Printer Friendly

Strengths and weaknesses of prominent data dissemination techniques in Wireless Sensor Networks.

1. Introduction

Wireless sensor networks can be applied in several application domains. In fact, a wide range of real-world deployments have been observed in the last few years [1-9]. This kind of network is constructed with a large number of tiny and smart sensor nodes deployed in an ad-hoc manner over an Area of Interest (AoI) for collecting the expected information [10-11], [96-97]. These nodes are expected to be inexpensive and can be deployed in a large number in harsh environments, which implies that the sensors are typically operating unattended without any human intervention for most of the network's lifetime. The communications from and to the network are performed using wireless technologies. Each sensor node has its control area to monitor the surrounding environment by perception equipment, optical equipment, chemical analysis equipment, and electromagnetic equipment. Some special functions can also be achieved by setting some functional equipment [12].

Among various roles and objectives of WSN, the most crucial objective is data dissemination [13-14], [98] which is also one of the key problems faced by the sensor nodes. In this environment, the network supervisor (or, administrator) may need to interrogate the sensors by spreading his interests over the whole network, whereas a sensor node needs to notify the supervisor when interested event occurs. During data dissemination processes, sensor nodes communicate with each other to deliver the sensed data to the supervisor via the sink node. Each node in the network acts as a router and may sense the data directly or receive it through other intermediate nodes.

One of the major barriers of a Wireless Sensor Network is that the sensor nodes have limited transmission range. Also their processing and storage capabilities as well as their energy resources are limited [15]. Hence, the limitations of resources are often noted as the key challenge to tackle for designing any operational protocol. Data dissemination within WSN is not an exception to this. In practice, data dissemination protocol for WSNs is responsible for delivering the sensed data using a valid path between source and destination node and has to ensure reliable multi-hop communications. Because of the relentless efforts of hundreds of researchers, several data dissemination protocols have been proposed for wireless sensor networks by this time. Considering all the inherent challenges in WSN as noted above, it is an interesting issue to investigate how the data disseminations are modeled for such networks. This is the core intent of this paper to analyze various aspects of the design methodologies of data dissemination of the most significant protocols. We describe the achievements so far in this area and highlight the relative strengths and weaknesses of the data dissemination models of various protocols for WSNs.

The rest of the paper is organized as follows: Following the Introduction, Section 2 describes the WSN data dissemination mechanism, notes the previous related surveys and provides new taxonomy of WSN data dissemination. Based on the classification of data dissemination protocols presented in Section 2, Section 3 and Section 4 present an overview of the major data dissemination strategies with static and mobile sinks respectively. The issues of single and multiple sinks are investigated in detail in separate subsequent sections. Section 5 resumes and compares the studied data dissemination protocols. Finally, Section 6 presents concluding remarks with directions on open research issues.

2. Data Dissemination Protocols

To design a data dissemination protocol for wireless sensor networks, it has to consider several parameters related to this environment [10]. Limited energy resources of sensor nodes, quality of the wireless channel, packet loss and latency constitute the main important issues that are needed to be considered. Energy consumption effectiveness represents the most significant performance metric which influences directly on the lifetime of network. This is why, several data dissemination protocols have ignored other performance metrics such as the data transmission time, latency, and have put more emphasis on energy consumption [16-27], [51].The goal of a data dissemination protocol is to find and maintain a valid path towards sink or base station by which data forwarding process would consume minimum of energy.

Several data dissemination strategies have been proposed for wireless sensor networks. Their principal ideas mainly are related to the class to which they belong. In literature, these approaches have been classified [10], [28-30], according to the network architecture, the initiator of communication, the path establishment, and so on. In [14], the authors highlight the special features of sensor data collection in WSNs, by comparing with both wired sensor data collection network and other WSN applications. The authors describe a basic taxonomy and propose to break down the networked wireless sensor data collection into three major stages: namely, the deployment stage, the control message dissemination stage, and the data delivery stage. A literature survey on data collection in WSNs with mobile elements has been presented in [31]. In this work, the data collection issue has been studied through three separate phases. Discovery phase allows nodes to detect the presence of the mobile elements; data transfer phase defines the communication process between a mobile element (ME) and its one-hop neighbors. In the last phase of routing to mobile elements, the authors present and discuss some data dissemination protocols with mobile elements into flat routing and proxy-based routing classes.

Our classification in this work is mainly based on the number of sink(s) (single or multiple) and their nature (static or mobile). Some protocols require that the sink node has to be static and sensor nodes cannot achieve their requirements without this assumption. Other protocols support the mobile sink concept and try to exploit this possibility to provide a good performance. Moreover, in these kinds of protocols, some use more than one sink which requires additional management and coordination operations.

Considering these key parameters, we suggest classifying the existing data dissemination protocols according to the taxonomy shown in Figure 1. Two great classes can be found; (i) Static sink data dissemination protocols and (ii) Mobile sink data dissemination protocols. Each class is again divided into two subclasses related to the number of sink(s).

3. Static Sink Data Dissemination Protocols

Sink node in a WSN is the most important entity. The collected sensor readings from sensory field have to be disseminated to a predefined sink for analysis and processing. The data dissemination strategies with single static sink try to prolong the network lifetime by unbalancing the traffic load using multiple data dissemination paths. Nevertheless, sensor nodes near the sink still deplete their energy faster than that of other nodes due to their heavy overhead of relaying messages. This uneven energy depletion phenomenon causes degraded network performance and limits network lifetime. If all sensors around the sink consume their energy, the sink will be isolated from the network, and then the entire network would fail. Using multiple static sinks can significantly improve the network performance in terms of latency and energy consumption. Having multiple sinks in the network reduces the distance between sensor nodes and a sink, thus can improve both energy consumption and latency [32-37]. A WSN with multiple sinks can be regarded as set of subnetworks, each of which is composed of a single sink. The number and the locations of sinks should be thoroughly studied as they could directly affect the network lifetime [38-41].

3.1 Single Static Sink Data Dissemination Protocols

3.1.1 Low Energy Adaptive Clustering Hierarchy (LEACH)

LEACH [42] is the first and most commonly known energy-efficient hierarchical clustering algorithm for wireless sensor networks. LEACH is a cluster-based protocol, which includes distributed cluster formation (Figure 2). LEACH randomly selects a few sensor nodes as cluster-heads and rotates this role to evenly distribute the energy load among the sensors in the network. In LEACH, the cluster-head node aggregates the sensed data arriving from nodes that belong to its cluster, and sends an aggregated data to the base station in order to reduce the amount of information that must be transmitted to the base station.

LEACH process starts with the entire nodes organizing themselves through the clustering algorithm to form a cluster where one node will be elected as a head node or cluster head. Energy will be depleted more if the cluster head is fixed into one node, thus LEACH has the ability to rotate the cluster head among the nodes in the local cluster. LEACH protocol uses aggregation method to gather all information from the sensor nodes in the local cluster where the cluster head will collect the information for sending to the base station. LEACH protocol design can be divided into three different angles:

--Nodes clustering,

--Data gathering for aggregation,

--Cluster head rotation.

In node clustering setup, each sensor node will select in which cluster it belongs based on the distance between the node and cluster head. The process needs the cluster head to broadcast a message to all its neighboring nodes which alerts them that it is a cluster head. After receiving all the messages from the nodes that would like to be included in the cluster and based on the number of nodes in the cluster, the cluster-head node creates a TDMA (Time Division Multiple Access) schedule and assigns each node a time slot to transmit its data. This schedule is broadcast to all the nodes in the cluster. This schedule permits the nodes to turn off their transmitters if there is no activity in the cluster. Hence, this mechanism reduces inter-cluster collision and energy consumption.

Data fusion or aggregation is to compact the data in cluster head for sending to the base station when all the information is being gathered from the sensor nodes in local cluster. The most important part in LEACH cluster head is the way it handles the rotation among the nodes for cluster head elects. Nodes have to elect the head by themselves based on the energy remaining in the nodes and some given probability calculated individually by each node.

Although LEACH is able to increase the network lifetime, there are still a number of issues about the assumptions used in this protocol. LEACH assumes that all nodes can transmit with enough power to reach the base station (BS) if needed and that each node has computational power to support different MAC (Medium Access Control) protocols. Therefore, it is not applicable to networks deployed in large regions. It also assumes that nodes always have data to send, and nodes located close to each other have correlated data. It is not obvious how the number of the predetermined cluster-head is going to be uniformly distributed throughout the network. Therefore, there is the possibility that the elected cluster heads will be concentrated in one part of the network. Hence, some nodes will not have any cluster-head. Furthermore, the idea of dynamic clustering brings an extra overhead, which may increase the energy consumption.

In [27], a centralized cluster formation version of LEACH has been proposed, where the base station organizes and controls the network. This protocol provides a centralized cluster formation, local processing for aggregation of sensed data and the rotation of cluster heads for every round. These activities are aimed at achieving uniform energy consumption among sensor nodes and maximizing network lifetime. Since, the base station does not have energy constraint, centralized cluster formation methods can be attractive alternatives. In this protocol [27], the cluster formation is formulated as a p-median problem [43], which is one of the well-known facility location problems (Just to clarify a bit here, the p-median problem can be stated very simply, like this: given a set of customers with known amounts of demand, a set of candidate locations for warehouses, and the distance between each pair of customer-warehouse, choose p warehouses to open that minimize the demand-weighted distance of serving all customers from those p warehouses [44]). This algorithm produces better clusters by dispersing the cluster head nodes throughout the network.

3.1.2 Threshold-sensitive Energy Efficient Protocols (TEEN and APTEEN)

Threshold-sensitive Energy Efficient sensor Network protocol (TEEN) [45], and (Adaptive Periodic Threshold-sensitive Energy Efficient sensor Network protocol (APTEEN) [46] are two hierarchical dissemination protocols proposed for real-time application. In TEEN, the authors assume that base station and sensor nodes have same initial energy and base station can communicate directly with each sensor nodes in the network. In this protocol, the sensor nodes sense their environment continuously, but the transmission is done less frequently. As shown in Figure 3, the network consists of three communication levels: simple nodes communicate directly with their cluster head and constitute the first communication level; then, cluster head can communicate directly with the base station, or via another intermediate cluster head.

Cluster head sends two parameters to its neighbors, hardware threshold and software threshold--the hardware threshold being the minimum value of an attribute permitting a sensor node to power-on its transmitter and transmit to its cluster head. It permits reducing the number of transmissions by allowing a sensor node to transmit its data if the sensed attribute is in the range of interest. The software threshold reduces the number of transmissions which could have differently occurred when there is little or no change of the sensed attribute.

