Many sensor nodes exist in a WSN that are tiny and cheap in cost which can be sensed cooperatively in a physical phenomenon. The results that exist from the research made the products which provide the possibility of building the WSN effectively in many of the applications. As the traffic features that are present inside WSN are understood better then WSN could be designed much more effectively. Consider an example as the routing protocols that are better and the strategy of sensor deployment is designed as the burden of traffic present on the sensors present in the WSN could be better understood. Considering the traffic features the abnormal and the normal traffic can be kept apart to make the management security and better fault to be applicable.
In the literature the investigation was performed on many kinds of traditional network traffic dynamics for both the wireless and wired networks in particular and also due to the WSN specialty the investigation was moved on. The tractable and accurate analytical models were constructed in the sensor traffic that provides the future work basis for a design of the network, the security provided and its optimization. The research was performed on the analysis and the modeling of traffic (John Paul Walters, 2006).
Data Traffic Arrival Process:
The scenarios of WSN are different in the data traffic dynamics as the analysis and the modeling of traffic are dependent on the application. Periodic data generation or Event-driven is the applications that are categorized from a WSN application.
In the periodic data generation scenario, the data traffic arrival process is modeled by using the constant bit rate (CBR) as the bit rate remains constant. If the bit rate is treated as a variable then the poison process comes into action for modeling the traffic data as it is not bursts.
In the Event-driven scenario, which includes the target tracking and target detection as the burst traffic may rise through any of the sensing area corner as an event gets detected by the sensors that are located locally. Also the poison process is being used.
The burst phenomenon of the data traffic is modeled by using the ON/OFF model in an event-driven scenario. Then the distributions present in the WSN scenario are made to follow the generalized Pareto Distribution. A different scenario that includes the mobile sensor network was introduced where the new dynamics is being introduced by the mobility of node to the traffic of network (Ankit Kesharwani, 2010).
Sequence Relations among General kinds of Packets:
In some packet kinds the existence of the sequence relation can be seen. Let us consider an example as the arrival of a Routing Reply message appears only after the Routing Request message occurs which is specified according to a routing protocol. A finite state machine (FSM) is proposed by the authors, which can specify the routing behavior correctly for the ad hoc on demand distance vector (AODV) routing. FSM can depict the sequence relations that are present in between different kind of messages.
The routing operations of the AODV and FSM are specified clearly the obtaining of the sequence relations among the different kinds of routing packets is easier and are then manually abstracted within the FSM. Considering the protocol specifications many kinds of packets sequence relations can be specified and also the on-line training can be used. The dissertation explains that firstly arriving packets are classified depending on their attributes such as addresses, packet type. Then the mapping the sequence of arriving packets is performed with a character string that is infinite.
The online learning is performed by considering the unique character string extracted from the scanning process performed based on the window. Then the normal traffic is built based on the learned packet sequence for the node present in a static WSN. The dynamic WSN where the nodes are mobile the profile of traffic is learnt which evolves quickly depending on the over time, thus making it less meaningful (Axel Jantsch, 2008).
Data Traffic Load Distribution:
The load of data traffic is distributed uneven over the nodes in a WSN. Let us consider an example such as there is sensors present one hop away from the sink and relay upon the entire traffic data. The network's functionality and the lifetime of it can be degraded due to the data traffic load distribution that is imbalanced. These efforts are devoted in the data traffic load characterization. As a node becomes nearer to the sink the data traffic load increases and reduction is seen in the traffic load data. In the dense planar WSNs the data traffic load is considered as the function for the sink distance and in linear WSN a node traffic load may increase as it moves nearer to the sink.
In a symmetric sensor network the nodes are distributed evenly in the field of sensing. The data traffic load expected for a node is inversely proportion to the mean routing hop length and is directly proportional to the radius of the network and also it is independent of the node density. The energy consumption is related by the distribution of data load traffic which makes an impact upon the WSNs performance. The results of the research made shows that the data load distributed can optimize the WSN performance which is related to the consumption of energy in the network (Adel Gaafar A.Elrahim, 2010).
