The Total Network Capacity of Wireless Mesh Networks for IoT Applications

—Computing and measuring the total capacity of a data network are a remarkably difficult problem. These metrics are directly linked to the available bandwidth to each wireless internet of things (IoT) device of the network. In this paper, the authors study the performance metrics associated with capacity traffic in multi-hop wireless mesh networks (WMNs). It is dedicated to Internet access, assuming a time division multiple access (TDMA). They focus simultaneously on three key operating metrics, the total network capacity (TNC), total application network capacity (TANC), and the Average message time (AMS). They also analyze how parameters such as forward error correction (FEC) and acknowledgments (ACK) affect the overall network capacity under different operating conditions. Theoretical network capacity for WMNs, in this paper, is explored to draw attention to the number of open research issues


Introduction
WMNs are emerging as the more lucrative technology for determining wideband connectivity [1]. However, for ensuring the smooth operation and meeting the requirements of the applications, the applications require higher throughput; consequently, WMNs need a higher capacity for innovating solutions. The network capacity of a wireless mesh network (WMN) [2] implemented in the internet of things (IoT) is one of the significant criteria for quality of service.
With rapidly expanding numbers of WMNs [3] available for IoT [4] applications, developers need to understand how these networks differ concerning the use cases and expected performance.
sidering two types of networks, arbitrary, and random networks. The authors of [17] expect the total capacity of wireless ad hoc networks, mainly; it studies 802.11 MAC interactions with ad hoc forwarding, their effect on network capacity In this work, the authors analyze the network capacity of WMNs. It is assumed that each node is located in the middle of the region of the area, and it can transmit at bits per second over a shared wireless channel. Packets are sent from station to station in a multi-hop mode until they reach their single final destination (remote station). They can be buffered at intermediate stations while awaiting transmission. To simplify the tasks of the nodes, the medium access mode to be used is a TDMA, as applied in [13]. TDMA is a successful method for avoiding packet collisions. However, time slot distribution must be realized to use TDMA.
The paper presents wireless mesh networks and the effect of FEC and acknowledgments (ACKs) in IoT networks. It analyses and discusses network capacity, gives an overview of the system model, and presents contributions on some measurements and numerical results obtained for the described model.

Wireless mesh network
Wireless mesh network (WMN) is a wireless local area network (WLAN) using one or more decentralized IoT devices connection, include complete mesh topology or partial mesh topology. In a complete mesh topology, each network IoT node is connected directly to each of the other IoT nodes. In a partial mesh topology, some IoT nodes are connected to all the others, but others are only connected to those IoT nodes with which they interchange the most data. WMN is a promising wireless communication technology for various applications [18] to comply with specific requirements and uses.
In WMN, all IoT devices can access each other randomly and spontaneously, and each network IoT node can forward data to the next IoT node. The networking infrastructure is decentralized and simplified because each IoT node need only transmit as far as the next IoT node. Wireless mesh networking could permit users living in remote areas operating in rural neighborhoods to connect their networks for available IoT connectivity.
In wireless mesh-type networks, all IoT devices can access each other arbitrarily and spontaneously. WMNs can also accomplice polling or report-by-exception applications.
Each link between IoT nodes (and) could be noisy, so the information in the packets also needs to be encoded to verify and correct the errors. In this paper we analyze FEC, and ACK approaches, a class of time-diversity techniques, in the context of IoT network at the link and transport layers

Effect of FEC and acknowledgments (ACKs) in IoT network
FEC (Forward Error Correction) is an instrumental technique to understate IoT wireless channel weakness. The transmitter subjoins some redundant data into its messages. This redundancy permits the receiver to detect and correct the erroneous packet. The improvement comes at the expense of the data transfer rate concerning each IoT devices.
is the IoT device data rate, m is the modulation rate, and FEC r = FEC ratio. The above formula shows that the lower the FEC ratio, the better the ability of error correction and the lower the IoT device data rate.
ACK is a setting, which requires additional bandwidth to repeat corrupted frames. To announce that the packet was received correctly, each transmitted packet has to be acknowledged by the receiving IoT device, using the very short service packet (ACK). If ACK is not received, the IoT device will retransmit the packet depending on its set of new attempts. The authors in [19] analyze the effects of packet forwarding in Mesh Network on transmission performance.
When the IoT device has successfully received the transmitted packet, it no needs to reconstruct any dropped, lost, and missing packets at the receiver. Consequently, there is no requirement to receive ACK from the receiving IoT device, i.e., the packet is transmitted only once, and it is not re-iterated. It is noted that the acknowledgment/retransmission scheme is an integrated part of the radio protocol. It can work individually and separately of any new attempts at higher protocol levels (e.g., TCP or IoT devices application protocol).