Based on the two thresholds, data transmission can be controlled and reduced which decreases the energy consumption and improves the effectiveness and usefulness of the receiving data. However, in TEEN, a sensor node may waste its time slot if it does not have any data to transmit. Also, cluster head has to keep its transmitter power "on" to receive data from its members; thus, more energy would be consumed.

APTEEN [46] is an extended version of TEEN. It uses the same network model as of TEEN (Figure 3). APTEEN supports both periodic data collection and time-critical situations.

After creating the clusters and selection of the cluster heads by the base station in each round, the cluster head sends to its member nodes some parameters concerning the physical parameters; the hard threshold and soft threshold values, the time slot to each node using TDMA and the maximum time period between two successive reports sent by a node. In APTEEN, the cluster head aggregates all the data received from its member nodes and sends it to its higher level cluster head or to the base station which allows reducing the network overhead and the overall energy consumption. Moreover, APTEEN is suitable in both proactive and reactive applications. However, this protocol generates an additional cost and more overhead to organize the sensor nodes in complex multiple levels of clusters.

3.1.3 Power Efficient Gathering in Sensor Information Systems (PEGASIS)

In LEACH [42], each node sends the collected data to its cluster-head, which unnecessarily implies a transmission of a great mass of information and it consumes much energy especially if these data are redundant. PEGASIS [47] is proposed to solve this problem and improve the LEACH protocol. In this protocol, each node communicates only with its nearest neighbor and only one node is selected to transmit to the sink which creates a chain communication shape. The chain is constructed in greedy way by assuming that all nodes have global knowledge of the network. The chain construction is started by the farthest node from the sink (node N3) to ensure that nodes farther from the sink have close neighbors.

Figure 4 shows a chain example (N3, N1, N5, N2, N4) after executing the greedy algorithm. When a node dies, the chain is reconstructed in the same manner to bypass the dead node. Node N5 presents the chain-head and it is responsible to transmit the gathered data to the sink. The chain-head is equitably rotated among the nodes of the chain. Chain-head is selected randomly, and each node has the chance to be the leader once every N round (N is the number of nodes).

For gathering data in each round, the chain-head sends a control packet to its neighbor to start the data transmission from the end of the chain. In Figure 4, node N5 sends this control packet along the chain to node N3. Node N3 will send its data towards node N5. After node N5 receives data from node N1, it will pass the control packet to node N4, and node N4 will pass its data towards node N5 via node N2. Each intermediate node has to aggregate the received data with its local data before sending it to its next neighbor node in the chain. Thus, chain-head sends only one message to the sink by round.

This technique of communication used by PEGASIS allows saving more energy compared to that of LEACH [42] to increase the lifetime of the network and to reduce the bandwidth consumed by using local collaboration between the nodes and by tolerating the failure of the sensor nodes. However, the direct communication between chain-head and sink consumes more energy especially when the distance between them is longer. Moreover, the latency is more important; thus, this protocol cannot be used for real-time applications.

3.1.4 Simple Energy-Efficient Routing (SEER)

SEER's [48] data dissemination is source-based which eliminates the need for the sink to flood an interest for data through the network. Nodes only transmit data when new data are observed. Data are routed along a single path, which is dynamically established. Every time when a node needs to send data, it selects one neighbor to send the message based on the neighbor's hop count and available energy.

Once the network has been deployed, the sink transmits a broadcast packet with header field of 64 bits. 16 bits are reserved for node, each node in the network is assumed to have a unique address within the network and the header field contains source and destination node addresses. 8 bits are reserved to identify new broadcast messages using a sequence number. The sink increments the sequence number every time it sends a new broadcast message. Nodes store the sequence number locally and forward the broadcast messages only if the sequence number of the message is different from the stored one. The sequence number permit to avoid redundant forwarding of old broadcast messages. An 8 bit hop count ensures that nodes can be up to 255 hops away from the sink.

When a node receives this initial broadcast message, it checks whether it has an entry in its neighboring table for the node that transmitted the message. If not, it adds an entry that consists of the neighbor's address, hop count, and energy level. The node then increments the hop count stored in the message and stores this hop count as its own hop count. It then retransmits the broadcast message, but changes the source address field to its address and the energy level field to its remaining energy level. Every node in the network retransmits the broadcast message once, to all of its neighbors. If a node receives a broadcast message with a lower hop count than the hop count it currently has, it updates its hop count. When this initial broadcast has been flooded throughout the network, each node knows its hop count and has the address, hop count, and energy level of each of its neighbors.

When a node observes new data, it initiates the data dissemination process and specifies in the header of the new message the type of new data (normal or critical), source, destination and creator addresses. A critical message has to transmit to two neighbors instead of only one. The neighbor is selected based on the hop count and the remaining energy. A node searches the neighbor with smaller than or at least the same hop count that it has. If there is only one, this neighbor is selected as the destination for the message. If there is more than one neighbor with a smaller hop count, the node selects the neighbor who has the highest remaining energy. If a node does not have any neighbor with hop count smaller or equal to its own hop count, the message is discarded.

Before the message is sent, the remaining energy entry for the selected neighbor is decreased in the neighboring table. If the message is a critical one, a second neighbor will be selected using the same process. Here, using hop count as the routing metric ensures that the message is always sent to the direction of the sink.

When nodes receive a data message, they update the remaining energy value of the sending node in their neighboring table and forward the data using the same dissemination process; the sending node has to be excluded from the list of the neighboring nodes to avoid any routing loop in the network. When a node's remaining energy decreases than a certain threshold, it transmits an energy message to all of its neighbors to inform them about its energy level.

3.1.5 Energy Aware routing Protocol (EAP)

In [49], authors proposed a novel energy-aware routing protocol (EAP) to prolong the lifetime of sensor networks. EAP introduces a new clustering parameter for cluster head election. As LEACH, EAP is divided into rounds, each round begins by a set-up phase in which clusters are organized and routing tree is constructed, followed by a working phase to collect and send data to the sink node. In EAP protocol, each node needs to maintain a neighborhood table to store the information about its neighbors. Each node located in the cluster range is seen as neighbor. At the beginning of each round, each node broadcasts its residual energy ([E.sub.r]) to its neighbors and setup its state as cluster head candidate. Each node receives the residual energy from all the neighbors in its cluster range. Then, accordingly it updates its neighborhood table and calculates the average residual energy ([E.sub.a]) of the cluster range and the broadcasting delay time T using the following equations:

[E.sub] = [[summation].sup.n.sub.i=1] [E.sub.ri]/n

where n is the number of neighbors in the cluster range.

T = K x P x [E.sup.a]/[E.sub.r],

where K is a real value uniformly distributed between 0 and 1, and P is the time duration for cluster heads election.

During the T time, a sensor waits to receive any proposed cluster head message from its neighbors. If it does not receive any proposition, it proposes itself and broadcasts its proposition to be cluster head to its neighbor nodes. After broadcasting its cluster head proposition, it has to wait 2 x [DELTA]t, where [DELTA]t is the time interval which can ensure that all neighbor nodes can receive the cluster head proposition message, to make sure whether there exists another cluster head proposition broadcasted by other nodes in its cluster range. If it does not receive any proposition from its neighbors over [DELTA]t, it sets its state as "Head", or else, it compares its weight with the weights of other broadcasting neighbors. If its weight is the largest one, it sets its state as Head and other broadcasting neighbors give up the competition. Otherwise, the node sets its state as member sensor of this cluster.

To reduce energy consumption, EAP adopts the same intra-cluster coverage scheme introduced in [50]. This scheme permits the cluster head to choose randomly active nodes to ensure a certain required coverage limit. The remaining nodes perform as redundant nodes and turn their radios off to minimize energy consumption.

To define the routing tree after clustering (Figure 5), each cluster head broadcasts within a cluster range a weight message, which contains node ID and its weight W defined as below:

[W.sub.i] = D ([RSS.sub.i]) x [E.sub.a]/[D ([RSS.sub.max]) x [E.sub.r]]

where [RSS.sub.i] is the node i's received signal strength for the signal broadcasted by the base station, [RSS.sub.max] is a constant which is determined by the location of the base station, and the function D is used to estimate the distance between node % and the base station. After the deployment of sensors, the base station broadcasts probing message to all sensors and sensors acquire the RSS according to the received signal strength. RSS remains constant during the network lifetime unless base station varies its location or sensor nodes are mobile.

The cluster head compares its own weight and the received weight of the other neighbor cluster heads. If it has smaller weight, it selects the node that has the largest weight as its parent and sends a message to notify the parent node. After a specified time, a routing tree is constructed. The root node has the largest weight among all cluster heads in the same independently connected component. The node that is closer to the base station and located in a sub-region with full energy would be the root node of routing tree due to its higher weight. After routing tree construction, cluster heads broadcast a TDMA schedule to their active member nodes to be ready for data gathering.

The intra-cluster coverage scheme presents the main advantage of EAP--in fact, this technique improves energy consumption and reduces the TDMA schedule overhead. According to the authors and compared with the previous works [27], [52], the selection cluster head technique prolongs the lifetime of the nodes that have low residual energy within the cluster range. Nevertheless, the drawback of this protocol dwells in the high control packet overhead provided during each round which incurs an extra energy and delay cost.

3.1.6 Directed Diffusion dissemination protocol (DD)

Directed Diffusion data dissemination protocol [53-54] is the first proposed data-centric communication protocol for wireless sensor scenarios. The data generated by the producer is named using attribute-value pairs. Figure 6 shows the operation of data-centric communication protocol. Directed diffusion is based on query, where sink queries the sensors in an on-demand fashion by disseminating an interest. As shown in Figure 6, DD consists of three phases: Interest propagation, initial gradient setup, and data delivery along reinforced path.

In directed diffusion, the data generated by the nodes are named by attribute-value pairs and the data dissemination process is a destination-initiated reactive routing technique in which routes are established when requested.

In the Interest propagation phase, the sink node broadcasts its interest message to all its neighbors. All nodes have an interest table in which the received interest message has to be saved. Each entry in this table has several fields. The most important fields are timestamp field which contains the last received matching interest, gradient fields which contain the data rate specified by each neighbor and duration field which contains the lifetime of the interest. When a node receives an interest, it checks its interest cache to check if it has an entry.

It creates one if there is no matching interest and a single gradient field is created towards the neighbor from which the interest is received and forwards the requested interest message to its neighbors. A gradient is removed from its interest entry when it expires. A gradient specifies both the data rate as well as the direction in which the events are to be sent. If the interest exists, the timestamp and the duration fields are updated in the entry and the second step starts.

