sequence number seqdstsrc by one. These two sequence numbers are indicated in the RREQ packet and denoted by seqsrc RREQ and seqdst RREQ respectively. Each time the destination node receives an RREQ packet, it computes a new sequence number: Fig 2.Performance evaluation of SMR Figure 4 illustrates the number of packets dropped by each protocol. Both data and control packets are measured. The reasons for packet drops can be incorrect route information, mobility, collisions, and congestion. DSR cannot maintain precise routes and drops more packets as nodes move more often (i.e., less pause time). The usage of state routes from caches is the major reason of DSR packet drops. Both SMR schemes have considerably fewer packet drops compared with DSR. SMR-2 has fewer packet drops than SMR-I. 3. MULTIPATH- AODVM We propose modifications to the AODV protocol so as to enable the discovery of multiple node-disjoint paths from a source to a destination. Instead of discarding the duplicate RREQ packets, intermediate nodes are required to record the information contained in these packets in a table which we refer to as the RREQ table. Fig. 3.1 (a) Structure of each RREQ table entry in AODVM, (b) Structure of each routing table entry in AODVM We see that intermediate nodes make decisions on where to forward the RREP messages (unlike in source routing) and the destination, which is in fact the originator of these messages is unaware as to how many of these RREP messages that it generated actually made it back to the source. Thus, it is necessary for the source to confirm each received RREP message by means of a Route Confirmation message (RRCM). The RRCM message can, in fact, be piggybacked onto the first data packet sent on the corresponding route and will also contain The destination then generates an RREP message that contains a sequence number seqdst RREP. In order to route information reliably in cases wherein, multiple node-disjoint paths are not available, a certain number of ``reliable nodes'' should be placed in the network. In the next section we describe the functionality of these reliable nodes and describe a methodology to control their trajectories to achieve higher routing reliability. IV. AODVM WITH PATH DIVERSITY We first describe AODVM and some modifications we made to AODVM. We then describe an extension to AODVM that can find routes with smaller correlation factors called AODVM with Path Diversity (AODVM/PD). 4.1 AODVM AODV-Multipath (AODVM) [2] is an extension to AODV for finding multiple node-disjoint paths. In AODVM, intermediate nodes are not allowed to send a route reply directly to the source. Also, intermediate nodes do not discard duplicate RREQ packets. Instead, all received RREQ packets are recorded in an RREQ table at the intermediate nodes. The destination sends an RREP for all the received RREQ packets. An intermediate node forwards a received RREP packet to the neighbor in the RREQ table that is along the shortest path to the source. 4.2. MODIFIED AODVM Each node can store multiple routes to a destination, as in AODVM. In our protocol, a route is not only identified by the destination node, but also by the source node. A set of routes in a node's routing table is uniquely identified by the three-tuple. The structure of a routing table entry in our modified AODVM is shown in Figure 2. Because AODV uses hop-by-hop routing rather than source routing, the source cannot readily specify a certain route that a packet should take. More specifically, each intermediate node can route the packet along whatever path it has to the destination in its routing table. Therefore, by identifying all routes by a source and destination address, the source can establish a set of routes
4 that are used exclusively for routing packets from the source to the particular destination. The following are additional modifications we have made to AODVM. 4.2.1. ROUTE DISCOVERY. When a source needs a new set of routes to the destination. If the RREP contains a larger source sequence number than one in the routing table, indicating a new route discovery, then the route list is reinitialized to contain just the new route. If the source sequence numbers are equal, indicating a new route from a current route discovery, the new route is inserted in the route list. 4.2.2. DELAYED ROUTE REPLIES. Our protocol requires an adequate delay between RREPs sent by the destination. In our protocol, as in AODVM, when the source receives an RREP, it sends an RRCM message along the path traversed by the RREP. When the destination receives the first RREQ for a particular route discovery, it immediately sends an RREP. However, upon receiving subsequent RREQs, the destination does not immediately send back an RREP, but instead stores the RREQs in its RREQ table. 4.2.3. GRATUITOUS ROUTE REPLIES. DSR allows intermediate nodes the ability to send gratuitous route replies. If upon inspection of the route in an overheard data packet a node detects that it is a downstream node of a route, the node sends a gratuitous route reply to the source of the packet. This is because a shorter route can be achieved by routing directly to the node. For mobile situations, since the topology is constantly changing, gratuitous replies can provide a significant performance enhancement. Therefore, we implement this same feature in our version of AODVM. When a node receives a gratuitous reply, it verifies the route, and sets its next hop field for the route equal to the source of the reply. Fig 4.1 Structure of routing table entry for modified AODVM 4.3. AODVM/PD AODVM/PD is based on our modified version of AODVM, and finds node-disjoint paths with lower correlation. The main objective of AODVM/PD is to find more diverse paths, or more specifically, to minimize the correlation factor of the discovered paths. To achieve this, AODVM/PD uses information gained from overheard packets, much like AODVM. Note that the correlation factor of two paths is determined by the number of nodes along both paths that are in range of each other. Suppose node A overhears an RREP from node B, as in Figure 4.Node A can determine that it is in range of node B, and that node B is along a potential path. If another route for this same route discovery were to go through node A, the A-B link would contribute one