Throughput Comparison In Large Scale Networks

Published Date: 02 Nov 2017

CHAPTER 3

In this section, we determine the three different application-specific protocols that we have introduced in the previous section. The proposed CRP protocol is analyzed using five different parameters namely packet delivery ratio, jitter, average delay, node density and throughput through varying the node speed, pause time, and number of nodes. We also provide the intuition justifying why they work well by examining a non-adaptive version of CRP with statically selected, fixed values for the zone radius. The analysis and the discussions for each parameter are as given in the following subsection.

As mentioned earlier, the overall goal of this simulation study is to evaluate and analyze the performance of CRP (Consecutive Routing Process) that comprises EDYMO, FSR and ZRP over Mobile Ad-hoc Networks (MANETs) environment. The simulations have been performed using NS2 simulations that provide scalable simulation results for MANET. The simulation has been made over a terrain of 1500m × 1500m area by varying the number of nodes and its node speed in the MANET. Moreover, in this section, the performance of CRP is analyzed and demonstrated based on the results obtained from the simulation A number of experiments are performed to explore the performance of these protocols with respect to jitter. Simulations are performed by varying Packet size and keeping mobility high.

The mobility in the environment was simulated using a random-waypoint mobility model, wherein each node randomly chooses a point in the field and moves towards it at a randomly chosen velocity. The node then pauses for a specified period at the destination before continuing the same pattern of motion. In our simulations, velocities ranged between 0 m/s and 20 m/s, while the pause time was set to 0 seconds, which corresponds to constant motion. We controlled the mobility rate by changing the number of mobile nodes within the network

3.1 ANALYSIS OF PROTOCOLS WITH THROUGHPUT AND DELAY

The nine protocols that are described above have been taken for the analysis. Mainly, the analysis was made with the parameters such as throughput and delay. Throughput is a dimensional parameter that denotes the average rate of successful message delivery with respect to time over a communication channel. It also provides the information whether the data packets are correctly delivered to the destination or not. While analysis, it is found that at the node densities of 200, 300, 400 and 500, the STAR and RIP protocols provided lower throughput, whereas the rest shows consistent throughput in packet delivery.

Figure 3.1: Throughput Comparison in large scale networks

Figure 3.1 exemplifies the throughput comparison of nine protocols that we have taken for analysis.

Concerning end-to-end delay analysis in figure 3.2, it is given that AODV and Bellman ford protocols are having higher end-to-end than others. Typically, end-to-end delay denotes that the time taken for a packet to be transmitter through a network from source to destination. It is also analyzed that the LANMAR and LAR protocols affect the speed of simulation in large scale networks. The analysis mainly concerns the network speed and communication effectiveness. Selection of these protocols producing higher delay that tends to the possibility of packet drop and the requirement of fault tolerance approach. The figure 3.2 presented below shows the comparison of aforementioned nine protocols in the evaluation of end-to-end delay.

Figure 3.2 : Average End-to-End Delay Comparison

From the work of module 1, the effective protocols for packet transmission has been found with respect to throughput and delay. Moreover, we have chosen three protocols for enduring the rest of the process. Those protocols are Fisheye State (FSR), Dynamic MANET On-demand (DYMO) and Zone Routing Protocol (ZRP).

3.2 EXPERIMENTAL RESULTS OF EDYMO

All the results shown below are an average over 20 simulation runs. We predicted the effectiveness of the proposed EDYMO protocol in terms of packet delivery ration, end-to-end delay, and throughput at different levels of mobility. For this analysis we used CRB to maintain the congestion level in the MANET. The EDYMO protocol is compared with the DYMO protocol based on the following quantifying measures.

Throughput

Packet Delivery Ratio

Delay

Energy consumption

The graph presented in the Figure 3.3 expresses the throughput of DYMO and EDYMO protocol. Here, the throughput is measured in terms of bits/sec (bps). It explicitly denotes that the proposed enhanced DYMO outperforms the general DYMO. The throughput increases with the increase in the number of nodes, i.e. the throughput is directly relational to the number of nodes. In this experiment, the number of nodes is increased by 5 from 10 to 50. The maximum throughput attained only when there are 50 nodes in the network for both the reactive protocols. EDYMO has the throughput as 670 bps when there are 50 nodes in the network. Similarly, the 544 bps is obtained at the same 50 node scenario. The throughput of the proposed is greater for all set of nodes than the existing routing protocol.

Figure 3.3 : Throughput EDYMO vs. DYMO

Similar to throughput comparison, the efficiency of the proposed protocol is measured using the factor packet delivery ratio. The graph in Figure 3.4 depicts that the PDR for both the proposed and existing DYMO protocol. This measure expresses the ratio between number of data packets that are sent by the source node and total number of data packets received by the destination. It is explicit from the figure that the PDR rate of proposed technique is higher than the existing technique. The PDR values given in the graph represent the average value taken from 20 runs. The PDR values are given in percentage with respect to number of nodes. Delivery ratio is also directly proportional to the number of nodes.