In the initial gradient setup, the sensor node which has a matching interest entry generates event samples and sends an event to all its neighboring nodes for which it has gradients. The last phase begins when the sink starts receiving this event, possibly along multiple paths. The sink then sends a reinforced packet to the neighbor node which is the first one receiving the target data. The neighbor node which receives the reinforced packet can also reinforce and select the neighbor node which can receive the new data first. Consequently, a path with maximum gradient is formed; hence, in future, received data packets can be transmitted along the best reinforced path. Finally, the real data are sent from the source to the sink using the selected path.

3.2 Multiple Static Sinks Data Dissemination Protocols

3.2.1 A Stateless Protocol for Real-Time Communication in Sensor Networks (SPEED)

SPEED [48] is a real-time communication protocol designed for sensor networks. Speed provides three types of real-time communication services namely: real-time unicast, real-time area-multicast, and real-time area-anycast. These communication types are defined as follows.

Real-time unicast. This is "more to one" communication mode which occurs when one part of a network detects some activity that needs to be reported to a remote base station.

Real-time area-multicast. Contrary to the first communication type, real-time area-multicast is "one to more" communication mode. This type of communication occurs when the base station initiates the communication by sending its query to an area in the sensor network.

Real-time area-anycast. This communication mode can be used when the response of any sensor node is sufficient. SPEED is specifically customized to be a stateless protocol. That means, it only maintains immediate neighbor information and does not require a routing table. SPEED provides a uniform delivery speed across the sensor network to meet the requirement of real-time applications such as disaster and emergency surveillance in sensor networks. To avoid congestion, SPEED uses a novel backpressure rerouting scheme to re-route packets around large-delay links with minimum control overhead. It also uses nondeterministic forwarding to balance each flow among multiple concurrent routes.

The routing module in SPEED is called Stateless Nondeterministic Geographic For- warding (SNGF) and it works with four other modules at the network layer [55]. Figure 7 shows these different modules.

The beacon exchange mechanism is used to collect information about nodes and their locations. Delay estimation at each node is made by calculating the elapsed time when an ACK is received from a neighbor as the response to a transmitted data packet. SNGF scheme selects nodes that would meet the speed requirement by estimating delay values.

In case no such node is found, relay ratio of the nodes is calculated. Neighborhood Feedback Loop (NFL) module is responsible for providing relay ratio of a node, which is fed to the SNGF module. The relay ratio of a node is calculated by looking at the miss ratios of its neighbors that could not provide the desired speed. The packet is dropped if the relay ratio is less than a randomly generated number between 0 and 1. When a node fails to find a next hop node, the backpressure-rerouting module is finally used to prevent voids and to eliminate congestion by sending messages back to the source nodes so that they would pursue new routes. In comparison to Dynamic Source Routing (DSR) [56] and Adhoc on-demand vector routing (AODV) [57], SPEED performs better in terms of end-to-end delay and miss ratio. SPEED reduces transmission energy consumption, control packet overhead, and traffic distribution. It is also able to achieve load balancing in the network to a great extent. SPEED, although is a successful real-time WSN routing protocol based on simple routing algorithm, it is not really energy efficient. SPEED uses only one delay threshold overall to manage transmission of data packets at the highest transmission velocity. As a result, it cannot satisfy different requirements for transmission delay and causes huge energy consumption.

The protocol indeed results in energy exhaustion of nodes quickly because it selects nodes having high transmission velocity without considering the remaining energy of nodes. Therefore, for a more realistic understanding of SPEED's energy consumption, there is a need for comparing it to a routing protocol, which is energy-aware.

In addition to these issues, the idea of per-flow reservation appears to be non-scalable in SPEED due to the highly dynamic links and route characteristics. So, SPEED might not be scalable well for large WSNs. There is an extension of SPEED, called FT-SPEED [58], which is proposed to handle the void problem caused by high sensor failure probability in WSN. In FT-SPEED, a "void announce" scheme is designed to prevent packets from reaching the void through other routing paths. It also introduces a void bypass scheme to route the packets around two sides of a void to guarantee that the packets are delivered rather than just being dropped.

3.2.2 Multi path Multi SPEED (MMSPEED)

Multi-path and Multi-SPEED Routing Protocol (MMSPEED) [59], an extension of SPEED is designed to support multiple communication speeds, which provides differentiated reliability. A key feature of MMSPEED is that it addresses both real-time issue and reliability separately. The main goals of MMSPEED design are.

* Localized packet routing decision without global network state update or a priori path setup.

* Providing differentiated QoS (Quality of Service) options in isolated timeliness and reliability domains.

For the first goal, geographic routing mechanism based on location awareness is used. Each sensor node is assumed to be aware of its geographical location. This location information can be exchanged with immediate neighbors with "periodic location update packets". Thus, each node is aware of its immediate neighbors within its radio range and their locations.

For the second goal, MMSPEED provides multiple delivery speed options that are guaranteed network-widely. For this, the idea of SPEED protocol [48] which can guarantee a single network-wide speed is used. MMSPEED assumes a few important assumptions.

(1) All nodes know their geographical location.

(2) Location of the packet destination is known.

(3) The underlying MAC protocol allows prioritizing between different classes at least stochastically.

(4) Each speed level is mapped onto a MAC layer priority class.

Associating messages with deadlines focuses on the problem of providing timeliness guarantees for multi-hop transmissions in a real-time sensor application. In such application, each message is associated with a deadline and may need to traverse multiple hops from the source to the destination. Message deadlines are derived from validity of the accompanying sensor data and start time of the consuming task at the destination. The protocol reduces deadline misses by scheduling message based on their perhop timeliness constraints. It supports a probabilistic QoS guarantee by provisioning QoS in two domains -Timeliness and Reliability. QoS differentiation in timeliness is provided through multiple network-wide packet delivery speed guarantees. The scheme employs localized geographic packet forwarding augmented with dynamic compensation, which compensates for local decision inaccuracies as a packet travels towards its destination. The intermediate nodes can lift speed level if they find that the packet may miss the delay deadline with current speed but may meet it at a higher level. To reduce the number of collisions, the QoS has been enhanced in [60] by adapting the Contention Window Adapter (CWA) mechanism in which a dynamic contention window has been used.

In supporting service reliability, probabilistic multi-path forwarding is used to control number of delivery paths based on the required end-to-end reaching probability. In this scheme, each node in the network calculates the possible reliable forwarding probability value of each of its neighbors to a destination by using the packet loss rate at the MAC layer. According to the required reliable probability of a packet, each node can forward multiple copies of it to a group of selected neighbors from the forwarding neighbor set to achieve the desired level of reliability. These mechanisms for QoS provisioning are realized in a localized way without global network information, which is desirable for scalability and adaptability to large scale dynamic sensor networks.

Though, MMSPEED [59] does some improvements over SPEED and differentiates among different real-time levels, it does not dynamically adjust routing paths according to the node's energy state. Both SPEED and MMSPEED have a common deficiency that is they do not take into account the energy consumption metric. This metric has been considered by EAMMSPEED protocol [61] which tries to balance the load and energy consumption of individual nodes in the network and improve the overall network lifetime. Therefore, each node makes routing decisions based on the following four parameters: geographic progress towards the destination sink, required end-to-end total reaching probability, delay, and residual energy at the candidate forwarding node. The performance evaluation shows that EAMMSPEED protocol provides stable service in the sensor network and maximizes the lifetime of the entire network while maintaining the QoS guarantees provided by MMSPEED.

3.2.3 Sequential Assignment Routing (SAR)

SAR [62] is the first protocol for WSN-oriented QoS. SAR calculates multiple paths from the source nodes to the sink, by building trees rooted from a 1-hop neighbor from the sink (Figure 8) and growing outward until it reaches leaf nodes while avoiding paths with low energy or low QoS guarantees. At the end of this process, each leaf node would belong to multiple trees and thus, would have multiple paths to reach the sink.

For each node, two parameters are associated with each path:

(1) Energy resource estimated by the maximum number of packets that can be routed before all of the energy is depleted.

(2) Additive QoS metric, where higher metric implies lower QoS.

Each node generating packets makes a decision about which path to choose. This decision is based on the energy resource and a weighted QoS metric which is the additive QoS metric multiplied by a weight coefficient associated with the priority level of the packet.

SAR shows an optimized performance focusing on lowering of the energy consumption of each packet without considering its priority. A routing table update revolves around the network so as to update all the routing tables of the network in order to find out the depleted nodes in the network and ignore any further communication through the ruined path.

The objective of the SAR algorithm is to maximize the lifetime of the network while minimizing the average weighted QoS metric. One of the drawbacks of this protocol is the high overhead due to the large number of tables being kept on each node, especially when the number of nodes becomes huge.

3.2.4 Hierarchy-Based Multipath Routing Protocol for Wireless Sensor Networks (HMRP)

In HMRP [63], sensor nodes are assumed to be fixed for their lifetimes, and the identifier of sensor nodes is determined a priori. Additionally, these sensor nodes have limited processing power, storage and energy, while sink nodes have powerful resources to perform any task or communicate with the sensor nodes.

HMRP is based on the hierarchical tree architecture, in which the sink nodes serve as root nodes. Each sensor node must be a member of the architecture. The protocol has two phases: Layer Construction Phase (LCP) and Data Dissemination Phase (DDP).

In the first phase, HMRP forms hierarchical relations by broadcasting a network construction packet (NCP) to all its neighbors. This packet contains Seq_Number, Hop_Count, SourceJD, Sink_ID, Packet_Type. The sink node initiates the Hop_Count by one, updates the other fields and broadcasts the packet with Layer Construction Request type to discover the one hop nodes. Each sensor node that receives this packet compares the Hop_Count field with its hop value in its routing path formation table. If Hop_Count field is smaller than its own hop value, then it keeps the packet during some period and updates its routing path formation table, else it drops the packet. If the time of the period duration is finished, the node selects the Source_ID of the received packets with the lowest Hop_Count values as its candidate parents. If the node receives more packets with the same lowest Hop_Count, it saves all Source_ID of the received packets as its candidate parents. This node then updates the Hop_Count and the Source_ID fields in the packet and rebroadcasts it again. Every node continues flooding the Layer Construction Request type packet until the network level is constructed.

