Chapter 5 Computation of Multiple Node Disjoint Paths 5.1 Introduction In recent years, on demand routing protocols have attained more attention in mobile Ad Hoc networks as compared to other routing schemes due to their ability and efficiency. There exist many on demand routing protocols for Mobile Ad Hoc Networks. Most of the protocols, however, use a single route and do not utilize multiple alternative paths. Multipath routing allows the establishment of multiple paths between a single source and single destination node and when a path breaks an alternate path is used instead of initiating a new route discovery. Hence multipath routing represents a promising routing method for wireless mobile Ad Hoc networks. There exist some multipath Ad Hoc on-demand routing protocols such as AOMDV [120]. Compared to AODV, the AOMDV protocol performs better in terms of QoS metrics such as end-to-end Delay and Packet delivery ratio [121]. The route discovery of the original single path AODV protocol can be altered by using the path accumulation feature [122] of DSR to compute more than one path during route discovery. Among the multiple paths obtained at the destination, only node disjoint paths can be selected using the path matrix method [123]. The modified multipath AODV is compared with the original AODV. Multipath routing improves the QoS metrics like reliability, route Some part of this chapter published in the International Journal of Applied Information Systems, Vol.3, No.5, 2012, ISSN: 2249-0868 67
discovery frequency and end-to-end delay. The goal of the research work is to improve the performance of the existing Multipath Ad Hoc On-demand Routing Protocol by providing modifications to Route discovery mechanism so that the algorithm is multi-objective on demand QoS routing algorithm for Mobile Ad Hoc Network which will be highly adaptive, energy efficient, scalable, reliable and mainly reduces end-to-end delay in high mobility cases. 5.1.1 Unipath Routing in MANETs The unipath routing protocols [124] discover a single route between a pair of source and destination. A new route discovery is required in response to every broken route which leads to high overhead and latency. The most commonly used unipath routing protocols are Ad Hoc On-demand Distance Vector (AODV), Dynamic Source Routing (DSR), and Destination Sequenced Distance Vector (DSDV) [125]. Among these AODV and DSR are on-demand protocols whereas DSDV is a table driven routing protocol. In on-demand protocols [126][127], nodes compute routes only when they are needed. Therefore, on-demand protocols are more scalable to large dynamic networks. When a node needs a route to another node, it initiates a route discovery process to find a route. On demand protocols consist of two main phases: (i) Route discovery It is the process of finding a route between two nodes. If a source has no entry for a destination in its routing cache, it starts a route discovery process. The sample route discovery process using unipath routing protocol in MANET is shown in Figure 5.1. In order to send data from node S to node D, source node S first identifies the route to destination node D through the intermediate node B. Once the route is identified, then node S sends data along the path. 68
A A A S D S D S D B B B C C C Figure 5.1: An example of Route discovery process in MANET (ii) Route maintenance It is the process of repairing a broken route or finding a new route in the presence of a route failure. If a node tries to forward a message, but detects that there is a link break, i.e., the next node is no more reachable, it starts a route maintenance process. Figure 5.2 shows the sample route maintenance process in MANET. The node S sends data to node D along the established route through the node B. When the node D moves out of range of node B, then the route breaks. So a new path is identified from node S to node D through the intermediate node C. Now the data is transmitted from node S to D along the new path through node C. A A A S D S S B B B C C D C D Figure 5.2: An example of Route maintenance process in MANET 69