Network Capacity
Starting from a given state of the network in which it is assumed that only TCP traffic is being processed. The network capacity is defined as the maximum TCP throughput that all IoT devices in their network could carry set. If each of them were to download as much as they can. Consider a WMN using a single radio channel shared among all routers that are assumed static. We distinguish the network capacity, defined as the maximum capacity of a link or network path to convey data from source (SCADA center node) in the network to the destination (remote station) via intermediate nodes.
In the context of the Internet of Things (IoT), a better network-wide capacity is necessary for providing a better quality of service to a more significant number of users.
In this paper, the TNC in bytes per second (includes IP packet overhead) is defined as the resulting number mentions to the maximum number in the ideally designed IoT node network. Notwithstanding, that figure can be solely completed when there is a considerable portion of the communication load among the IoT nodes (remotes) themselves.
When all control messages have ultimately to reach the SCADA center station over the same links employing the same radio-channel, any calculation of TNC loses its sense. This problem can be eliminated by adding additional channels or bypass dominant IoT devices using more radio hops. Usually, every network of devices using narrowband radios demands a capacity-aware layout.
Consider the network in Fig.1, which consists of three ( 3) n = wireless routers coverages areas. Router 1 sends its packets to router 2, and, thus, the link 1,2 e has to be able to forward traffic. Router 2 starts forwarding its packets to router 3 via the link 2,3 e .Router 1 can exchange data with Router 2. Router 3 can also exchange information with Router 2. However, Router 1 and 3 cannot communicate with each other, but their signals can interfere at Router 2. Digital radio has developed ways in which more than one information conversation of the data center can be accommodated (multiplexed) inside the same physical radio channel. For achieving this, the TDMA method is used. It must be noted that in that case, TNC is dependent on the transmission range of each IoT node. So that the collisions cannot be excluded entirely; hence, a collision-solving system must be a complete part of the protocol in the radio channel.
Whenever the traffic increases over a specific limit, the number of collisions augments considerably, decreasing the instant network capacity well below normal situation.

System Model
In this paper, the fixed infrastructure of the WMN is represented by an acyclic directed graph   SCADA center sends control information to all IoT nodes. SCADA application typically uses a specific address for such control information. Along the paths of the highway system, one node is chosen per edge, which acts as a relay for the following node. The receiving IoT node 2 (directly connected to the IoT node 1 (one radio hop), converts such message to a customized IP broadcast and sends it to IoT node 3 over one IoT node 2 (two radio hops) respectively to all node's units within the network. Each link in the chain is constrained to send only when the other links in its vicinity are inactive. An objective to expand the system in WMNs is to stretch the coverage range of wireless IoT networks without resorting to the sacrifice of the channel capacity. Another aim is to provide non-line-of-sight (NLOS) connectivity among the IoT devices without direct line-of-sight (LOS) links. To meet these needs, the multi-hop paths are required [20]. Wireless mesh networking has become ideal for IoT, for smart car parking mobile application [21] , and it can be incorporated with different network capacities such as IEEE 802.11, 802. 16, 802.15, etc. Assume that each IoT node receives traffic to be forwarded to the RS and that each IoT node can only receive packets from its immediate neighbors. The traffic that has to be forwarded by each link is computed. It is clear from the picture that nodes closer to the RS have to forward more traffic than nodes farther. The RS is connected to the SCADA center through n IoT nodes, with the last one is considered as a remote station.
The packet must be completely received before being forwarded to the next IoT node. Consider D as a total delay for sending N packets crossing n nodes over M links (pipelining delay) that is given in the following form [22]: D = Propagation delay + Transmission Delay + Pipelining delay Where TP D = transmission time of the packet. Each packet contains L bits of data and a header of size H bits with total packet size LH + bits. The transmission of a message (control information) of S bits needs / SL packets; therefore, the time to transmit the message over M links is where R is the data rate. In this network, there is one source node (SCADA center) and one destination node (remote station). The source node does not have any incoming edges, and a destination node does not have any outgoing edges. SCADA center has a message for monitoring remote station system as in [19].
Consider that the total network capacity TNC C is the relationship between the useful data trafficked, without any headers (IP, TCP, UDP,…), trafficked by all IoT devices and the total time takes them, so: where data ACK TT + is the total time it takes for a node i to download a full packet payload L .
Rewriting the equation (3): Consider that i C is the capacity that each IoT node obtained. Consider the case where only the one connected to the network which is equivalent to: This last result represents the total network capacity assuming that all connected users are reporting as much as possible.
Total network capacity in bytes per second (includes IP packet overhead) is the resulting number refers to the maximum number in the optimally designed IoT devices network. While total application network capacity is total network capacity in bytes per second, but no IP packet overhead is included.
The following Table 1 lists the most critical measured parameter selected for the used topology. bles until it reaches the max W limit, the moment from which W stops growing. Then the plot discards a certain number of retransmissions. Also, it is assumed that there are only one SCADA center and one remote station in the network. Furthermore, every IoT node is capable of creating, receiving, or transmitting data over a communications channel. The topology illustrated in Fig.2 is respected for the simulation with 9 n = .