In the second phase, Sensor nodes can start disseminating the sensed data to the sink via the parent node. A Received Data Acknowledge (RDACK) packet is sent when the data packet is successfully transmitted to the parent node. The parent node then replies with this packet to notify the source node, and forwards the data packet to next hop. In case of several parents, the source node chooses the parent node as next hop using Round Robin Scheduling when it wishes to send a data packet to a sink. When the source node receives an ACK from the selected parent, it moves this record of this parent in its routing path formation table to the last position and transfers the data packet to the next parent. If no ACK reply is obtained from the parent during some period of time, the source node deletes the record of the concerned parent from its routing path formation table.

The main advantage of HMRP is that the sensor node needs only to know to which parent node to transfer, without maintaining the whole path information. This can reduce the overhead of sensor node. Furthermore, HMRP supports multipath data forwarding path which distributes the energy and prolongs the lifetime of network. However, this protocol has some weaknesses like using an ACK to notify the reception of each data packet increases the network overhead and consumes energy too. This information can be recorded from MAC layer. Moreover, HMRP supports multiple sink nodes scenario, but it does not specify any sink node management procedure--in fact, sink nodes work without any coordination among them, and thus, it has an impact on the overall network performance.

3.2.5 Sinks Accessing data From Environment (SAFE)

SAFE [64-65] is a data dissemination protocol for wireless sensor networks. In this protocol, sensor node can disseminate its sensed data to sinks that explicitly present their interests by sending data requests. Each data sink is allowed to specify its own desired data update rate. SAFE has two major phases: query transfer and dissemination path setup.

In query transfer phase, user sends -via sink node- its query specifying the location, the sensor data type, the desired data update rate, and the service duration. Every node maintains a recent query table and a data management table. The query table records the most recent queries that have been received, and the data table keeps the status of sensor data being or to be distributed by the node. Each node that receives the query performs the tasks as noted below.

* Check the query table if the same query has recently been dealt with. If so,

* Ignore the new query, Otherwise,

* Save the query into its query table

When the node is the data source, it sends a PathSetup message to the inquiring node via unicast. If the node is not the source but on a dissemination path, which is called a junction node, it sends a JunctionInfo message to the sink via unicast. When the node is neither the data source nor a junction, it forwards the query to the next hop, as long as it is not farther away from the queried location than the previous hop node. The hop sender information might be extracted from the packet header filled by the routing protocol in use, or injected by this data dissemination protocol before forwarding a query.

In the second phase of dissemination path setup, each intermediate node has already inserted the necessary information in its data management table while receiving the PathSetup message during the last phase. The intermediate node sets a timer for waiting an Ack message from its descendant, which confirms the path is activated. When the sink node receives the PathSetup and the JunctionInfo messages, it waits for a certain amount of time and then subscribes to the node that sent the best feedback until then. If the best one is a junction node, the sink sends a Subscribe message to this node. When junction node receives this message, it sends a TrailSetup message to that sink and establishes the dissemination path. Otherwise, when the source is eventually the best subscription point, the sink sends an Ack to its progenitor and every progenitor acknowledges its progenitor in turn until the source gets an Ack message and establishes the dissemination path.

4. Mobile Sink Data Dissemination Protocols

Mobile sink wireless sensor network has recently attracted a lot of attention from the research community. Recent works [66-72] have shown that the use of mobile sink can enhance connectivity and lifetime of WSNs. Mobility has been proposed as an alternative way in the literature for reducing the communication distance between sensor nodes and sinks. Network lifetime can be improved with mobile sinks by reducing multi-hop communication and avoiding the bottleneck problem, which appears on the nodes close to the static sink.

In wireless sensor networks, mobility can appear in three main forms [73]: mobility of the sensor nodes that sense the environment and transmit the sensing data, mobility of sinks that gather the information from the network and forward data to the applications, and mobility of the observed event. Sinks can adopt mobility schemes according to the nature of WSN application and its requirements. This mobility can be classified into three categories [66].

(1) Uncontrolled or Random Mobility [74-77].

(2) Predictable Mobility [70], [78-79].

(3) Controlled Mobility [66], [68-69], [80].

4.1 Single Mobile Sink Data Dissemination Protocols

4.1.1 Congestion Avoidance Energy Efficient routing protocol (CAEE)

In [81], the authors tried to solve the problems of data loss due to congestion around the static sink, and the high energy consumption of the sensor nodes located in the vicinity of the static sinks. Therefore, the authors present a routing protocol that is based on an in-network storage model [82], and a mobile sink.

In an environment where the sensor nodes are uniformly but randomly deployed, the authors assume the following points.

* Sensor nodes are grouped into clusters, where cluster head is designated for each cluster.

* Cluster head forwards the received data from neighboring head nodes and the nodes of its cluster towards the sink.

* Depending on the node density and sensor field coverage requirements, the cluster head manages the nodes of its cluster by assigning them "awake" or "sleep" status.

* Each node has a list of all its neighboring nodes. CAEE is based on discrete sink mobility along a fixed trajectory. In CAEE, Mini-sinks (MS) are created utilizing the in-network storage model [82] along the mobility trajectory of the sink. Each MS (Figure 9) is considered as a cluster of sensor nodes managed by cluster head called a data collector node (DC). The DC node receives the collected data from the sensor field and stores it in the MS nodes. The mobile sink periodically visits each MS and retrieves the stored data.

The CAEE protocol does not impose any restriction on the shape of the mobility trajectory of the sink. The mobility path of the sink is along the periphery of the sensor field. During its first trip along the periphery of the sensor field, the sink selects a subset of sensors as DC nodes. Each DC node sets up its MS, and broadcasts this information to the sensor nodes. The sink node starts its first mobility round along the periphery of sensor field to select the DC nodes. It chooses the first or the starting node as DC1 if the last one is cluster head. Otherwise, the sink queries the start node about its cluster head node. On retrieval of the required information, the sink assigns the status of DC1 to the obtained cluster head node. Thus, the sink starts its mobility along the periphery of the sensor field, and selects the second data collector node DC2 that is located at least H hops away from DC1. Similarly, the third data collector node DC3 is located at least H hops away from DC2, and so on. In this way, a set of DC nodes are created along the periphery of the sensor field.

To create the MS nodes, each DC node broadcasts a message to invite the sensor nodes to joint its K-hop cluster. The message contains the ID of the DC node and the hop count that is initialized with 1. Each sensor node receiving this message does the following tasks:

* Compares the available routing path to a data collector node with the newly reported route and keeps the shortest one.

* Increments the hop count by 1 in the received message and forwards it.

After a certain period of time, each node knows a shortest possible route to one of the data collector nodes as shown in Figure 9.

Each sensor node sends the collected data to the nearest DC node which stores it in one of the buffer nodes of its MS nodes. The mobile sink stops at each MS and requests data transfer from the DC node. The DC node reports the total number of bytes that it wants to transfer to the mobile sink and then the data transfer starts.

In this protocol, the collected data from the sensor node to the MS node are transmitted over the shortest path which increases the lifetime of the network. Also, congestion and data delivery delay have been improved because of the mobility of sink and the multiple MSs. However, this protocol may suffer from latency which can be increased as the number of MSs in the network increases. Hence, this protocol is not recommended for a large scale sensor network and real-time applications.

4.1.2 Sink Mobility Protocols for Data Collection (SMPDC)

Sink mobility has been investigated as a method for efficient and robust data delivery in wireless sensor networks [83], [84]. The authors proposed four mobility patterns for the sink, mostly randomized (such as the simple random walk, biased random walks, and walks on spanning sub graphs), as well as predictable mobility (moving on a straight line or cycle). These patterns assume and exploit different degrees of freedom, simplicity and network knowledge. To get data from sensors, the sink movement is combined with three data collection strategies: a passive, a multi-hop, and a limited multi-hop.

The authors consider an environment composed of a huge number of small homogeneous sensor devices with limited capabilities. They suppose that the sensor devices are randomly deployed in a flat square area. The sink does not have any resource limitation, it can calculate accurately its position using Global Positioning System (GPS) and it is aware of the dimension of the network area.

The first proposition called Random Wall and Passive Data Collection (RWPDC) is a simple mobility pattern in which the mobile sink moves randomly towards a chosen direction with constant speed. The mobile sink selects a random uniform angle in [-[tau], [tau]] radians. This angle defines the deviation of the mobile sink's current direction. To determine the new position, the mobile sink selects a uniform random distance d [member of] (0, dmax) which is the distance to travel along the newly defined direction. If the new position is outside the network area, the sink decreases the distance to the network area border. The data are collected in passive manner. The sink broadcasts periodically a beacon message. Each sensor node that receives this message replays by transmitting its data to the sink. RWPDC presents the simplest possible movement, guarantees visiting all sensors in the network, and thus, collects data even from disconnected areas in case of few/faulty sensors or in the presence of obstacles. However, the latency is the biggest problem of this method.

The second proposition is called Partial Random Walk with Limited Multi-hop Data Propagation (PRWLMDP). In this proposition, the authors assume that the network area is partitioned as equal square regions. The center of each region is connected to the center of the adjacent regions. Initially, the mobile sink is positioned on or near one of the center nodes. Then, it calculates the next position by selecting randomly one of the neighbors of the current center. Thus, the sink moves toward this new position with predefined constant speed.

The data collection protocol forms propagation trees. The sink periodically broadcasts a beacon message and indicates the depth of the propagation trees by setting a TTL (Time To Live) parameter. This process creates a number of propagation trees within the network with the roots of these trees being one hop away from the sink. Sensor nodes that belong to propagation tree may begin immediately forwarding their data to the sink.

As the sink moves on, the propagation trees may become disconnected. When the root node loses the communication with the sink, it simply caches all data both generated and relayed, and waits to hear another beacon message to begin the propagation process again.

In this scheme, the distance traveled by the sink is reduced and the time to visit network nodes is accelerated. However, PRWLMDP uses more knowledge of the network which is more expensive in terms of communication and computational costs on the sensor devices.

The third proposition called Biased Random Walk with Passive Data Collection (BRWPDC) extends the previous one (PRWLMDP). The authors use the same assumptions and the same data collection protocol. In this proposition, the sink calculates its next position based on two parameters: the visiting frequency of the region and the number of sensor nodes in the region. The center of the region that has the low visiting frequency and the high number of sensor nodes will be selected as the next position to move toward it. The low visiting frequency is preferred to speed up the coverage of new areas. To increase data delivery in areas with many nodes, the region of high number of sensor nodes is preferred.