5.1.2 Multipath Routing in MANETs Multipath routing consists of finding multiple routes between a source and a destination node. These multiple paths between source and destination node pairs can be used to compensate for the dynamic and unpredictable nature of Ad Hoc networks. Multipath routing consists of three components: route discovery, traffic allocation and route maintenance. These three components are discussed below: (i) Route Discovery- Route discovery and route maintenance consist of finding multiple routes between a source and a destination node. Multipath routing protocols can attempt to find node disjoint, link disjoint, or non-disjoint routes. Node disjoint routes, also known as totally disjoint routes, have no nodes or links in common. Link disjoint routes have no links in common, but may have nodes in common. Non disjoint routes can have nodes and links in common. The Figure 5.3 shows two types of disjoint QoS paths from source node S to the destination node D. A A S D S D B C B C Path1 Path1 Path3 a) Node disjoint paths SAD, SBD b) Link disjoint paths SBD and SCD and SABCD Figure 5.3: Disjoint paths between Source S and Destination D 70
Figure 5.3 (a) shows the node disjoint paths and Figure 5.3 (b) shows the link disjoint paths. Among these, node-disjoint path is better, because it is the strongest measure of path independence. In this case if any node fails in between, then any of the remaining paths can be used for transmission through route maintenance without any problem. (ii) (iii) Traffic Allocation - Once the source node has selected a set of paths to the destination, it can begin sending data to the destination along the paths. The traffic allocation strategy used deals with how the data is distributed amongst the paths. The choice of allocation granularity is important in traffic allocation. The allocation granularity specifies the smallest unit of information allocated to each path. For instance, a perconnection granularity would allocate all traffic for one connection to a single path. A per-packet granularity would distribute the packets from multiple connections amongst the paths. A per-packet granularity results in the best performance [128]. This is because it allows for finer control over the network resources. It is difficult to evenly distribute traffic amongst the paths in the per-connection case, because all the connections experience different traffic rates. Path Maintenance - Over time, paths may fail due to link/node failures or, in Ad Hoc networks, due to node mobility. Path maintenance is the process of regenerating paths after the initial path discovery. It can be initiated after each path failure, or when all the paths have failed. Some multipath protocols use dynamic maintenance algorithms to constantly monitor and maintain the quality or combined QoS metric of available paths. 5.1.3 Benefits of multipath routing (i) Fault tolerance Multipath routing protocols can provide fault tolerance by having redundant information routed to the destination via alternative paths. This reduces the probability that communication is 71
disrupted in case of link failure. More sophisticated algorithms employ source coding [129] to reduce the traffic overhead caused by too much redundancy, while maintaining the same degree of reliability. This increase in route resiliency largely depends on metrics such as the diversity, or disjointedness, of the available paths. Disjoint routes are discussed in the section 5.3. (ii) Load balancing When a link becomes over utilized and causes congestion, multipath routing protocols can choose to divert traffic through alternative paths to ease the burden of the congested link. (iv) Bandwidth aggregation By splitting data to the same destination into multiple streams, each routed through a different path, the effective bandwidth can be aggregated. This strategy is particularly beneficial when a node has multiple low bandwidth links but requires a bandwidth greater than an individual link can provide. End-to-end delay may also be reduced as a direct result of larger bandwidth. (v) Reduced delay For wireless networks employing single path on demand routing protocols, a route failure means that a new path discovery process needs to be initiated to find a new route. This results in a route discovery delay. The delay is minimized in multipath routing because backup routes are identified during route discovery. 5.1.4 Ad Hoc On-Demand Multipath Distance Vector Routing (AOMDV) The main idea in AOMDV is to compute multiple paths during route discovery. It is designed primarily for highly dynamic Ad Hoc networks where link failures and route breaks occur frequently. When single path on-demand routing protocol such as AODV is used in such networks, a new route discovery is needed in response to every route break. Each route discovery is associated with high overhead and latency. This inefficiency can be avoided by having multiple redundant paths available. Now, a new route discovery is needed only when all paths to the destination break. To 72
