Abstract: Background and Objective: Mobile ad hoc networks require routing protocols to discover routes between nodes, which is challenging due to unpredictable network environments. Over the years, different versions of the conventional routing models have been proposed, but still proficient routing is quite essential. This paper examined DSR (dynamic source routing), a well-known routing protocol in mobile ad hoc networks for possible improvements. This study determined a comparative performance analysis of the standard and revised routing models of the DSR. Materials and Methods: These analyses were performed using NS-3.25 (Network Simulator, version 3.25) and with the help of various performances evaluating metrics such as the throughput, PDR (packet delivery ratio), EED (end to end delay), PL (packet loss) and NRL (normalized routing load). Results: Average numerical values of different performance evaluating metrics of either routing models of the DSR were obtained for different network node densities from the simulation results. Performances of the standard and revised dynamic source routing models were evaluated and compared through the graphs. Conclusion: This study concluded performance improvements in the revised DSR routing model as compared to its standard version for different node population scenarios.
INTRODUCTION
Mobile Ad hoc Networks (MANETs) are new-age wireless networks that operate without the need for centralized network equipment (i.e., network gateways, routers, network access points, etc.). This is because MANET nodes act as both routers and hosts. Mobile ad-hoc networks are self-forming and self-healing when needed, they are simple and economical to use. These ad hoc networks are very helpful in diverse applications such as military applications, emergency and rescue operations, collaborative and distributed computing, wireless sensor networks, wireless mesh networks and hybrid wireless networks1. These networks are very useful in establishing instant communication at battlefields, flood and earthquake hit areas2. To meet such critical applications, a better-performing routing protocol is very much needed for accomplishing efficient and error-free communication paths between the nodes. In the past few decades, different versions of the conventional routing models were proposed but still an efficient routing is quite essential.
Mobile ad hoc networks operate on multi-hop radio transmissions, a source node intensive to communicate with a destination node has to communicate across multiple hops to get the shortest and most efficient path to that destination3,4. A connected mobile ad hoc network with nine numbers of nodes ‘N’ has been demonstrated in Fig. 1. Scalability of mobile ad hoc networks depends upon packet forwarding capacity and the size of the network5.
The process of finding a route from a source node to the destination node in a network is called routing6, where, routing protocols ensure efficient and error-free routes. Some features such as resource allocation, node mobility and dynamic topographies mark the routing procedure a key challenge3.
This paper presents simulation-based comparative analysis on DSR protocol. The unicast routing protocols communicate using a single channel whereas multicast protocols communicate using multiple channels7. The DSR operates as a unicast, reactive routing method that relies on a single channel8,9.
DSR: Dynamic source routing is an on-demand type routing protocol which supports unicast communication and takes advantage of the source routing algorithm10,11. The DSR is also called a demand driven routing protocol developed for multi-hop relaying MANETs12-14. In DSR, bandwidth restrictions on control over heads are achieved by removing periodic table update messages, it does not require periodic transmissions of beacons or hello messages1. It employs source routing, as the complete path is included in the DSR heading and ropes unidirectional links. The DSR is used in Microsoft mesh networks15.
Path discovery in DSR: If a source node ‘S’ initiates a communication with the destination node ‘D’, it generates RREQ (Route Request) message and floods it throughout the network because node ‘S’ does not have a valid route to ‘D’. The RREQ messages hold information such as, source address, source ID (Identity), destination address and destination ID with an exclusive sequence number generated by the source node ‘S’. Upon receiving the RREQ, intermediate nodes check the sequence number and either generate a RREP (Route Reply) message for the source node ‘S’ or forward the RREQ message to other nodes in the network. Intermediate nodes generate RREP only when they have valid routes to the destination. Before forwarding the RREQ, intermediate nodes add their address and ID in the RREQ and store it in their cache.
The process of route discovery in DSR was illustrated in Fig. 2, dotted lines represent connectivity between the nodes, forward arrows signify RREQ transmission and backward arrows denote transmission of the RREP message. When a source node ‘S’ floods the RREQ, immediate node ‘P’ receives the RREQ, but it has no valid path to ‘D’ and hence forwards it to nodes ‘R’, ‘T’ and ‘Q’ including its address to the source path. Node ‘Q’ does not have a valid path to node ‘D’, it forwards the RREQ to nodes ‘U’, ‘V’ and node ‘D’. Node ‘D’ is the destination, so it generates the RREP message and passes it to the source node through nodes ‘Q’ and ‘P’. The shortest route between node ‘S’ and ‘D’ is S-P-Q-D. During the path discovery process, every node which forwards the RREQ, stores the path information in the send buffer. The RREQ, RREP and RERR (Route Error) messages live a short period and are discarded by the nodes after their time-out period. During path breaks, the source node re-initiates the route discovery process16. During path discovery, every node acquires new path information, repetition of hop paths is avoided to elude control overheads. Path cashing in DSR involves cache structure, cache capacity and cache timeout17.