The last proposition is called Deterministic Walk with Multihop Data Propagation (DRWMDP), in which the sink's movement is predefined. The trajectory is characterized by its length (L). The authors use a particular trajectory (line or circle trajectory). The linear trajectory consists of a horizontal or vertical line segment passing through the center of the network. The sink moves from one edge of the line to other and returns along the same path. The circle is centered at the center of the network and its radius is defined as R = L/2[tau]. Initially, the sink is positioned on the perimeter of the circle and continues along this path. In this kind of sink mobility, the authors use a data collection protocol similar to the one presented in the second proposition (PRWLMDP), without the timeout and TTL mechanism, thus paths are created according to minimum hop distance and span throughout the whole network area. The deployment of this protocol imposes a high cost on the sensor devices that perform tree formation and multi-hop propagation. However, it seems that the delivery latency is lower than any of the three previous propositions. Furthermore, the selection of the trajectory length introduces a trade-off between the cost at the sink and the cost at the sensors.

The simulation results show that for applications where time efficiency is not critical and the energy saving is important, it is better to let the sink traverse the whole network area, as given in the first and the third propositions. When the latency is important, the second proposition is more appropriate. For the applications where the mobility capabilities of the sink are limited but can tolerate some loss of information and increased energy consumption, the last proposition is more suitable.

4.1.3 Density-based Proactive Data Dissemination Protocol (DEEP)

In [85], the sensing data are proactively distributed and stored throughout the network. The mobile sink is free to choose its own trajectory in any way and at any time. The only condition imposed on the mobile sink in order to retrieve a representative view of the monitored data is on the total number of nodes the mobile sink needs to visit. In DEEP, data dissemination strategy uses a combination of density sensitive probabilistic forwarding with deterministic corrective measures, as given in [86].This technique permits to ensure a predefined average number of transmissions and retransmissions of each message. Based on calculated probability, each node can decide to broadcast any message after receiving it for the first time. If the node does not decide to retransmit the message, it should wait for a given delay and if it does not receive this message from any of its neighbors, then this node retransmits the message. Moreover, in this protocol, the node can store the received message based on another calculated probability.

The simulation results show that DEEP [85] is the more viable solution, especially for sparse networks, when the frequency of sending messages is low, and when the amount of sensed data reported in each message is large. However, the simulation results illustrate that the data storage is well distributed in the network.

4.1.4 Data Collection with Adaptive Stopping Times (DCAST)

In this protocol, the authors propose biased sink mobility with adaptive stop times, as a method for data collection in wireless sensor networks [87-88]. The system model contains a single mobile sink and a vast number of ultra-small homogeneous static sensor devices. Each sensor is a fully-autonomous computing and communication device, characterized mainly by its limited power supply. The sensors are deployed randomly in flat square area. The authors assume the existence of some regions, called pockets, in the network with high sensor node density. Each pocket represents a circular area and does not overlap with another pocket. Moreover, the authors suppose that the mobile sink is not resource constrained. This sink is assumed to be powerful enough in terms of computing, memory, and energy supplies. The sink can accurately calculate its position using GPS and it is aware of the dimensions and the boundaries of the network area. Also, it moves with constant speed according to a given mobility function. The mobility function can be invoked at anytime even before reaching the designated point.

As shown in Figure 10, the network area is partitioned, during the network initialization, as equal square regions, called cells. The center of each cell is connected with unidirectional edges only to the four centers corresponding to adjacent cells.

Thus, when the sink is located at the center of a cell, it can communicate with every sensor node within the cell area. The sink collects the data in a passive manner and broadcasts beacon messages within the cell. Nodes that receive a beacon start transmitting the data stored in their memory to the sink. Initially, the mobile sink is positioned on one of the central nodes. In the Figure 10, two sink mobility schemes proposed by the authors: deterministic walk and biased random walk are represented by the blue-thin and the red-thick arrow lines respectively.

In the deterministic walk, the sink visits cells from left to right and vice versa according to the blue-thin arrow. By moving on this trajectory, the sink can communicate with each node in the network. This walk assumes some global network knowledge to know the boundaries of the cells and the network. It avoids visit overlaps and multi-hop communication, which optimizes the time needed to cover the network. However, in this kind of walk, it may not be feasible to traverse the network with the presence of obstacles that may hinder the movement of the sink. Also, the network topology may not be known to the sink or may change dynamically. To avoid these inconveniences, the authors proposed the second sink mobility scheme (biased random walk) represented by the red-thick arrows in the Figure 10.

In this walk, the next position of the sink is determined by selecting the center of one of the neighboring cells. A frequency is associated with each cell - the sink increases this frequency for each visit to the corresponding cell. The selection of the next area to visit is done in a biased random manner depending on this frequency and the less frequently visited regions are favored.

For each mobility scheme, the authors proposed two different types of stopping for the sink. In the Constant stop time, the mobile sink takes a constant and an equal pause time at each cell. In the Adaptive stop time, the sink can leave the cell before the end of the stopping time to avoid spending a lot of time in a cell without collecting data, when the sensors empty their memory before stopping time expires. Also, at each cell, the sink waits for some time after the end of the stopping time to receive the eventual pending data.

4.2 Multiple Mobile Sinks Data Dissemination Protocols

4.2.1 Coordination-based data dissemination protocol for wireless sensor network (CODE)

CODE [89] considers energy efficiency and network lifetime, especially for sensor networks with high node density. In this protocol, all sensor nodes are stationary except the sinks nodes. The authors assume that each sensor is aware of its residual energy and its location using the Global Positioning System (GPS) [90]. As shown in the Figure 11, the sensor network field in CODE is divided into grids. Each grid is indexed based on its geographical location. During the data dissemination process, each grid participates by only one coordinator node. The other sensors remain in the sleeping mode using GAF (Geographical Adaptive Fidelity) protocol [91]. The coordinator acts as an intermediate node to cache and relay data.

CODE has two major phases: query transfer phase and data dissemination phase. For example in Figure 11, if an event is detected (grid [1, 0]) the source generates a data-announcement message and sends the message to all coordinators using a simple flooding mechanism. Then, the interested sink sends a query (query transfer phase) to the source node along the path [2, 2] [right arrow] [1, 1] [right arrow] [1, 0] which will be used to transport the sensing data during the data dissemination phase. However, the sink checks its geographical location periodically. If the sink moves out of the grid (from [2, 2] to [2, 1]), it has to send a message to remove the previous data dissemination path and then resends a query to set up a new one ([2, 1] [right arrow] [1, 0]).

CODE establishes a better data dissemination path based on the grid ID without flooding and additional phase. The sinks do not need to periodically propagate their geographical location to the sources. Moreover, CODE takes into account query and data aggregation to reduce the amount of data transmitted from multiple sensor nodes to sinks. However, the random sink mobility presents the major inconvenience of this protocol. The mobile sink can move at anytime, goes away from the source node. Thus, it may increase the latency and energy consumption.

4.2.2 Two-Tier Data Dissemination Protocol (TTDD)

This protocol [77] provides data delivery to multiple mobile sinks based on a decentralized architecture. TTDD uses homogeneous sensors and assumes that each sensor is aware of its own location using GPS [90]. TTDD is grid-based structure. A virtual grid should be created at any new sensed data by the source node. As shown in Figure 12, when the source (S) senses new data, it calculates the locations of its four forecasted neighboring dissemination points (D) based on its geographical position and cell size. The source node then sends a data-announcement message to these neighbors to select the real four dissemination nodes (D). Each dissemination node resends the data-announcement message with the same manner until the construction of the virtual grid as shown in Figure 12.

Instead of broadcasting the location information of mobile sinks to all sensor nodes, TTDD uses a two-tier data dissemination model to deal with the sink mobility problem and reduces energy consumption.

Only sensors located at a cell boundary need to forward the data. The sink proactively builds the two-tier grid structure throughout the network and sets up forwarding points in the sensors closest to the dissemination nodes. The lower tier is the cell at the sink's current location and the higher tier contains the dissemination nodes at cell boundaries. The sink broadcasts its query within its own cell. When the nearest dissemination node in the cell receives the query, it forwards it to its adjacent dissemination node in another cell. This process continues until the query reaches the source node or one of the dissemination nodes that have the corresponding data. During the query propagation, the network establishes the reverse path towards the sink so that the data could be forwarded on the same path as that of the query propagation. TTDD exploits local flood within a local cell of a virtual grid which sources build proactively. However, it does not optimize the path from the source to the sinks. When a source communicates with a sink, the restriction of grid structure may increase the length of a straight-line path. Also, TTDD creates new virtual grid for each new data source. It therefore, increases energy consumption and connection loss ratio. Moreover, sink mobility in this protocol has random scheduler like CODE [89] which affects negatively on the network performance.

In the protocol, EGDD (Energy-Aware Grid-based Data) [92], the author tries to overcome the above TTDD's drawback. In EGDD, the grid is constructed only when no valid grid is present in the sensor field. Also, the dissemination node is selected based on its residual energy. Hence, it can be replaced by another one when its energy becomes equal to the predefined energy threshold. Moreover, EGDD network model ensures query and data forwarding through the shortest path between source and sink. However, sink mobility in this protocol is uncontrolled which brings other challenges for this protocol.

4.2.3 Pseudo-Distance Data Dissemination protocol (PDDD)

In PDDD [93], network partitioning is not considered and mobile sink nodes are assumed to have unlimited battery power. Also, it is assumed that the links between sensor nodes are bidirectional and no control messages are lost. The main idea of this protocol is to create and maintain a Totally Ordered Graph (TOG) using pseudo-distance. As depicted in the Figure 13(a), when a sink node (S) wants to collect data from sensor nodes, it broadcasts an interest message. By receiving this message, each node can set its pseudo-distance and corresponding level from the sink, and then it broadcasts the received interest message to its neighbor nodes with its own level metric. At the end of this operation, the hierarchical levels of communication are created (Figure 13 (b)). Thus, each sensor node uses this TOG to disseminate the requested data.

Mobile sink nodes generate periodical heartbeat messages to their direct neighbors. Therefore, if the mobile sink nodes move, the direct sink's neighbors can detect the sink mobility by losing the heartbeat messages. For the other sensor nodes, PDDD uses ACK packet. Each sensor node that transmits data packets to its next hop should receive an ACK from the latter one. If it is not the case, then the link is considered as failed link. Therefore, the sensor node has to choose another redundant path. However, if a node loses all of its parent nodes, then it has to update its own level locally to find new parent.

This protocol can achieve an acceptable data dissemination level in stable sensor network. However, PDDD needs many control messages (interest, heartbeat and ACK messages) to create and maintain the TOG graph. The number of control messages increases when the number of mobile sinks increases. Thus, this fact affects directly and negatively on the network performances like latency and energy consumption. Also, PDDD did not consider the energy parameter in its data dissemination process. Some nodes may be more soliciting than others which would accelerate the death of these nodes. Moreover, like CODE [89] and TTDD [77], PDDD did not give any strategy of sink mobility which is still random and uncontrolled.