keep track of multiple routes, the routing entries for each destination contain a list of the next hops along with the corresponding hop counts. All the next hops have the same sequence number. For each destination, a node maintains the advertised hop count, which is defined as the maximum hop count for all the paths. This is the hop count used for sending route advertisements of the destination. Each duplicate route advertisement received by a node defines an alternative path to the destination. To ensure loop freedom, a node accepts an alternative path to the destination, only if it has a lower hop count than the advertised hop count for that destination. AOMDV can be used to find node-disjoint or link-disjoint routes [120]. To find node-disjoint routes, each node does not immediately reject duplicate RREQs. Each RREQ arriving via a different neighbor of the source defines a node-disjoint path. This is because nodes cannot broadcast duplicate RREQs, so any two RREQs arriving at an intermediate node via a different neighbor of the source could not have traversed the same node. In an attempt to get multiple link-disjoint routes, the destination replies to duplicate RREQs regardless of their first hop. To ensure link-disjointedness in the first hop of the RREP, the destination only replies to RREQs arriving via unique neighbors. After the first hop, the RREPs follow the reverse paths, which are node-disjoint and thus link-disjoint. The trajectories of each RREP may intersect at an intermediate node, but each takes a different reverse path to the source to ensure link-disjointedness. The performance study of AOMDV relative to AODV under a wide range of mobility and traffic scenarios reveals that AOMDV offers a significant reduction in delay. It also provides reduction in the routing load and the end-to-end delay. 5.1.5 Experimental Results The performances of two on-demand routing protocols, viz., AODV and AOMDV are compared using NS-2 simulation. The two models used for the simulation are the Traffic generation model (TGM) and the Mobility generation model (MGM). 73
Delay (in ms) PDR (in %) The performance of AODV and AOMDV has been analyzed by fixing simulation parameters as given in Table 4.1: node speed to 10m/s, pause time to 40 seconds, single CBR flow per node and by varying the node density from 10 to 100 nodes. The mobile nodes are placed randomly within a 1000mX1000m area. Radio propagation range for each node is 250m. 5.1.5.1 PDR vs. number of nodes: Table 5.1 and Figure 5.4 show the PDR for each protocol versus number of nodes at low mobility. Table 5.1: PDR versus no. of nodes for different protocols PDR in % No. of nodes AODV AOMDV 10 99.5 99.9 20 99.2 99.8 30 99 99.75 40 98.75 99.6 50 98 99.5 60 97 99.3 70 95.5 98 80 90 96.5 90 82 95 100 73 93 120 100 80 60 40 20 PDR vs. number of nodes 0 No. of nodes AODV AOMDV Figure 5.4: PDR versus No. of nodes 5.1.5.2 Delay vs. number of nodes: Table 5.2 and Figure 5.5 show the end-to-end delay for each protocol versus number of nodes at low mobility. Table 5.2: Delay versus no. of nodes for different protocols Delay in milliseconds No. of nodes AODV AOMDV 10 65 60 20 82 70 30 88 82 40 102 90 50 116 105 60 125 112 70 145 130 80 162 145 90 184 160 100 194 175 Delay vs. number of nodes 250 200 150 100 50 0 No. of nodes AODV AOMDV Figure 5.5: Delay versus No. of nodes 74
5.1.5.3 Control overhead: Total number of control packets such as route requests (RREQs), route replies (RREPs) and route error (RRER). AOMDV allows for more RREQ and RREP packets in the network in order to build multiple paths to each destination for each node. The number of control packet sent for movement of 20 m/s and 30 traffic sources with zero pause time is shown in Table 5.3. Table 5.3: Control overhead of AODV and AOMDV Protocol Total no. of Control packets AODV 1,35,202 AOMDV 95,054 5.1.6 Analysis of Simulation Results The performance of AODV and AOMDV routing protocols is compared and analyzed using NS-2.34 simulator. The QoS metrics Average delay and Packet delivery ratio is measured by varying the number of nodes from 10 to 100. a) PDR comparison: It is observed that the PDR of AOMDV is better in increasing the number of nodes as compared to AODV at low mobility. For example, at 80 nodes, the PDR of AODV and AOMDV is 90% and 96.5% respectively. But, at high mobility the PDR of AOMDV reduces to 75%. b) Delay comparison: It is observed that the delay of AOMDV is better than AODV at low mobility. For example, at 80 nodes, the delay of AODV is 162ms. Whereas for AOMDV, delay is 145ms. But at high mobility conditions, the delay of AOMDV may reach up to 250ms. Thus, from the experimental analysis, it is observed that the AOMDV routing protocol is better in performance as compared to AODV. But the performance degrades, as the mobility and traffic is increased. This is due to, more routing overhead caused in the MANET. 75