Route maintenance in DSR: The DSR maintains routes in five different strategies namely, circulation of RERR messages, packet salvaging18, auto-route shortening, RREQ hop limits and preventing RREP squalls.
The process of route maintenance in DSR was illustrated in Fig. 3. Here, the source node ‘S’ initiates a route discovery by generating a RREQ (S-D) to fetch the route up to the destination node ‘D’. The shortest available route from ‘S’ to ‘D’ is S-B-F-D, but there is a route break between nodes ‘F’ and ‘D’.
| Fig. 1: | Connectivity in a mobile ad hoc network |
| Fig. 2: | Path discovery in DSR |
| Fig. 3: | Route maintenance in DSR |
Node ‘F’ generates a RERR (F-D) message and spreads it all around the network. Upon receiving RERR (F-D) message, the source node ‘S’ looks into its cache for other available routes to the destination node ‘D’. If there are no routes in the cache, the source node again generates a new RREQ to get a route up to the destination node ‘D’. Other alternate routes to the destination node ‘D’ are S-C-D and S-A-E-D. Upon detecting a broken link, a node initiates route maintenance by generating a route error message and floods it. Then the node deletes the broken link route from its cache3.
MANET routing protocols are helpful in establishing shortest routes and their maintenance. Over the years, much research were conducted on mobile ad hoc network routing protocols, many of these researches were based on performance comparison studies of some well-known conventional routing protocols considering network parameters, mobility models, transmission region, transmission range and type of offered load etc. In this study, a combination of network and DSR parameters was considered to achieve possible improvements in routing data packets considering different network sizes.
MATERIALS AND METHODS
Study area: This study was conducted recently at the Electronics Engineering Department of Indian Institute of Technology (Banaras Hindu University). Duration of this study was three months (from January, 2023 to March, 2023). These research works were performed with the help of network simulator -3 (NS-3, version 3.25) over the server grade CentOS (version 5.1) Linux operating system on an Acer server. Obtained packet data (results from the simulation based experiments) were used to calculate performance evaluating metrics. Various experiments were performed for different sets of node densities to test the revised DSR model.
Performance evaluating metrics: There are many performances evaluating metrics available to measure the performances of the routing protocols in MANETs. Some of them considered for this analysis are, the throughput, PDR (packet delivery ratio), EED (end-to-end delay), PL (packet loss) and the NRL (normalized routing load)8,19.
Throughput: The number of data packets transmitted by the network between the source node and the destination node is known as the throughput. Higher values of throughput provide improved network performance. The unit of the throughput is Kbps (Kilobits per second):
| (1) |
Packet delivery ratio: The packet delivery ratio is the sum of data packets received to the sum of data packets sent. Higher values of this metric deliver better performance, deriving unit of PDR is percentage:
| (2) |
End-to-end delay: End-to-end delay can be defined as the average time interval between data packets generated at the source and their actual transfer at the destination. To get better protocol performances, delay values must be lesser. The unit of this metric is milli seconds (ms):
| (3) |
Packet loss: Packet loss can be termed as the difference between total sent data packets and total received data packets. For better and improved protocol performances, values of packet losses must be minimal. The unit of the PL is in numbers:
PL = Total sent data packets-Total received data packets |
(4) |
Normalized routing load: Normalized routing load is the fraction of sent routed data packets to the received data packets. The NRL can be derived in numbers, higher NRL values deliver enhanced protocol performances though higher NRL values indorse lesser efficiency concerning consumption of bandwidth8,20:
| (5) |
Simulation tools: Simulation based analysis on the dynamic source routing protocol (DSR) were performed with the aid of NS-3. The NS-3 is discrete event-based open source software available for educational purposes including research. It holds the GNU GPLv2 license and is openly accessible for R and D (Research and Development) activities. Usage and debugging of this network simulator are easy as “it forms a compact simulation core”. This network simulator provides all the support from simulation conjuration to the results for simulation based experiments. It has extensive research provisions on IP (Internet Protocol) and non IP based computer networks. Network simulator-3is popularly used for IP based wireless network simulations by researchers. It was developed from scratch using C++ along with optional Python bindings. As compared to NS-2 (Network Simulator-2), NS-3 has improved simulation reliability. In this simulator, APIs (Application Program Interfaces) of NS-2 have been replaced10.