4.2.4 Topology-based Rendezvous Data Dissemination (TRDD)

This protocol [94] assumes a network model of three tiers. Sensor nodes are deployed randomly in the lower tier, each sensor is assumed to be aware of its geographical position using GPS [90]. Sink nodes are placed randomly on the periphery and they constitute the middle tier. The higher tier represents the administration site. TTDD consists of two phases: the events propagation phase and query propagation phase. The dissemination structure of TRDD is based on a simple geometric idea. It considers the network perimeter as a polygon that provides a closed region for the interior nodes called query region (QR). TRDD adapts a modified version of the algorithm presented in [95] to identify dynamically the outer boundary nodes. As shown in Figure 14, the construction of the QR starts once the sinks receive the query from the management site and after discovering the boundary nodes. Thus, QR contains the sensor nodes of the network perimeter and their one-hop neighbors.

When the sensor nodes of one-hop neighbors of the boundary nodes in the QR region receive the query packet, a beacon packet has to be transmitted by those nodes. This beacon allows interior nodes located outside QR to update their neighbor table, to be informed about the creation of the QR region, and to start sending their sensing data toward the QR. After selecting a direction's path (TRDD proposed eight possible directions) based on three policies: Random-based, Round Robin-based or centroid-based, sensor node selects its closest next hop in this direction. In the first policy, sensor node randomly selects one direction. In the second policy, sensor node selects its direction in order after selecting the first one randomly.

In the last policy, sensor node calculates its position relative to the virtual gravity center of the network and selects the contrary direction relative to this center. Intermediate nodes use the same direction chosen by the initiator node except in failure case, where the intermediate nodes should change the direction. Thus, the sensed data will intersect the QR region and will be transmitted in the reverse path to the root sink. In TRDD, the authors consider and evaluate two sink mobility patterns: random and controlled mobility. In the controlled mobility, sinks move along the network diagonal or along the network periphery.

According to the simulation results presented in [94], the diagonal and the random mobility produce extra costs than that of the peripheral mobility. Moreover, TRDD generates low communication overhead in comparison with TTDD [77]. In TRDD, the sink mobility does not need to transmit any additional control packet contrary to TTDD. However, detecting the boundary nodes and creating the QR may add an additional cost. Also, in TRDD when sensor node detects an event, it creates and stores the new data locally until it receives the beacon packet of the query. Thus, this may influence negatively on the latency and the storage capability of the nodes, especially if the query comes with more delay.

5. Comparative Summary of the Works

5.1 Data Dissemination and Sink Mobility Scheme

In this study, several previous works on data dissemination issues have been analyzed. Most of these works have proven that their methods improve the data dissemination process in terms of energy usage, network lifetime, and reliability. Some protocols also address the issue of being real-time. Our study has been principally based on two main parameters: the number of sink and the type of sink (stationary or mobile). We have investigated how these parameters affect the data dissemination process. Table 1 and Table 2 show general overviews of the data dissemination and sink mobility methods of the previously proposed protocols. The character 'slash' (/) in the table means no information is available or not applicable. These tables present respectively the data dissemination protocols with single sink and multiple sinks.

5.2 Data Dissemination protocol and Application Requirements

Data dissemination protocol represents the main important issue of WSNs, as the primary task of these networks is to collect useful data by monitoring phenomena in the nearby environment and transmit the sensed data to the sink for analysis and further processing. Data dissemination and reporting in WSN depends both on specific needs of the application and also on time sensitivity of the collected data. To have clear idea about the application type of each data dissemination protocol, we define, according to the sensor node data sending behavior, the following Data Application Models (DAMs).

* Periodic-Based Data Sending (PBDS). This data dissemination model is required for applications that require periodic data monitoring. In this model, sensor nodes can sleep and periodically wake up, sense the environment, and transmit the sensed data to the sink in periodic intervals.

* Query-Based Data Sending (QBDS). Represents a typical way for extracting data from a sensor network. In this model, sensors only transmit data when it is explicitly requested by the sink.

* Event-Based Data Sending (EBDS). In this model, sensor node reports its sensed data to the sink at any time if the related occurred event meets the conditions required by the application.

* Hybrid-Based Data Sending (HBDS). This model combines the three data reporting models mentioned above. In networks, where different data reporting models coexist, the data dissemination protocol should change--in case of PBDS and EBDS applications for example; the operation mode depends on the importance of the sensed information.

To compare the studied data dissemination protocols using the above data reporting models, we define the following formula to calculate the percentage of each representative model.

DAM(%) = NDAM/NDDP

where, NDAM is the number of data dissemination protocols using the same Data Application Model (DAM). NDDP is the number of the studied data dissemination protocols.

Figure 15 shows that the dominant application models are: the query-based and the event-based models. Typically, a data dissemination protocol is highly influenced by the application data dissemination model in terms of energy consumption and route calculations. While Periodic data reporting based applications may tolerate delay and loss of data, timely and reliable delivery of data may become very important concerns for query-based applications and event-based applications. Hence, both the query-based and event-based approaches are data-centric and well suitable for real-time applications.

6. Conclusions and Future Research Directions

In this paper, we have studied the recent significant research results on data dissemination in wireless sensor networks and classified these protocols into two main categories based on the number of sink (single or multiple) and the nature of its movement (static or mobile). Whatever the category in which any data dissemination protocol belongs, network resource like energy consumption still remains the major concern while designing protocols for wireless sensor networks.

In a static sink approach, sensor nodes do not need to know the geographical position of the sink at each time. Usually, sink broadcasts its location information in the network only once, just after the network deployment. Moreover, sensor node keeps no more than one valid path to forward its data toward the sink. Thus, such stability can help improve the network performance and reduce the network overhead. However, in static sink approach, sensor nodes relatively closer to the sink can be loaded with relaying a large amount of traffic from other nodes. This situation results in energy exhaustion at the nodes near the sink too soon, leading to the separation of the sink from the rest of the nodes that still have plenty of energy.

As presented in the second category, many protocols try to exploit sink mobility to improve the lifetime of the network. However, there is still some kind of skepticism in the research community about the practicality of deploying moving sinks in WSN scenarios. One of the major concerns behind this skepticism is that mobility inevitably incurs additional overhead in data communication protocols and the overhead can potentially offset the benefit brought by mobility. Further research works may investigate this particular issue in-depth and analyze how effective a mobile sink could be in comparison with static sink in the network for data dissemination.

Acknowledgement

This project was supported by NDC Lab, KICT, IIUM. References

[1] C. Hartung, R. Han, C. Seielstad, and S. Holbrook, "FireWxNet: A Multi-Tiered Portable Wireless System for Monitoring Weather Conditions in Wildland Fire Environments," in ACM MobiSys, pp. 28-41, 2006,.

[2] M. Ceriotti, L. Mottola, G. P. Picco, A. L. Murphy, S. Guna, M. Corr'a, M. Pozzi, D. Zonta, and P. Zanon, "Monitoring Heritage Buildings with Wireless Sensor Networks: The Torre Aquila Deployment," in ACM/IEEE IPSN, pp. 277-288, 2009.

[3] S. Kim, S. Pakzad, D. Culler, J. Demmel, G. Fenves, S. Glaser, and M. Turon, "Health Monitoring of Civil Infrastructures Using Wireless Sensor Networks," in ACM/IEEE IPSN, pp. 254-263, 2007.

[4] G. Tolle, J. Polastre, R. Szewczyk, D. Culler, N. Turner, K. Tu, S. Burgess, T. Dawson, P. Buonadonna, D. Gay, and W. Hong, "A Macroscope in the Redwoods," in ACM SenSys, pp. 51-63, 2005.

[5] L. Selavo, A. Wood, Q. Cao, T. Sookoor, H. Liu, A. Srinivasan, Y. Wu, W. Kang, J. Stankovic, D. Young, and J. Porter, "LUSTER: Wireless Sensor Network for Environmental Research," in ACM SenSys, pp. 103-116, 2007.

[6] G. Barrenetxea, F. Ingelrest, G. Schaefer, and M. Vetterli, "SensorScope: Out-of-the-Box Environmental Monitoring," in ACM/IEEE IPSN, pp. 332-343, 2008.

[7] G. WernerAllen, K. Lorincz, J. Johnson, J. Lees, and M. Welsh, "Fidelity and Yield in a Volcano Monitoring Sensor Network," in USENIX OSDI, pp. 381-396, 2006,.

[8] W.Z. Song, R. Huang, M. Xu, A. Ma, B. Shirazi, and R. LaHusen, "Air-dropped Sensor Network for Real-time High-fidelity Volcano Monitoring," in ACM MobiSys, pp. 305-318, 2009.

[9] Y. Kim, T. Schmid, Z.M. Charbiwala, J. Friedman, and M.B. Srivastava, "NAWMS: Nonintrusive Autonomous Water Monitoring System," in ACM SenSys, pp. 309-322, 2008.

[10] J. Al-Karaki, and A.E. Kamal, "Routing Techniques in Wireless Sensor Networks: A Survey," IEEE Communications Magazine, vol. 11, no. 6, pp. 6-28, December 2004.

[11] S. Tilak, N. B. Abu-Ghazaleh, and W. Heinzelman, "A taxonomy of wireless micro-sensor network models," ACM Mobile Computing and Communications Review, Vol. 6, issue. 2, pp. 28-36, Apr 2002.

[12] K. Lin, L. Shu, L. Wang, A.S. K. Pathan, "Data gathering with Multi-attribute Fusion in Wireless Sensor Networks", Advancements in Distributed Computing and Internet Technologies: Trends and Issues (Edited by Al-Sakib Khan Pathan, Mukaddim Pathan, and Hae Young Lee), ISBN: 978-1-61350-110-8, IGI Global Publishers, USA, pp. 159-181, 2011.

[13] R. Rajagopalan and P.K. Varshney, "Data-aggregation techniques in sensor networks: a survey," IEEE Communications Surveys & Tutorials, Volume: 8, Issue: 4, pp. 48-63, Fourth Quarter 2006.

[14] F. Wang and J. Liu, "Networked Wireless Sensor Data Collection: Issues, Challenges, and Approaches," IEEE Communications Surveys & Tutorials, Volume: 13, Issue: 4, pp. 673-687, Fourth Quarter 2011.

[15] I. Akyildiz, W. Su, Y. Sankarasubramaniam, and E. Cayirci, "A survey on sensor networks," IEEE Communications Magazine, vol. 40, issue. 8, pp. 102-114, 2002.