Network Optimization for WSNs:
In WSNs many problems related to optimization are to be resolved still. The problems may be related to the control flow, congestion control, medium access control, rate control, topology control and power control.
Energy Efficient Routing Design:
As the critical energy consumption is dominated by the communication and routes the design that considers the sensor network core design. In the research many prior algorithms are proposed. For the problems related to the flow of routing in the network a shortest path is considered. The minimum hop routing present in the simple translation is considered most AODV can be considered as an example for it. The main fundamental of the sensor network is that the routing algorithms attempt to minimize the valuable resource utilization.
Each data bit's transmission energy is attempted to be reduced by the minimum total energy and the minimum total transmission power routing. The path length is considered as the energy sum extended per data bit while the transmission is performed on the forwarding path on each and every link. Improving the packet ratio that is transmitted for the network which consumed the energy itself is not a better measure for the network efficiency. The routing of the maximum residual energy path is performed based on algorithm having similar conditions that attempts in first node death postponement making the remained path energy maximization (JOHN STANKOVIC, 2004).
The first battery expires faster as a network is maximized when the source set nodes and the sink nodes of an ad-hoc network are chosen. The minimum expenditure of total energy is not that desirable as the nodes have the relay burden in excessive amounts and these kinds of nodes would expire soon. Due to which the connection may be lost among the other nodes. To overcome this kind of problem the routes are to be chosen having an ultimate objective as the time is maximized till the first battery gets expired. The linear programming problem is reduced by the energy efficient routing problem and is described as follows:
Max lifetime
s.t. 1. Energy constraint
2. Flow conservation constraint
The lifetime represents the operational time taken by the network until the first battery gets expired. The energy constraint is described as the expended energy for the communication, sensing and other operations which cannot surpass the reserves of the initial energy surpass. The flow conservation constraint is used to specify outgoing data flow number from each of the node that is made equal to the summation of the number of data flows incoming for a node and the data flows number that originate at the node. The optimal routing strategy corresponds to the data flows which can maximize lifetime (Rob Hoes, 2008).
The new concept called application-tolerable network which works at the run time to collect the information depending on the network lifetime which orients the information for the comparison. In the definition of new network lifetime the permission of death is given to the nodes while the network operational lifetime. Due to which the transmission of data among the two nodes may become unstable. The optimization problem formulated here is described as follows:
Max lifetime or Total Information Collected statements
1. Energy constraint
2. Flow conservation constraint
3. Application-Dependent Requirement on Network Information-Collecting Ability
The lifetime is defined as the network operational time till the ability to collect the information about the network falls down the requirement of the application-development. The total information collected is defined as the performance metric that is much suitable to those WSNs which collect the information and represents the entire information which is collected by the network in its lifetime. The energy constraint is described as the expended energy for the communication, sensing and other operations which cannot surpass the reserves of the initial energy surpass. The flow conservation constraint is used to specify outgoing data flow number from each of the node that is made equal to the summation of the number of data flows incoming for a node and the data flows number that originate at the node (David Richard Raymond, 2008).
The Application-Dependent Requirement on Network Information-Collecting Ability is used to specify the worst information collected by the network that is tolerable by the application.
Energy-Efficient MAC Design:
In most of the sensor nodes the energy consumed is much higher by the MAC protocols as they influence the transceiver utilization directly. The latency is minimized and the fairness is provided by the MAC protocols. A minimum amount of energy is considered in WSN as the energy expended is in the idle mode. Much energy is wasted by the idle listening as the receiver node does not when the node turns out as a receiver when a message is sent by the other neighbor node so the radio is to be kept in receiver mode all the time.
Many solutions are given addressing the problem about the wastage of energy because of the idle listening. A duty cycle involves as it allows periodically every node to sleep and based on the energy preserving of TDMA-based protocols naturally. The slots of TDMA allotting are treated as a big problem as it requires the coordination. So an extra radio is required to save the energy (Yi Shang, 2007).