c) Control overhead: AODV allows only for a single RREP packet, for the first RREQ the destination node received to be sent back via the reverse route it arrived in. By considering 3 types of control messages(rrep, RREQ and RERR) the overall control overhead of AOMDV is less than AODV is given in the Table 5.3. 5.2 Establishing Path Accumulation features in AODV DSR and AODV are the two well-known on-demand routing protocols for MANETs. A major disadvantage of these two protocols is, the problem associated with routing of packets through flooding, resulting in congestion and unnecessary traffic in the network. There exists minimal flow control in the network. Thus, DSR and AODV are combined into one hierarchical routing protocol [130]. AODV can be modified to enable path accumulation as in the case of DSR, during the route discovery cycle. Each node also updates its routing table with all the information contained in the control messages. Each RREQ and RREP contains a source route for the nodes along the path, so that each node can have a routing table entry to the rest of the nodes. The main benefit of obtaining the additional routing entries is to reduce the route discovery overhead by eliminating some of the RREQs that would be required to discover these nodes. Since RREQs are the major source of control overhead due to flooding the whole network, any reduction in RREQs is expected to improve the performance significantly. The tradeoff is that the RREQ and RREP packet header will become larger to accommodate the source route. Figure 5.6 clarifies the difference between route discovery without path accumulation and route discovery with path accumulation [131]. 76
A A A A RREQ A B C D E E E E E RREP Figure 5.6 (a): Route Discovery Mechanism without Path Accumulation A A B A B C A B C D RREQ A B C D E B C D E C D E D E E RREP Figure 5.6 (b): Route Discovery Mechanism with Path Accumulation Figure 5.6: Route Discovery Mechanism in AODV 5.3 Proposed method of Multiple Node Disjoint Paths for a MANET Consider the sample network and possible paths from source A to destination D given in Figure 5.7. For this graph we can construct the path matrix as shown in Figure 5.8. In the path matrix, number of paths is placed in rows, and number of vertices is placed in columns. For every path we identify the vertices. If the vertex is there in the path, then we assign the value 1 for the corresponding vertex otherwise we assign the value 0. B C A Source Node D Destination Node A G D E F Figure 5.7: Sample graph and possible paths 77
The possible paths are: (i) P1 = ABCD (ii) P2 = ACD (iii) P3= AGD (iv) P4 = AFD (v) P5 = AEFD (vi) P6 = ACFD (vii) P7 = ABCFD A B C D E F G Weight P1 1 1 1 1 0 0 0 4 P2 1 0 1 1 0 0 0 3 P3 1 0 0 1 0 0 1 3 P4 1 0 0 1 0 1 0 3 P5 1 0 0 1 1 1 0 4 P6 1 0 1 1 0 1 0 4 P7 1 1 1 1 0 1 0 5 Figure 5.8: Path matrix In the above path matrix, paths P2, P3 and P4 contain the lowest weight 3. Among these 3 paths all the paths are treated as node disjoint paths. Suppose by default, among them first path is selected, i.e., P2. Then rest of the node disjoint paths are computed using the Hamming distance matrix [122]. Hamming distance is the difference in bit pattern of each path with the other paths. The hamming distance matrix for the above graph is represented in Figure 5.9. 78
P1 P2 P3 P4 P5 P6 P7 P1 0 1 3 3 4 2 1 P2 1 0 2 2 3 1 2 P3 3 2 0 2 3 2 4 P4 3 2 2 0 1 1 2 P5 4 3 3 1 0 2 3 P6 2 1 2 1 2 0 1 P7 1 2 4 2 3 1 0 Figure 5.9: Hamming Distance Matrix In the Hamming distance matrix, the corresponding P2(row2) the maximum value is under P5, i.e., 3(circled in the Figure 5.9). So P5 is selected as the second path. Now, in order to find the third path, we have to find the maximum of the sum of P2 and P5 as shown in Figure 5.10. P2 1 0 2 2 3 1 2 P5 4 3 3 1 0 2 3 Sum 5 3 5 3 3 3 5 Figure 5.10: Determining the third disjoint path Now the maximum sum is 5. It is under column1. So the third path selected is P1. In order to find the next node disjoint path the maximum sum of P2, P5 and P1 are used as shown in Figure 5.11. 79