[16] N. Dimokas, D. Katsaros, and Y. Manolopoulos, "Energy efficient distributed clustering in wireless sensor networks," Journal of Parallel and Distributed Computing, vol. 70, pp. 371-383, 2010.

[17] M. Younis, M. Youssef, and K. Arisha, "Energy-aware routing in cluster-based sensor networks," In Proceedings of the 10th IEEE/ACM International Symposium on Modeling, Analysis and Simulation of Computer and Telecommunication Systems (MASCOTS), pp. 129-136, 2002.

[18] A.E. Tumer and M. Gunduz, "Energy-efficient and fast data gathering protocols for indoor wireless sensor networks," Sensors, vol. 10, 2010, pp. 8054-8069.

[19] M. Chen, T. Kwon, and Y. Choi, "Energy-efficient differentiated directed diffusion (EDDD) in wireless sensor networks," Computer Communications, vol. 29, no. 2, 2006, pp. 231-245.

[20] C. Moser, L. Thiele, D. Brunelli, and L. Benini, "Adaptive power management for environmentally powered systems," IEEE Transactions on Computers, vol. 59, pp. 478-491, 2010.

[21] M. H. Anisi, A. H, Abdullah, and S. A. Razak, "Energy-efficient and reliable data delivery in wireless sensor networks," Wireless Networks, vol. 19, pp. 495-505, 2013.

[22] J.H. Chang and L. Tassiulas, "Energy conserving routing in wireless ad-hoc networks," in Proc. IEEE INFOCOM, vol. 1, pp. 22-31, 2000.

[23] K. Kar, M. Kodialam, T. Lakshman, and L.Tassiulas,"Routing for network capacity maximization in energy-constrained ad-hoc networks," in Proc. IEEE INFOCOM, vol. 1, pp. 673-681, 2003.

[24] A. Sankar and Z. Liu, "Maximum lifetime routing in wireless ad-hoc networks," in Proc. IEEE INFOCOM, vol.2, pp.1089-1097, 2004.

[25] L. Li, J. Halpern, P. Bahl, Y.-M. Wang, and R. Wattenhofer, "Analysis of a cone-based distributed topology control algorithm for wireless multi-hop networks," in Proc. ACM PODC, pp. 264-273, 2001.

[26] X.Y. L. W.Z. Song and W. Wang, "A unified energy-efficient topology for unicast and broadcast," in Proc. ACM MobiCom, pp. 1-15, 2005.

[27] W.B. Heinzelman, A.B. Chandrakasan, and H. Balakrishnan,"An application-specific protocol architecture for wireless microsensor networks," IEEE Transactions on Wireless Communications, vol. 1, no. 4, pp. 660-670, Oct 2002.

[28] M. Chu, H. Haussecker, and F. Zhao, "Scalable Information-Driven Sensor Querying and Routing for ad hoc Heterogeneous Sensor Networks," The International Journal of High Performance Computing Applications, vol. 16, no. 3, pp. 293-313, August 2002.

[29] K. Akkaya, and M. Younis, "A survey on routing protocols for wireless sensor networks," Ad Hoc Networks, Elsevier, vol. 3, no. 3, pp. 325-349, May 2005.

[30] S. Singh, M. Singh, and D. Singh, "Routing Protocols in Wireless Sensor Networks - A Survey," International Journal of Computer Science & Engineering Survey, vol.1, no.2, pp. 63-83, November 2010.

[31] M. Di Francesco, S.K. Das, and G. Anastasi, "Data collection in wireless sensor networks with mobile elements: A survey," ACM Transactions on Sensor Networks, Article 7, Vol. 8, no. 1, 31 pages, August 2011. doi:10.1145/1993042.1993049.

[32] I. Raicu, L. Schwiebert, S. Fowler, and E.K.S. Gupta, "Local load balancing for globally efficient routing in wireless sensor networks," International Journal of Distributed Sensor Networks, vol. 1, issue. 2, pp.163-185, 2005.

[33] H. Dai and R. Han, "A node-centric load balancing algorithm for wireless sensor networks," In Proceedings of the 2003 IEEE Global Telecommunications Conference, GLOBECOM'03, vol. 1, pp. 548-552, December 2003.

[34] P. Chatterjee and N. Das, "A distributed algorithm for load-balanced routing in multihop wireless sensor networks," In Proceedings of the 9th international conference on Distributed computing and networking, ICDCN'08, Springer-Verlag, pp. 332-338, 2008.

[35] G. Gupta and M. Younis, "Load-balanced clustering of wireless sensor networks," In Proceedings of the 2003 IEEE International Conference on Communications, ICC'03, vol. 3, pp.1848-1852, May 2003..

[36] C.P. Low, C. Fang, J.M. Ng, and Y.H. Ang, "Efficient load-balanced clustering algorithms for wireless sensor networks," Computer Communications Journal, Elsevier, vol. 31, pp. 750-759, March 2008..

[37] Y. Zhang and Q. Huang, "A Learning-based Adaptive Routing Tree for Wireless Sensor Networks," Journal of Communications, vol. 1, issue. 2, pp.12-21, 2006.

[38] K. Akkaya, M. Younis, and W. Youssef, "Positioning of base stations in wireless sensor networks," IEEE Communications Magazine, vol. 45, issue. 4, pp.96-102, 2007.

[39] D. Luo, D. Zuo, and X. Yang, "An optimal sink selection scheme for multi-sink wireless sensor networks," In Proceedings of the International Conference on Computer Science and Information Technology, ICCSIT'08, pp. 544-548, September 2008.

[40] W.Y. Poe and J.B. Schmitt, "Self-organized sink placement in large-scale wireless sensor networks," In Proceedings of the 17th IEEE International Symposium on Modeling, Analysis Simulation of Computer and Telecommunication Systems, MASCOTS'09, pp. 1-3, September 2009.

[41] E.I. Oyman and C. Ersoy, "Multiple sink network design problem in large scale wireless sensor networks," In Proceeding of IEEE International Conference on Communications (ICC 2004), volume 6, IEEE, pp. 36633667, 2004.

[42] W. Heinzelman, A. Chandrakasan and H. Balakrishnan, "Energy-Efficient Communication Protocol for Wireless Mi-crosensor Networks," In Proceedings of the 33rd Hawaii International Conference on System Sciences (HICSS '00), doi:10.1109/HICSS.2000.926982, January 2000.

[43] P. B. Mirchandani and R. L. Francis, "Discrete Location Theory," Ser. Wiley-Interscience Series in Discrete Mathematics and Optimization, Wiley & Sons, New York, Total Pages 576, July 1990.

[44] Modeling the P-Median problem with OPL. IBM ILOG OPL documentation for ODM Enterprise, IBM, from: http://pic.dhe.ibm.com/infocenter/odmeinfo/v3r3/index.jsp ?topic=/ilog.odms.ide.odme.help/Content/Optimization/Do cumentation/ODME/_pubskel/ODME_pubskels/startALL _ODME_Eclipse_and_Xplatform_doc258.html [last accessed: 7 July, 2013]

[45] A. Manjeshwar and D.P. Agarwal, "TEEN: a routing protocol for enhanced efficiency in wireless sensor networks," In 1st International Workshop on Parallel and Distributed Computing Issues in Wireless Networks and Mobile Computing, pp. 2009-2015, April 2001.

[46] A. Manjeshwar and D.P. Agarwal, "APTEEN: A hybrid protocol for efficient routing and comprehensive information retrieval in wireless sensor networks," Parallel and Distributed Processing Symposium. Proceedings International, pp. 195-202, 2002.

[47] S. Lindsey, C. Raghavendra, "PEGASIS: Power-Efficient Gathering in Sensor Information Systems," IEEE Aerospace Conference Proceedings, vol. 3, pp. 1125-1130, 2002.

[48] T. He, J.A. Stankovic, C. Lu, and T.F. Abdelzaher, "SPEED: A Stateless Protocol for Real-Time Communication in Ad Hoc Sensor Networks," In International Conference on Distributed Computing Systems (ICDCS '03), USA, pp. 46- 55, May 2003.

[49] M. Liu, J.N. Cao, G.H. Chen, and X.M. Wang, "An Energy-Aware Routing Protocol in Wireless Sensor Networks," Sensors, Vol. 9, No. 1, pp. 445-462, 2009.

[50] M. Liu, J. Cao, H.g. Gong, L.j. Chen, and X. Li, "A Distributed Power-Efficient Data Gathering and Aggregation Protocol for Wireless Sensor Networks," Lecture Notes in Computer Science, Volume 3758, Springer Verlag, pp. 743-756, 2005.

[51] A.-S. K. Pathan and C. S. Hong, "An Efficient Scheme for Secure Data Transmission in Wireless Sensor Networks," Proceedings of the 17th Joint Conference on Communication & Information (JCCI 2007), 2-4 May 2007, PyeongChang, p. 48 (CD), Korea.

[52] O. Younis and S. Fahmy, "HEED: A Hybrid, Energy-Efficient, Distributed Clustering Approach for Ad Hoc Sensor Networks," IEEE Transactions on Mobile Computing, vol. 3, pp. 366-379, 2004.

[53] C. Intanagonwiwat, R. Govindan, and D. Estrin, "Directed Diffusion: A Scalable and Robust Communication Paradigm for Sensor Networks," In Proceedings of the Sixth Annual International Conference on Mobile Computing and Networks (MobiCOM 2000), pp. 56-67, August 2000.

[54] C. Intanagonwiwat, R. Govindan, D. Estrin, J. Heidemann, F. Silva, "Directed Diffusion for Wireless Sensor Networking," IEEE/ACM Transaction Networking, 2003, vol.11, no.1, pp. 2-16, February 2003.

[55] S. Sharma, D. Kumar, and R. Kumar, "QOS-Based Routing Protocol in WSN," Advances in Wireless and Mobile Communications, Vol.1, pp. 51-57, 2008.

[56] D.B. Johnson and D.A. Maltz, "Dynamic Source Routing in Ad Hoc Wireless Networks," Mobile Computing, chapter 5, Kluwer Academic Publishers, pp. 153-181, 1996.

[57] C.E. Perkins and E.M. Royer, "Ad-Hoc On Demand Distance Vector Routing", In Proceedings Workshop Mobile Computing Systems and Applications, pp. 90-100, Feb. 1999.

[58] L. Zhao, B. Kan, Y. Xu, and X. Li, "FT-SPEED: A Fault Tolerant, Real-Time Routing Protocol for Wireless Sensor Networks," In proceedings of international conference on wireless communications, Networking and Mobile Computing, WiCom'07, pp. 2531-2534, Sept. 2007.