In-Network Processing:
An enormous amount of data is being collected by WSN upon the space and time as the transmitted data from a sensor node to the central processing location is raw. The raw data collection is not an ultimate goal rather the parameters or functions of interest estimation is much important. The Distributed In-network Processing is used in eliminating the transmission of raw data need to the central point which would significantly reduce the energy resources and communication consumed.
Many In-network Processing approaches exist which are combined along with the routing algorithms as the ultimate objective of it is in computing all the measurements related to the average or the quadratic cost functions and the objective parameter that can be passed through. Many cycles may be required to achieve the objective sometimes by a network. A LEACH protocol is considered as an elegant solution for the problem of data aggregation in the clustering formed by a self-organized manner for the fusion of data before it gets transmitted to the sink or base station (Wadhai V, 2010).
Load Balancing:
In WSNs when many sensor nodes send the information to one or more number of data sinks by the multi-hop transmission that dominates the pattern of communication. An imbalance rises drastically in the distribution of traffic load in a network as the nodes move towards the sink. As the communication dominates the consumption of energy of a sensor node that provides the energy sources limitedly the traffic load distribution is imbalanced is much harmful to the network.
The problem related to harm that is resulted by the unevenly distributed traffic load is countered by the researchers by turning their attention towards the load balancing problem. Due to the traffic load distribution imbalance the nodes may die in a network resulting in the degradation of the performance of network. To overcome this problem many routing algorithms are introduced such that the path capacity measurement is proposed.
Resource Allocation:
To counter the resulted harm by the traffic load distribution imbalance a fair resource allocation approach is used an alternative. In this allocation the objects are allocated with the resources such as bandwidth, nodes and energy accordingly to the object's workload. The performance upper bounds in WSN increases linearly along with the reservation of energy within a bottleneck zone identified assigning the energy resources available which can lighten the bottleneck effectively.
The radio range is made adjustable so that the consumption of energy is saved on the path that is routing. The shorter relaying ranges present for the nodes that relay at the sink closely would add the imbalance to the already imbalanced traffic load distribution. As the traffic load indicates the energy consumption rate the fair energy proposed would allocate the scheme which maximizes lifetime of a network and further equalizing the lifetime expected by each and every individual sensor node (Shensheng Tang, 2006).
Dominating Traffic Pattern
The wireless sensor network missions are used for gathering the data from different kinds of surroundings. These environments are used to report the base station nodes which are corresponding to the distributed nodes in the network. Various traffic flows will represent the communication designing Parts. This kind of sensor nodes is essential for configuring the distributed network. The traffic flows are started from sources to destination. There are various distributed nodes. These nodes are existed for network protocols.
Specific network protocols in the networks are used for several concrete applications. There are various other kinds of network protocols. Traffic routing protocols include link layer message protocols for specifying the sensor nodes. There is a kind of application specific traffic for displaying the message as "hello message". (Qinghua Wang3. 2010)
packet sequence modeling
Some relations would be existed in different kinds of packets in the corresponding network. Routing requests will be preferred by conventional routing protocols. This kind of routing protocol use a machine called finite state machine in order to specify the different kinds of packets with information. This machine will indirectly refer to the routing behavior. FSM model can be used for representing the routing protocol. Sequence relationships between nodes can be used for specifying the various protocol requirements. Sequential relations can be followed by the concept of automatic learning. (Adel Gaafar A.Elrahim1. 2010)
Packet classification
In order to know about the sequential relations, first step is to initialize the general packets. Several packets in the network can be considered as unmanageable packets in case of size. There the size is considered as the learned sequential relations that are specifically based on classification of packets. The packet classification is indirectly reflecting the unique and conventional behavior of a node in a network with specific interest. Protocols specifications are specified in order maintain the routing reply.