to the correlation factor. In Figure 4, if a route were to go through node A, both links A-B and B-C would contribute to the correlation factor. Clearly, it is desirable to avoid routing RREPs to nodes that have over heard many previous RREPs. Local correlation factor (LCF): Measure of how many RREPs associated with a given route discovery that a node has overheard. Area correlation factor (ACF): A weighted average of a node s local correlation factor and the average of its neighbors local correlation factors. It is defined as α*my_local_corr + (1-)*avg_local_corr_of_nghbrs Correlation threshold (CT): When a node s area correlation factor is over the correlation threshold, the node is no longer allowed to participate in any routes for the particular route discovery. Besides just a node s local correlation, the area correlation factor takes into account the correlation of the surrounding nodes. Each route discovery is uniquely identified in the RREQ table by the three-tuple {Source ID, BroadcastID, Destination ID} (see Figure 4(a)). These values are included in all control messages sent for the particular route discovery, such as the RREQs and RREPs. The route discovery process of AODVM/PD like AODVM, consists of 3 phases: route request, route reply, and route confirmation. The route request phase is the same as in AODVM. A. ROUTE REPLY : When a node overhears an RREP, ents its local correlation factor by 1 (see hen a node receives an RREP addressed d its area correlation is over the correlation threshold, threshold, it sends a Route Discovery Error (RDER message to the sender. Otherwise, the node selects the neighbor from the RREQ table with the shortest hop count to the source, and forwards the RREP to that neighbor. When there is a tie, and more than one path has the shortest hop count, the neighbor with the lowest local
5 correlation factor is chosen. Like AODVM, the source sends an RRCM upon receiving an RREP. B. ROUTE CONFORMATION : When a node overhears an RRCM, it broadcasts a Correlation (CORR) packet (Figure 4(b)) containing its local correlation factor. When a node receives a CORR packet from a neighbor, it updates the local correlation of its neighbor in the RREQ table to the value contained in the CORR packet. The node then calculates its area correlation factor (see Figure 4D). If its area correlation factor is above the correlation threshold, then it broadcasts another CORR packet with the OVER_THRESHOLD flag set to true. When a node receives a CORR packet with the OVER_THRESHOLD flag set, it deletes the sending node from its RREQ table. When the destination receives an RRCM, it sends the next RREP to a neighbor from its RREQ table. The purpose of the route confirmation phase is to purge nodes with a relatively high area correlation from RREQ tables. demonstrates that our algorithm, we can control the level of correlation between the discovered paths by varying the threshold values. With thresholds of 2.5 and 3, the correlation is nearly the same. This is plausible, because at a certain point, no nodes will go over the de mobility. In this model, a node chooses a random point in the network, and moves towards that point at a constant speed. 4.4. PERFORMANCE EVALUATION We now present our simulation model, along with a performance comparison of AODVM and AODVM/PD. 4.4.1. Optimal Correlation Threshold for AODVM/PD. In this simulation, we try to determine an optimal correlation threshold for AODVM/PD. The optimal threshold for an application should find paths with lower correlation as compared to AODVM, while still being able to find a suitable number of paths. Normalized correlation facts. The maximum number of multiple paths is set 3. Therefore, if more than 3 paths are found, the correlation factor is only calculated for the 3 shortest s found. The results are then averaged over all the sources. This
6 CONCLUSION Multipath routing can be used in ad hoc networks to help achieve lower end-to-end delays and better fault tolerance.. Providing multiple routes helps minimizing route recovery process and control message overhead. Our protocol uses a per-packet allocation scheme to distribute data packets into multiple paths of active sessions. This traffic distribution efficiently utilizes available network resources and prevents nodes of the route from being congested in heavily loaded traffic situations. AODVM enable the computation of multiple node-disjoint paths without incurring the overhead generated by link-state routing methods. V. FUTURE SCOPE The mobile ad hoc networks have been a subject of quite a number of investigations in recent years. Most of these investigations have been motivated by the need to design an efficient routing protocol for an ad hoc network. A good routing protocol needs to provide reliability and energy efficiency with low control overhead. To ensure reliability, load balancing and QoS, multipath routing protocols have been proposed for MANET. The surveyed protocols showed that multipath routing can improve network performance in terms of delay, throughput, reliability and life time. Yet it is hard to find a single protocol or a set of protocols that can improve all these performance parameters. Selection of a multipath routing protocol depends on a particular application and tradeoffs. Some of the objectives are energy efficiency, low overhead, reliability and scalability. With this, we can get what has been investigated, and we can identify which protocol to use, and what are the trade-offs. VI. REFERENCES [1] Survey of multipath routing protocols for mobile ad-hoc networks by Mohammad Tarique a, Kemal E.Tepe b,sesanadibi c,shervinerfani b, 2009 [2] Yao-Nan Lien and Ho-Cheng Hsiao, A New TCP Congestion control mechanism over wireless ad- Hoc Network by Router-Assisted Approach IEEE 2007 [3] Carthy PM, Grigoras D. Multipath Associativity Based Routing. In: Proceedings of the second annual conference on wireless on-demand network systems and services (WONS), 2005. p. 60 79. [4] Lim H, Xu K, Gerla M. TCP performance over multipath routing in mobile ad hoc networks. In: IEEE international conference on communication (ICC), vol. 2, 2003, p. 1064 8. Costa LHMK, de Amorim MD, Serge F. Reducing latency and overhead of route repair with controlled flooding. ACM Wireless Networks 2004;10(4):347 58.