Figure 3.4 : Packet Delivery Ratio DYMO vs. EDYMO

The most important quantifying measure is the delay in delivering the packets. The experiments were conducted to determine the delay for the proposed and existing protocol. It is measured in terms of milliseconds. It is clear from the Figure 3.7 that the delay experienced by the packets while using the DYMO protocol is higher than the proposed protocol. The delay is measured with respect to number of nodes in the network. As the nodes in the network increases the delay in delivering the packets from the source to sink node is increased. This is explicitly viewed in the graph of Figure 3.5.

Figure 3.5: Delay Experienced by EDYMO vs. DYMO

Figure 3.6: ENERGY CONSUMPTION

From the above figure 3.6 shows the Energy Consumption comparison between DYMO and EDYMO. The energy consumption of DYMO shows higher energy consumption whereas EDYMO consumes less energy. Hence EDYMO shows the reduction of energy when compared with DYMO.

3.3 EXPERIMENTAL RESULTS OF CRP WITH COMPARISON

3.3.1 PACKET DELIVERY RATIO

This measure is used to detect the delivery ratio of the packets that are sent by the source station and the resultant value of this will be higher if maximum packets are delivered successfully. On the other hand if the packets losses due to increased channel contention then the value for PDR will go down. With PDR we estimated the reliability of packet transmission in the network. It can be defined as the percentage of data packets that are successfully delivered to the destination node from the transmitter node. Here, we measured the delivery ratio of the packets and their results are presented in the figure3.7. This analysis is carried out through the process of varying the node speed and the number of nodes in the network.

Figure 3.7 : PDR for various Node speed

The PDR rate varies drastically with number of nodes. As the number of nodes increases the network contention increases which interfere the packets that are routed from source to destination node. Similar to number of nodes, the node speed also has impact on delivery ratio. The percentage of delivery of data packet is higher when the number of nodes in the network and the node speed is lower. This can be realized from the above graph that, when there are 100 nodes in the network the percentage of delivery of packets is ranged between 93% - 96%. Whereas, the range of delivery ratio is 40%-93% when there exits 500 nodes in the network. Likewise, the percentage of packet delivery is ranged between 84-91%, 35% - 95%, and 27% - 95%, when there are 200, 300 and 400 nodes respectively. The percentage of delivery of packets is computed by varying the node speed, which ranges from 10mps, 20mps, 30mps, 40mps, and 50mps. It is implicitly understood from the figure that the number of nodes and the delivery ratio rate are inversely proportional to each other.

Similarly, we have computed the packet delivery ratio by varying the pause time. Pause time is a duration for which all the nodes hold the same positions at waypoints. We have used the random waypoint model that produces the waypoint at random. The nodes presented in the network moves to a specific random waypoint at a given speed and when hits the point, it chooses another waypoint at random and begins moving towards it. Figure 3.8, represent the packet delivery ratio at different pause time.

Figure 3.8: PDR obtained by varying the pause time

It is shown in the above figure that the PDR values are higher when the number of nodes is lesser while varying the pause time and at constant mobile speed. The performance of packet delivery depends on three factors namely (1) Number of Nodes, (2) Node speed, and (3) Pause time. Therefore, it is required to choose optimal number of nodes along the suitable mobile speed and pause time to improve the percentage of delivery rate.

In addition to the above analysis for PDR, we compared the PDR rate for CRP with three other protocol namely (1) EDYMO, (2) ZRP, and (3) Fisheye. Figure 3.9 portrays the comparison of those protocols. For analyzing the comparison we have taken 20 nodes and vary the node speed.

Figure 3.9 : Comparison of CRP with existing protocols in terms of PDR

From the above figure it is evident that the proposed protocol has higher delivery ratio than the hybrid protocol (ZRP), reactive protocol (DYMO), and fisheye technique. Our proposed technique has the delivery rate of above 90% whereas the other existing protocols have below 90%. The experiment is carried out by varying the node’s node speed.

With the comparison we can conclude that, even when the mobile speed is higher the performance of the proposed protocol in terms of delivery is extraordinary in all aspects.

3.3.2 THROUGHPUT

It is one of the dimensional parameters of the network which gives the fraction of the channel capacity used for useful transmission selects a destination at the beginning of the simulation i.e., information whether or not data packets correctly delivered to the destinations. The maximum number of packets made, by using each protocol in a finite simulation time is analyzed. Throughput is referred as the total amount of data transfer from one node to another node in the network within a specified amount of time. We computed the throughput by dividing the total number of packets delivered by total number packets originated in a unit time. Throughput can be computed from the equation (1).