Consider the AODV protocol, this protocol is specified as the routing reply can be initialized. In specific wireless sensor networks, the links between various nodes can be established and learned. There is a feature that can be called as "packet type" for naturally classifying the different kinds of packets. A cluster node is essential for sensor network in order to sensing the neighborhood nodes in the network. There are various pairs of packet type. These pairs can easily control the packet classifications. General address spaces can be some times specified as neighborhood nodes for specifying the hop distance. Neighbor represents the nodes in the network within the specified and determined hop length. Here local shows the already known nodes between the corresponding source and destination nodes. In the concept of packet sequence learning, no other node can be easily divided into the non local variables. This time the packet sequence collaborations are much required and get inherited. Here the node observation is specifically maintained and source and destination nodes will be signaled by the non local variables in the network. (Dr Helonde J B. 2010)
Packet translation
Fig: packet arrival events translation
Here the translation is done for converting the arrival events to characters. Sometimes several divided packets can be effectively mapped in order to set the ASCII characters. The above diagram will represent the character sets for arrival of the characters of large string. (Laurent Mounier. 2007)
Pattern extraction
In order to learn and understand the packet arrivals, the packets must be extracted. Here extraction will be the most significant step after classifying them. The classified packets are specifically turned into the extraction step. An extraction algorithm is used for successfully complete the extraction process. Actually, this algorithm is first used for the concept of intrusion detection. An arriving sequence can be utilized and used properly by the packed events of the network. There are unique sub sequences that are have been found by several researchers for constructing the pattern in a network.
The pattern construction s could be done for initializing the data transmission. If the hop length can be identified as k, then packed events can be scanned within the hop length between the source and destination nodes of a network. If the hop length can be identified then it is ensured that the process of pattern extraction can be completed. Below example depicts the pattern database extraction. Consider a variable called k=4 and the sample sequence will be AABBDCC, then after the completion of pattern extraction, the pattern will look like as follows. The patterns are AABB, ABBD, BBDC, and BDCC. This example is directly represents the process of pattern extraction. Here wireless sensor networks are significantly deployed for changing the various action sequences in order to evolve the packet arriving sequences that area hold for particular period of time. In order to maintain the packet arrival sequences, the pattern database is required. (Delphine Christin. 2010)
Overview of packet sequence modeling
This section will clearly learn the packet sequences with the arrival of them. The following packet translations could observe the recent patterns for the process of packet translations. Using these approaches of packet extraction and packet translation, the probability of new pattern sequences can be minimized. So the process of learning sequences could be minimized in order to observe the sensor nodes of the network. (E. Egea-López. 2005)
traffic load distribution
Here the traffic load distribution can be performed and operated in the networks called dense sensor networks. Wireless sensor networks are clearly guiding the energy allocation in wireless sensor networks. In order to minimize the energy allocation, the concepts of load distribution can be understandable. The networks can be wireless networks and the specific nodes of those corresponding networks are deployed and spread on the disk area.
Here the features of wireless networks are to be found as inverse proportion to the network intensity corresponding to hop length. The network scale effects can be deployed for investigating the hop length. The network intensity outcomes are specifically represented by the various simulation experiments. (Yimin Chen.2009)
Network scenario
The dense wireless network can be considered by the various approaches of deployed disk spaces. All the nodes in corresponding dense network can be deployed and distributed over the disk space. Here the routing paths between the nodes can be specified as sensor nodes for the communication. Various routing paths are maintained in case of traversals. This kind of traversal can be literally proportional to the distance that is between source and destination. Here the hop length can be traversed by the routing paths in the wireless network. This kind of networks will directly accept several routing protocols and the touting algorithms.
If the shortest routing path algorithm is considered, the main objective of this algorithm is to find the path between the various nodes of a wireless network. It considers the Euclidian distances between the nodes of a network. These kinds of communication paths are specifically move close to the higher communication range. Here several scenarios are to be considered by the progress based algorithms. This algorithm is considered as the routing algorithm for the nodes in wireless sensor networks.
The closed communication ranges in a wireless network can be clearly explains the arbitrary algorithms of dimensional network. In different kinds of optima routing policies, the no of hop lengths can be described and verified. Many researches are takes placed for proposing various routing algorithms for obtaining the routing algorithms. Some no of hops will effectively increase the dimensions of MANETS in wireless networks.