Results
This section describes the results of the experiments. The presented measurements show how Total application network capacity (TANC) varies depending on the sever-al Message sizes according to the modulation type (QAM). Various measurements are effectuated as a function of different values of AMS, without FEC/ACK or with FEC/ACK, respectively. Each device may support up to 1500 bytes of RF payload. In this case, the payload bitrate curve shape, depending upon the number of hops, AMS, and ACK/FEC, is illustrated in the figures that follow.
The Fig. 3 shows the results of TANC in bytes per second, (but no IP packet overhead is included) versus AMS with modulation type QAM. According to the results shown in Fig. 3, it can be observed that the TANC increases considerably for the AMS values increase until 500 bytes, then increases slowly from 500 bytes to 1500 bytes. Additionally, one can see that the number of hops can also promote a significant increase in TANC.
The results illustrated in Fig.4 show the comparison between Total IP network capacity and Total application network capacity without any correction. It is noted that the total IP network capacity is high relative to the capacity total application network capacity, especially between zero and 600 bytes. Fig.5 shows the effect of protocols (on network capacity), which can detect and recover the lost packets, either by data redundancy FEC=On or by retransmission ACK=ON.  The results in Fig.5 show that FEC, as well as ACK, have a significant effect on the network capacity. Additionally, our results reveal that packet loss recovery techniques promote a further decrease in network capacity.
The scenario is considered in Fig.2: direct transmission between the serial neighbor devices using nine routers. Fig. 6 shows the comparison of Total IP network capacity results obtained for different configurations of FEC and ACK by considering different fixed message sizes. The results in Fig. 6 show that the using different average message sizes (bytes) in wireless networks, without FEC/ACK or with FEC/ACK respectively, has a significant effect on the Total IP network capacity. Additionally, the results reveal that the FEC decreases the total IP network capacity channel more than ACK.
In the same way, Fig.7 shows the comparison of total application network capacity results obtained for different configurations of FEC and ACK. The introduction of ACK increases the Total IP network capacity and total application network capacity more than the presence of FEC. Also, without any correction (FEC=ACK=0) only increases capacity but does not improve robustness to link failures, network congestion, and other problems of the wireless network communication.
Moreover, it can be noted that the results illustrate that the IP packet overhead promotes a further decrease in Total application network capacity. The use of FEC and ACK has the disadvantage of low network capacity.
To ensure the conformity of the experimental results, it convenient to compare them with the simulation result, as illustrated in Fig.8.   Fig. 8. Comparison between the simulation and measurement results of Total application network capacity vs. AMS using FEC and ACK.
In the case where ACK and FEC have become operational, the curve illustrated in Fig.8 shows that the experimental and simulation results are almost the same

Conclusion
In this paper, the authors focus on the analysis of TNC over a multi-hop, wireless network technologies, designed for the IoT, where all links share the same physical channel. Our analysis shows that it is often difficult to simultaneously improve both the TNC and the Error Correction Code (ECC). Number hops and average message size adversely affects the TNC. They analyze the case in which the noise in a given link is unrelated to the signals traveling over other links. In that case, the authors show, the problems of error correction and capacity network can be separated without limiting the capacity of the network as a whole.
Furthermore, the analysis of the results shows that the Error-correcting codes, such as FEC and ARQ, have two opposite effects on the efficiency of TNC in IoT networks with different lengths. Although ECC dramatically increases the reliability and performance of the overall system, the ECC also makes signals more robust; in contrast, it reduces the total network spare capacity. It also shows that there is a straightforward connection between TNC characteristics and observed packet size FEC and ACK. The more hops per path, the less overlap, and consequently, more capacity left for simultaneous transmissions from different IoT devices. That is the reason for higher capacity with more hops in the network, and it is the subject of the future work using a new performance algorithm.