P2 1 0 2 2 3 1 2 P5 4 3 3 1 0 2 3 P1 0 1 3 3 4 2 1 Sum 5 4 8 6 7 5 6 Figure 5.11: Determining the fourth disjoint path It is observed that the maximum sum is 8. It is under column3 (circled in the Figure 5.11). So the fourth path selected is P3. Similarly the rest of the node disjoint paths can be computed using the same procedure. The algorithm first selects the node disjoint paths according to the Path matrix initially. For the given MANET, using the path matrix, there are 3 paths (P2, P3 and P4) having the same weight value (lowest weight). So, actually P2, P3 and P4 will be the first set of node disjoint paths. Upon selecting these three paths only, second set of node disjoint paths are selected, based on the hamming distance. Total number of node disjoint paths computed is based on the Timestamp. During the data transmission, initially whichever path is available first (primary path) is taken. When primary path breaks, it takes up the next alternative path from the routing table. 5.4 Algorithm for computing Multiple Node Disjoint Paths by making use of Path accumulation feature in AODV protocol (MQARP) Suppose n is the number of mobile nodes and N is the set of mobile nodes, N={N1,N2,..,Nn}. Assume that node Ni seeks to find a path to node Nj and Nt receives the RREQ packet, where Ni, Nj, Ntϵ N and 1<i, j, t<n and i j. 80
1. Broadcast the RREQ message from Ni to the intermediate node Nt if they are reachable. 2. [Using path accumulation feature of DSR in AODV] Accumulate each intermediate node address in the RREQ message and forward the RREQ message until destination Nj is reached. 3. Repeat steps 1 and 2 until all RREQ messages are received at the destination Nj or a time stamp is reached. 4. [Computing Path matrix by placing paths as rows and nodes as columns] For each path or row Begin If a node is present in the path put the value 1. Else place a zero in the corresponding column. Select the next path or next row. End 5. From the Node disjoint path matrix, for every path find the weight, by finding the row sum. 6. Now select the minimum weight from the path matrix. Include this path as a first path in the list of Node disjoint paths. 7. [Compute the rest of Node disjoint paths by making the Hamming distance matrix where row and column indicates the paths] For each path or row Begin Initialize the counter to zero. Take the bit pattern from the Path matrix for this path and Compare this bit pattern with the rest of the path s bit pattern and with the same path. If there is no change in corresponding bit position, then skip that bit Else add 1 to the counter. Store the counter value in the corresponding column of the Hamming distance matrix for that path. Select next path or row. 81
End 8. Using the hamming distance matrix compute the rest of the node disjoint paths by taking the maximum column value under the first node disjoint path or row. The path having the maximum value is taken as the second node disjoint path. 9. Add the corresponding bit positions in the bit patterns of the computed node disjoint paths or compute the column sum of the corresponding entries. 10. In the column sum, if all the values are same then select any one of the values and take the corresponding column s path as the third node disjoint path else go to Step 11. 11. Find the maximum value of the column sum, and take the corresponding column s path as the third path. 12. Repeat steps 9 to 11 to compute rest of the node disjoint paths, until the maximum limit or Time stamp reached. 13. At the node Nj forward the computed node disjoint paths to the source node Ni. 14. Stop 5.5 Experimental Results The performance of AOMDV and MQARP has been analysed by fixing node speed to 25m/s, single CBR flow per node, pause time to 10 seconds and by varying the node density from 10 to 100 nodes. The mobile nodes are placed randomly within a 1000mX1000m area. Radio propagation range for each node is 250m. 5.5.1 PDR vs. number of nodes: Table 5.4 and Figure 5.12 show PDR for each protocol versus number of nodes. 82
Delay (in ms) PDR (in %) Table 5.4: PDR versus no. of nodes for different protocols at high mobility No. of nodes PDR in % AOMDV MQARP 10 99 99.5 20 98.5 99 30 97 98.2 40 92 95 50 90 94.3 60 86 92 70 81 88 80 75 85 90 68 82 100 63 78.5 PDR vs. number of nodes 120 100 80 60 AOMDV 40 MQARP 20 0 No. of nodes Figure 5.12: PDR versus No. of nodes 5.5.2 Delay vs. number of nodes: Table 5.5 and Figure 5.13 show the end-to-end delay for each protocol versus number of nodes. Table 5.5: Delay versus no. of nodes for different protocols at high mobility Delay in ms No. of AOMDV MQARP nodes 10 82 78 20 90 86 30 105 98 40 118 112 50 125 118 60 138 123 70 150 135 80 180 155 90 225 176 100 250 189 Delay vs. number of nodes 300 250 200 150 100 50 0 No. of nodes AOMDV MQARP Figure 5.13: Delay versus no. of nodes 5.5.3 Control overhead: Table 5.6 shows overall control overhead of AOMDV and MQARP protocols at high mobility situations. 83