There are other source algorithms which are used to express the hop length based on the Euclidian spaces which are specifically considered as the independent routing paths from the nodes such as source and destination. Here the difference between mean length and hop length can be ignored. 'h' represents the mean hop length in the wireless sensor networks. All the nodes in this network will contain the same sensing rate which is average. This could be denoted by Å and various investigations have been distributed over the networks. If there is no node, even the action stability is performed, and then the energy remains constant in the corresponding network. (Yi Qian. 2007)
Traffic load Analysis
Here the amount of traffic load can be determined in a certain period of time. The complete amounts of packets are to be handled by the wireless network. These packets are handled by packets of the wireless network regarding sensing nodes. Total packets include other packets of traffic load. This should be expected by the original position of the nodes in a certain period of time. Here the number can be determined by the given nodes in the wireless network. Sometimes this conclusion can be considered as the straight forward issue for child nodes.
If a node b is as child node for the node 'a'. Then the node 'b' has to depend on the parent node a. this could be done in case of transmission of packets.
Fig: sensor network on disk space
Fig: infinitesimal area in proximity region
The above figures represent the deployment of nodes in the disk space and the child nodes in the proximity region respectively. In above the node's' has a distance rs with the dense network. This node is lies between the nodes that are in proximity region of the wireless sensor network. Here this node is closely deployed to the next node of the wireless network. For each infinitesimal dσ in ss can be defined as dσ = r dr dθ in case of polar coordinates. When the nodes can be sinking, the region can be again defined as follows. dσ' = (r+h) dr dθ. Here the nodes S's can be considered as the immediate child nodes of the proximity region.
Here the assumption can be done based on the child nodes of the network. The denseness assumption of the child node s's can be defined as following.
Σ dσ' = ∫∫Ss (r+h) dr dθ ≈ (1+h / rs ) ∫∫Ss r dr dθ = (1+h / rs) Σ dσ
The above equation represents the determination of infinitesimal region of immediate nodes in the wireless sensor networks. But this can be applied in distributed networks only. The total no of child nodes for the node s can be specified as follows.
This is done because the child nodes can be periodically creates the packets with an average transmission rates per a certain period of time. The traffic load in distributed networks can be defined as the following.
Here the node s will only represent the sensing region. This equation can be literally generalized as follows in case of sinking.
Here R denotes the finite radius in the distributed network and m can be defined as follows. m= [R-r]/h. it is much easier to determine the network radius R and the hop length h. these are the two parameters which describe the influences of network relationships. The traffic load expressions can be clearly observed and defined as below.
Here r' = r/R which is the unique space from sink node. Then,
This will specify the traffic load from the knowledge of h/R. this can be usually gained by the below expression.
Above expressions can be evaluated by the individual sensing nodes for distributed network. R represents the network radius. The small radius r' ε (0,1] here r' → 0. This will be the bottleneck situation in the corresponding distributed networks. (THEODORE ZAHARIADIS. 2009)
Simulation outcomes
The dense network is initially simulated and maintained with 20000 nodes in the distributed network. The average distance of in the network grids is 100 meters. This can be efficiently proved through the higher communication ranges. Additionally the sensor nodes regarding the network are deployed over the disk space. In this way the simulation has been done. Here the traffic loads can be estimated by determining the inner and grid radius of the network. Here the inner radius is r- ∆r/2 and the outer radius can be r+ ∆r/2. Here the grid radius can be defined as ∆r = √âˆR2 /n. it will indirectly shows the traffic load which is considered as the theoretical load.
Graph: distribution of traffic load
The theoretical traffic results would be expected for the load distribution on the dense sensor networks. Here the density of the networks would be considered with a poor density. Other practical experiment could be performed in order to show the traffic load on the dense networks. This corresponding result can be represented as follows.
Graph: traffic load when n=250 and R=50m
Graph: comparison between theoretical load and practical load
Overview of traffic load distribution
Special communication pattern can be used in the distributed networks with full traffic load. Here traffic load distribution can be determined by using the node deployment in the distributed network. The sinking node can be taken as the root for determining the traffic load using inner and outer radius of the distributed network. (Ana-B García. 2008)