(1)

We measured the throughput in two different aspects. One aspect is through varying the mobile speed and another through varying the pause time of the nodes. Here, we first discuss the MPR’s throughput in terms of mobile speed. We determined the throughput values by varying the mobile speed as well as the number of nodes in the network. Figure 3.10 shows the performance of throughput while varying the mobile speed.

Figure 3.10: Throughput of CRP with respect to Node speed

The above figure illustrates that the throughput increases as the node speed decreases. The rate of data transfer is higher when the number of nodes as well as the node speed is lesser. When the network has 100 nodes the experiment is conducted and computed the throughput through varying the node speed of nodes as 10mps, 20mps, 30mps, 40mps, and 50mps. Similarly, the number of nodes is varied and determined the throughput. When there is huge number of nodes in the network then, the data transfer rate is decreased since the amount of control packets transferred is increased. These packets occupy a huge amount of bandwidth, so it becomes harder to transfer the data packets. Because of this reason the throughput gets decreased when the number of nodes increased. Moreover, the throughput is also computed by varying the pause time of the nodes, which is shown in the figure 3.11Depending on number of nodes in the network the data rate is changed.

Figure 3.11:Throughput of CRP with respect to Pause Time

In addition to that, we have also compared the throughput analysis of CRP with other routing protocols namely (1) DYMO, (2) ZRP, and (3) Fisheye. Figure 3.12 shows the result of the comparison. It is illustrated that the throughput for the proposed routing protocol is higher than the other existing protocols. As the node speed increases the rate of data transfer is gradually decreasing for our protocol. Whereas, the existing protocols have comparably lower data transfer rate. It is also shown that the fisheye and DYMO protocols function better even when the mobility rate is higher from their stands. Though they have greater performance they are still lower than our proposed protocol. Therefore, our proposed protocol performs much better than the existing protocol. The throughput analysis is carried out on having 20 nodes and varying the node speed.

Figure 3.12:Comparison of CRP with existing protocols in terms of throughput

3.3.3 JITTER

In general, Jitter is the undesired deviation from exact periodicity of an expected periodic signal in the field of electronics and telecommunications, frequently in relation to a position clock source. Jitter may be perceived in characteristics such as the signal amplitude,the frequency of consecutive pulses, or point of periodic signals. Jitter is a substantial, and usually undesired, factor in the scheme of almost entire communications links.

Delay jitter, evaluated as the variance of inter-arrival time between packets. It is claimed as a critical performance metric in various real-time applications. For a case, applications involving periodic transmission of sensor data, such as object tracking, as well as multi-media applications for transmitting uncompressed voice streams are highly sensitive to the variance in packet arrival times. There is a trade-off between the overhead entailed by the two routing components and the delay performance they provide at different network conditions.

CRP proactive component exhibits lower variance in latency in highly mobile networks. The reactive component reconstructs routes upon route failures resulting in a temporary halt of data flow. Consequently, frequent route-failures increase jitter. In low mobility scenarios, the reactive component offers the desired jitter performance at a very low overhead compared to the proactive component. This motivates an adaptation strategy to utilize the trade-off and provide good jitter performance at low overhead as the network conditions vary. It is also been noted that, in addition to route failures, jitter is also affected by congestion in the network. Currently, our protocol does not take into account the effect of congestion on delay jitter. We computed the jitter in terms of seconds. The Figure 3.13 presented below shows the delay jitter with varying node speeds. It is obvious from the figure that jitter will be high for low node speed of nodes as 10, 20, 30, 40 and 50 respectively. It is shown from the graph that the delay jitter rate increases when there is an increase in node speed rate in the applied network.

Figure 3.13 : Jitter for various node speeds

As the above figure, we have also found the jitter values through varying the pause time of the nodes. The output of that is portrayed as graph in Figure 3.14.

Figure 3.14:Jitter with respect to Pause Time

Figure 3.15 portrays that the evaluation of jitter for various protocols. This graph is drawn for twenty nodes and varying the mobile speed.

Figure 3.15:Comparison of CRP with existing protocols in terms of jitter

4.2.4 AVERAGE END-TO-END DELAY

End-to-End delay is used to predict the time required to transfer the packet across the network from the transmitter node to the destination node. Generally the end-to-end delay consists of two parts namely variable and fixed part. The fixed part has the following components.

Propagation Delay

Packetization Delay

Coding delay

Whereas the variable part is defined as the sum of all the queuing delays that a packet encounters at the output part of the router. We have computed the end-to-end delay in terms of its average. The end-to-end delay is measured in two different angles through varying the speed of the node and pause time. Figure 3.16 represents the delay that has occurred while varying the node speed. It is clear from the graph that the proposed CRP has much lower delay when the nodes have low mobility speed. The mobility speed is varied from mps. As the mobility speed of the node is increased the delay also increased. The delay also depends on the number of nodes. When the network has more number of nodes then, a packet will have multiple-hop to transfer the packet to the destination. At each intermediate node the packet encounters a queuing delay that increases the overall delay. So the increase in number of nodes in the network will increase the delay.