Network modeling: The 50 numbers of moving nodes were placed inside a rectangular network region of size, 300×1500 m. Mobile nodes ensure their mobility at a velocity of 20 mps within this region as per the random way point mobility model (RWPMM). Ten numbers of fixed source-sink connections were used for data transmission. The communication channel capacity was set to 2 Mbps with 7.5 dBm transmit power. The simulation run time was set to 200 seconds (transient period: 50 seconds). The network parameters used in the simulation were shown in Table 1.
Protocol modeling: The standard DSR routing protocol has many core parameters which are responsible in achieving better functioning of the protocol. Here, standard attributes of some core parameters have been altered to study protocol behavior and to test the protocol performances in terms of throughput, PDR, EED, PL and NRL. The DSR parameters considered for this study were shown in Table 2.
Statistical analysis: In this study, a typical ad hoc network model was simulated for different node sets using the standard and revised DSR protocol models separately. Obtained packet data of each protocol was used to determine the performance evaluating metrics. Different performance evaluating metrics were calculated in Table 3 and performances of the protocol models were compared.
| Table 1: | Network parameters used |
| Network parameter | Value |
| MANET routing protocol | DSR |
| Data transmission rate | 2 Kbps |
| Rate of Wi-Fi | 2 Mbps |
| Number of source/sink connections | 10 |
| Network expanse (rectangular) | 300×1500 m |
| Node mobility speed | 20 m sec–1 |
| Node transmit power | 7.5 dBm |
| Wi-Fi mode | Ad-hoc |
| Mobility model | Random way point mobility model |
| Data packet size | 64 Bytes |
| Simulation time | 200 sec |
| Node pause time | No pause time |
| Node concentration | 30, 40, 50, 60, 70, 80, 90 and 100 |
| Table 2: | DSR parameters used in simulation |
| Protocol parameter | Assigned value |
| MaxSendBuffTime | 25 sec |
| MaxMaintLen | 40 |
| MaxMaintTime | 20 sec |
| RouteCacheTimeout | 400 sec |
| MaxEntriesEachDst | 25 |
| SendBuffInterval | 550 sec |
| NodeTraversalTime | 30 msec |
| RreqRetries | 12 |
| MaintenanceRetries | 3 |
| NonPropRequestTimeout | 25 msec |
| MaxSalvageCount | 12 |
| BlacklistTimeout | 5 sec |
| GratReplyHoldoff | 2 sec |
| RequestPeriod | 550 msec |
| MaxRequestPeriod | 15 sec |
| MinLifeTime | 2 sec |
| RetransIncr | 25 msec |
| MaxNetworkQueueSize | 500 |
RESULTS AND DISCUSSION
In this study, a performance comparison of the standard and revised DSR routing protocols was presented, whereby attributes of core performance parameters of the standard DSR routing protocol were revised and modeled as the revised DSR. Experiments on dynamic source routing models were carried out to study their behavior and performances for different node population scenarios. Various experiments were conducted to test the revised DSR model considering 30, 40, 50, 60, 70, 80, 90 and 100 sets of network nodes (small and large node sets) and obtained results were compared with the results of the standard model. A screen shot of a running program script was shown in Fig. 4. As compared to the standard DSR routing model, the revised model gained improved network throughput and better data packet delivery between the source and destination nodes. The revised model has attained minimum delays in delivering data packets to the destination nodes with the least packet losses and routing overheads.
Compared to previous studies, where only the performance of different standard routing protocols were analyzed by varying different network parameters6-8,10-13,18-21, offered traffic22, media access control protocols and node velocities23 using network simulator-2, the present study managed to obtain better performing revised DSR routing model taking protocol parameters into account with the help of network simulator-3. Similar studies on AODV and OLSR routing protocols were accomplished by researchers Lakshman et al.24,25 considering protocol parameters.