Table 5.6: Control overhead of AOMDV and MQARP Protocol Total no. of control packets MQARP 84,075 AOMDV 95,054 5.5.4 Analysis of Simulation Results The performance of AOMDV and MQARP routing protocols is compared and analyzed using NS-2.34 simulator. The QoS metrics Average delay and Packet delivery ratio are measured by varying the number of nodes from 10 to100. a) PDR comparison: It is observed that the PDR of MQARP is better in increasing the number of nodes as compared to AOMDV at high mobility. For example, at 80 nodes, the PDR of AOMDV and MQARP is 75% and 85%, respectively. b) Delay comparison: It is observed that the delay of MQARP is better than AOMDV at high mobility. For example, at 80 nodes, the delay of AOMDV is 180ms. Whereas for MQARP, delay is 155ms. The reason behind this is, MQARP identifies node disjoint paths, due to link break or node failures another route will be readily available in the routing table for data transmission. The route discovery latency of MQARP is lesser than that of AOMDV and the AOMDV which is available now is taking care of only link disjoint paths. The link disjoint paths provide fault tolerance, but cannot be used for resource sharing. The node disjoint path is the strongest measure of path independence. c) Control overhead comparison: As far as control overhead is concerned, AOMDV and MQARP are almost same for RREQ messages. But the control overhead in terms of RREP is less for MQARP than AOMDV. 84
The total number of paths computed for data transmission, depends upon the network topology and the reachability of nodes. For traffic sharing node disjoint paths are necessary. Since the multiple paths computed by AOMDV protocol is not node disjoint, it cannot be used for simultaneous data transmission. In the experiments, the paths computed are node disjoint. i.e., among the multiple computed paths, there is no common node except source and the destination. It s not always feasible to have multiple node-disjoint paths. For example, consider a sample MANET shown in Figure 5.14. B C S A E G H D F Figure 5.14: Sample MANET According to the algorithm only one node-disjoint path is considered. There exist three paths between S and D. They are: S->A->E->G->H->D, S->A->B->C->G->H->D and S->A->F->G->H->D. According to the path matrix method, two paths contain same weights. Assuming that, for example S->A->E->G->H->D path is obtained first at the destination, then this path is taken as the node disjoint path even though there exist one more path between S and D. At the destination, if the weights of the paths are same, then they are checked for node disjointedness. 5.6 Summary In the single path routing protocol as the number of nodes are increased, performance degrades in terms of QoS metrics. Multipath routing allows the establishment of multiple paths between a single source and single destination node and when a path breaks an alternate path is used 85
instead of initiating a new route discovery. Hence multipath routing represents a promising routing method for wireless mobile Ad Hoc networks. Multipath routing achieves load balancing and is more resilient to route failures. Compared to unipath AODV, the performance of AOMDV is better in terms of QoS metrics. The AOMDV works for both link disjoint and node disjoint paths. In the case of original AODV, since at a time only one path is identified, due to link break or node failure the current path becomes invalid, so immediate route discovery has to take place. In the case of original AOMDV, the path/paths that are existing in the routing table are not node disjoint. i.e., whenever there is link break or node failure in the network, the current path becomes invalid. In this work the AODV protocol is modified to get multiple node disjoint paths during route discovery using the path accumulation feature. In MQARP, when the primary path breaks, it takes up the next alternative path from the routing table. This message is given to the source node in terms of error messages. The purpose of the proposed work is not only to compute the maximum number of node disjoint path, but also the paths computed must be delay aware, link reliable and energy efficient. The experimental results show that the performance of the proposed MQARP better in all aspects as compared to the original AODV and AOMDV. This is due to the infrequent route discovery process. The performance needs to be improved further for high mobility and high traffic situations. So the MQARP is taken as the base protocol to improve further the QoS metrics. 86