Figure 3.16: Average End-to-End delay for various Node speed

The measure of delay with respect to the second angle is analyzed and their results stated through graph in the figure 3.17.

Figure 3.17:Average End-to-End delay of CRP

with respect to Pause Time

In addition to these analyses, we have also compared the proposed protocol’s delay with the existing protocol namely ZRP, FSR, DYMO, and Fisheye. The comparison graph is given in the figure 3.18.

Figure 3.18:Comparison of CRP with existing protocols - Average End-to-End delay Protocol

For carry out the experiment we have taken 20 nodes as total in the network and vary the mobile speed. Figure 10 shows that the proposed technique functions better than all other existing protocols. Our proposed technique has much lower delay on comparing with other protocols. This expresses that the packets are delivered quickly than other existing protocol.

3.3.5 NETWORK DENSITY

Table 3.1 provides the details of the density of the nodes proposed system. This is analyzed in the proposed system also since the density has higher impact on the delay factor

Table 3.1 Node density

Number of nodes

Simulation time

Real time

20

200

84

40

200

109

60

200

125

80

200

146

100

200

163

Table 3.2 SIMULATION TIME VS REAL TIME

Table 3.2 shows the comparison of simulation time with real time during simulation time. From the above table it can be concluded if the number of nodes increases there is a variation in real time while the simulation time remains same.

3.4 CONCLUSION AND FUTURE WORK

CONCLUSION

The work of this research involves in developing an efficient routing protocol that adapts to dynamic network characteristics and enhances the performance of MANET. As described earlier, MANET is a self-configuring infrastructure-less network of mobile devices linked by wireless. Concerning the routing protocols of MANET, it is claimed that the protocols are widely categorized under three categories namely, proactive, reactive and hybrid. The predominant intention of the work is analyzing some efficient protocols of those categories and developing an efficient Consecutive Routing Protocols. For that affair, the research work has been structured in three phases. The first phase of work is accomplished as study phase about nine efficient routing protocols. The protocols Ad-hoc On-demand Distance Vector (AODV), Fisheye State (FSR), Dynamic MANET On-demand (DYMO), Source Tree Adaptive Routing (STAR), Routing Information Protocol (RIP), Bellman Ford, LANd Mark Ad-hoc Routing Protocol (LANMAR), Location Aided Routing Protocol (LAR), and Zone Routing Protocol (ZRP) we have considered for analysis. The study has been made in terms of throughput and delay. From the results of the study phase, we have taken three adept protocols namely, EDYMO, FSR and ZRP for developing our CRP.

The second phase work comprises the network setup for the CRP process. Here, the network model of IEEE 802.11 DCF based CSMA/CA is used to avoid collision, the Random Waypoint Model (RWM) uses for mobility model, and Random Early Detection (RED) was applied to achieve the dropping policy.

The third phase is incorporated for the effective development of CRP that combines the effectiveness of aforementioned protocols with EDYMO in order to provide increasing packet delivery ratio, increasing throughput, reducing jitter and reducing end-to-end delay for varying number of nodes and node densities. The Enhanced DYMO is to support the routing in an adept manner and enhances the performance of the network. EDYMO efficiently reduces the energy consumption compared with DYMO reactive protocol. It shows better performance while analyzing the results with DYMO. Moreover, CRP is driven by the fundamental trade-off between proactive dissemination, reactive and hybrid discovery of routing information with the DYMO as underlying protocol. The experimental analysis that CRP is a hybrid protocol can automatically determine the balance point between three strategies for making an informed trade-off and dynamic network measurements. The implementation has been accomplished with the network simulations with the simulation statistics described in chapter 4. The comparison between the performance of DYMO in terms of PDR, delay, jitter, throughput, Energy consumption and node densities reveals the performance augmentation of EDYMO. Hence, the resultant protocol CRP that comprises EDYMO, FSR and ZRP produces improved results with respect to the considered parameters. Moreover, CRP uses efficient mechanisms for dynamically manipulating the zone size and simultaneously performs fine-grained adaption with low overhead. Further, CRP enables applications with different demands to control the performance of the routing layer. We also described how CRP could be used to reduce packet overhead and to control delay. Our evaluation evinces that CRP attains performance that is better than each one of its associated purely reactive and purely proactive protocols across a wide range of network conditions.

Future explorations:

Evaluating various metrics

Geographical Location based proposal for CRP

Improving efficiency by controlling routing packets