Packet data obtained from the simulation experiments were utilized to calculate different performance-evaluating metrics discussed in the materials and method section. The same was shown in Table 3.
| Table 3: | Data sheet showing performances of the DSR models |
|
Throughput in kbps |
PDR (%) |
EED (msec) |
Packet loss |
NRL |
||||||
| No. of nodes |
STD.DSR |
REV.DSR |
STD.DSR |
REV.DSR |
STD.DSR |
REV.DSR |
STD.DSR |
REV.DSR |
STD.DSR |
REV.DSR |
| 30 |
12.37 |
12.98 |
61.87 |
64.88 |
15.41 |
13.53 |
2288.00 |
2107.00 |
0.619 |
0.649 |
| 40 |
12.75 |
12.84 |
63.75 |
64.20 |
14.22 |
13.94 |
2175.00 |
2148.00 |
0.638 |
0.642 |
| 50 |
11.41 |
12.55 |
57.07 |
62.73 |
18.81 |
14.85 |
2576.00 |
2236.00 |
0.571 |
0.627 |
| 60 |
12.93 |
12.98 |
64.65 |
64.88 |
13.67 |
13.53 |
2121.00 |
2107.00 |
0.647 |
0.649 |
| 70 |
11.92 |
12.94 |
59.58 |
64.68 |
16.96 |
13.65 |
2425.00 |
2119.00 |
0.596 |
0.647 |
| 80 |
10.94 |
12.51 |
54.68 |
62.57 |
20.72 |
14.96 |
2719.00 |
2246.00 |
0.547 |
0.626 |
| 90 |
12.86 |
12.93 |
64.32 |
64.63 |
13.87 |
13.68 |
2141.00 |
2122.00 |
0.643 |
0.646 |
| 100 |
13.04 |
13.30 |
65.20 |
66.48 |
13.34 |
12.60 |
2088.00 |
2011.00 |
0.652 |
0.665 |
| PDR: Packet delivery ratio, EED: End-to-end delay, NRL: Normalized routing load, STD: Standard and REV: Revised | ||||||||||
| Fig. 4: | DSR script under execution NS-3 manet-routing-compare script |
| Fig. 5: | Throughput |
| Fig. 6: | Packet delivery ratio |
| Fig. 7: | End-to-end delay |
| Fig. 8: | Data packet loss |
| Fig. 9: | Normalized routing load |
Throughput: As compared to standard DSR, the revised DSR had better network throughput. The maximum bandwidth passed through the network was 13.30 Kbps for 100 sets of nodes. The throughput achieved for other node sets were presented in Table 3. Throughput graphs were shown in Fig. 5.
PDR: For different node densities, REV.DSR has shown constant and better packet delivery from the source node to the destination node. Packet delivery ratio of the STD.DSR was found to fluctuate. The PDR results were demonstrated in Fig. 6.
EED: Revised DSR has encountered lesser delays in delivering the data packets between the source and the destination nodes whereas, the standard DSR has met with large delays during data transmission sessions. Delay scenarios of both routing models were presented in Fig. 7.
PL: Data packet losses encountered in both the routing models were explored in Fig. 8. The revised model has shown better performances by achieving minimum packet losses as compared to the standard model.
NRL: Normalized routing load scenarios in either routing model were shown in Fig. 9. The REV.DSR has shown better results as compared to the STD.DSR by having minimal routing overheads. However, better-performing routing model may devour additional bandwidth.
Implications, applications, recommendations and limitations of the study: For different sets of the node populations (20, 40, 60, 80, 100 and 120), performance of the standard and revised DSR model was compared. As compared to the standard DSR model, the revised DSR gained improved throughput, better packet delivery, least end-to-end delays, minimum packet losses and least routing overheads. The revised dynamic source routing protocol will be helpful in establishing error-free and quick communication routes in critical applications such as military, emergency and rescue operations (during floods, earthquakes) etc. This study can be taken forward considering diverse network scenarios, other core protocol parameters and additional MANETs routing protocols. Results achieved in this study are limited to general network parameters and the attributes of some core parameters are taken into account.
CONCLUSION
According to simulation results and performance calculations, the revised DSR model has shown improved performance in terms of performance evaluating parameters as compared to its standard version. Attribute values considered for the revised DSR model may help improve the standard DSR routing protocol. Further research on standard and revised DSR routing models can be taken ahead for large number of network nodes with QoS (quality of service) concerns.
SIGNIFICANCE STATEMENT
Many researches were conducted on comparative analysis of MANET routing protocols considering network parameters. In this paper, network parameters as well as core attributes of the protocol were considered to study the behaviour of the revised DSR and compare its performance with the standard DSR. The routing protocols are considered as key protocols because they establish paths between network nodes effectively. The MANETs are very helpful at locations where network infrastructure does not exist. Some applications include, military operations, emergency rescue operations (during flood, earthquake, etc.). A better routing protocol is required to strengthen these networks in terms of effective connectivity. The purpose of this study is to enable scientists and engineers to consider better performing routing model while designing protocol suits.
ACKNOWLEDGMENT
The authors would like to thank Dr. Ramesh Singh Yadava, Computer Centre, Banaras Hindu University for providing his support in this study.