[59] E. Felemban, C. Lee, E. Ekici, R. Boder, and S. Vural, "Probabilistic QoS Guarantee in Reliability and Timeliness Domains in WirelessSensor Networks" In Proceedings of IEEE INFOCOM, vol. 4, March 13-17, pp. 2646-2657, 2005.

[60] W.Ben Jaballah and N.Tabbane, "Multi path Multi SPEED Contention Window Adapter," International Journal of Computer Science and Network Security, vol.9, no.2, pp. 113-118, February 2009.

[61] S.Sanati, M. H. Yaghmaee, and A. Beheshti, "Energy Aware Multi-path and Multi-SPEED Routing Protocol in Wireless Sensor Networks," In Proceedings of the 14th International CSI Computer Conference (CSICC'09), pp. 640-645, 2009.

[62] K. Sohrabi, J. Gao, V. Ailawadhi, and G. J. Pottie, "Protocols for Self-Organization of a Wireless Sensor Network," IEEE Personal Communications, vol. 7, no. 5, pp. 16-27, Oct. 2000.

[63] Y. Wang, C. Tsai and H. Mao, "HMRP: Hierarchy-Based Multipath Routing Protocol for Wireless Sensor Networks," Tamkang Journal of Science and Engineering, vol. 9, no 3, pp. 255-264, 2006.

[64] S. Kim, S. Son, J. Stankovic, Y. Choi, "Data dissemination over wireless sensor networks," IEEE Communications Letters, vol. 8, no. 9, pp. 561-563, Sep. 2004.

[65] S. Kim, S. Son, J. Stankovic, S. Li, Y. Choi, "SAFE: A Data Dissemination Protocol for Periodic Updates in Sensor Network," ICDCS Workshops, pp. 228-234, 2003.

[66] S. Basagni, A. Carosi, E. Melachrinoudis, C. Petrioli, and Z. M. Wang, "Controlled sink mobility for prolonging wireless sensor networks lifetime," Wireless Networks, vol. 14, issue. 6, pp. 831-858, February 2007.

[67] A. Gonga, O. Landsiedel, and M. Johansson, "MobiSense: Power-Efficient Micro-Mobility in Wireless Sensor Networks," In Proceedings of the 7th IEEE International Conference on Distributed Computing in Sensor Systems, IEEE DCOSS'11, Barcelona, Spain, pp. 1-8, 2011.

[68] A. Karmouch and S. Hashish, " Deployment-based Solution for Prolonging Network Lifetime in Sensor Networks," In Proceedings of IFIP International Federation for Information Processing, vol. 264, pp. 8596, 2010.

[69] M.E. Keskin, I.K. Altinel, N. Aras, and C.Ersoy, "Lifetime Maximization in Wireless Sensor Networks Using a Mobile Sink with Nonzero Traveling Time," The Computer Journal, 54 (12), pp. 1987-1999, 2011.

[70] D. Turgut and L. Boloni, "Heuristic Approaches for Transmission Scheduling in Sensor Networks with Multiple Mobile Sinks," The Computer Journal, vol. 54, issue. 3, pp. 332-344, December 2009.

[71] J. Luo and J.P. Hubaux, "Joint mobility and routing for lifetime elongation in wireless sensor networks," In Proceedings of the 24th Annual IEEE International Conference on Computer Communications (INFOCOM2005), vol. 3, pp. 1735-1746, 2005.

[72] W. Wang, V. Srinivasan, and K.C. Chua, " Extending the lifetime of wireless sensor networks through mobile relays," IEEE/ACM Transactions on Networking, vol. 16, issue. 5, pp. 1108-1120, October 2008.

[73] H. Karl and A. Willig. Protocols and Architectures for Wireless Sensor Networks. John Wiley & Sons, Ltd, Chichester, UK, ISBN 9780470095126, April 2005.

[74] C.S. Park, K.-W. Lee, Y.S. Kim, and S.J. Ko, "A Route Maintaining Algorithm using Neighbor Table for Mobile Sinks," Wireless Networks Journal, Volume 15, Issue 4, pp. 541-551, May 2009.

[75] K. Hwang, D. Eom, "Adaptive Sink Mobility Management Scheme for Wireless Sensor Networks," Lecture Notes in Computer Science Volume 4159, Springer Verlag, pp. 478-487, 2006.

[76] H. S. Kim, T. F. Abdelzaher, and W. H. Kwon, "Minimum-energy asynchronous dissemination to mobile sinks in wireless sensor networks," In Proc. of Embedded Networked Sensor Systems (SenSys'03), pp. 193-204, 2003.

[77] H. Luo, F. Ye, J. Cheng, S. Lu, and L. Zhang, "TTDD: two-tier data dissemination in large-scale wireless sensor networks," Wireless Networks, vol. 11, issue 1-2, Kluwer Academic Publishers, pp. 161-175, January 2005.

[78] K. Hwang, J. In, and D. Eom, "Distributed dynamic shared tree for minimum energy data aggregation of multiple mobile sinks in wireless sensor networks," Lecture Notes in Computer Science Volume 3868, Springer Verlag, pp. 132-147, 2006.

[79] A. Chakrabarti, A, Sabharwal, and B. Aazhang, "Using predictable observer mobility for power efficient design of sensor networks," In Proceedings of the 2nd international conference on Information processing in sensor networks (IPSN'03), pp. 129-145, 2003.

[80] E. Ekici, Y. Gu, and D. Bozdag, "Mobility-Based Communication in Wireless Sensor Networks," IEEE Communications Magazine, vol. 44, issue. 7, pp. 56-62, September 2006.

[81] M. Khan, W. Gansterer, and G. Haring, "Congestion Avoidance and Energy Efficient Routing Protocol for Wireless Sensor Networks with a Mobile Sink," Journal of networks, vol. 2, no. 6, pp. 42-49, Dec. 2007.

[82] M. Khan, W. Gansterer and G. Haring, "In-network storage model for data persistence under congestion in wireless sensor network," In Proceedings of the First International Conference on Complex, Intelligent and Software Intensive Systems (CISIS'07), pp. 221-228, 2007.

[83] I. Chatzigiannakis, A. Kinalis, and S. Nikoletseas, "Efficient data propagation strategies in wireless sensor networks using a single mobile sink," Computer Communications, vol. 31, no. 5, pp. 896-914, March 2008.

[84] I. Chatzigiannakis, A. Kinalis, and S. Nikoletseas, "Sink Mobility Protocols For Data collection in Wireless Sensor Networks," In Proceedings of The International Workshop On Mobility Management and Wireless Access, MobiWac'06, pp.52-59, 2006.

[85] M.Vecchio, A.Viana, A.Ziviani, and R.Friedman, "DEEP: Density-based proactive data dissemination protocol for wireless sensor networks with uncontrolled sink mobility," Computer Communications, vol. 33, no. 8, pp. 929-939, May 2010.

[86] V. Drabkin, R. Friedman, G. Kliot, and M. Segal, "RAPID: reliable probabilistic dissemination in wireless ad hoc networks," In Proceedings of the 26th IEEE International Symposium on Reliable Distributed Systems (SRDS), 2007, pp. 13-22.

[87] A. Kinalis, S. Nikoletseas, D.Patroumpa, and J.Rolim, "Biased Sink Mobility with Adaptive Stop Times for Low Latency Data Collection in Sensor Networks," GLOBECOM '09, pp.1-6, 2009.

[88] A. Kinalis, S. Nikoletseas, D. Patroumpa, and J. Rolim, "Biased Sink Mobility with Adaptive Stop Times for Low Latency Data Collection in Sensor Networks," Information Fusion, Elsevier, April 2012, Available online 30 April 2012. To appear.

[89] H. Xuan and S. Lee, "A Coordination-Based Data Dissemination Protocol for Wireless Sensor Networks," Proceedings of the Sensor Networks and Information Processing Conference, pp. 13-18, December 2004.

[90] US Naval Observatory (USNO) GPS Operations, http://tycho.usno.navy.mil/gps.html

[91] Y. Xu, J. Heidemannn, and D. Estrin, "Geography informed energy conservation for ad hoc routing," In Proceedings of the Seventh Annual ACM/IEEE International Conference on Mobile Computing and Networking (MobiCom 2001), Rome, Italy, pp. 70-84, July 2001.

[92] S.K. Yadav, "A Novel Survey on Various Grid Based Data Dissemination," International Journal of Advanced Research in Computer Science and Software Engineering, vol. 2, issue 7, pp. 153-158, July 2012.

[93] M.G. Lee and S. Lee, "Data Dissemination for Wireless Sensor Networks," In Proceedings of the 10th IEEE International Symposium on Object and Component-Oriented Real-Time Distributed Computing (ISORC'07), pp. 172-180, 2007.

[94] S. Hashish and A. Karmouch, "An adaptive rendezvous data dissemination for irregular sensor networks with multiple sinks," Computer Communications, vol. 33, issue 7, pp. 176-189, 2010.

[95] Q. Fang, J. Gao, and L.J. Guibas, "Locating and bypassing routing holes in sensor networks," In 23rd INFOCOM, vol. 4, pp. 2458-2468, 2004.

[96] S. Tang, "Traffic Flow Analysis of a Multi-hop Wireless Sensor Network Subject to Node Failure," International Journal of Communication Networks and Information Security, Vol. 3, No. 2, pp. 163-169, August 2011.

[97] A.M. Popescua, I.G. Tudorachea, B. Peng, A.H. Kempa, "Surveying Position Based Routing Protocols for Wireless Sensor and Ad-hoc Networks," International Journal of Communication Networks and Information Security, Vol. 4, No. 1, pp. 41-67, April 2012.

[98] N. Meghanathan, "Link Expiration Time and Minimum Distance Spanning Trees based Distributed Data Gathering Algorithms for Wireless Mobile Sensor Networks," International Journal of Communication Networks and Information Security, Vol. 4, No. 3, pp. 196-206, December 2012.

Mohamed Guerroumi (1), Al-Sakib Khan Pathan (2), Nadjib Badache (3), and Samira Moussaoui (1)

(1) Electronic and Computing Department, USTHB University, Algiers, Algeria

(2) Computer Science Department, International Islamic University Malaysia, Kuala Lumpur, Malaysia

(3) DTISI, CERIST Research Center, Algiers, Algeria

mguerroumi@usthb.dz, sakib@iium.edu.my, badache@mail.cerist.dz, and smoussaoui@usthb.dz

Table 1. Data Dissemination Protocols with Single Sink.

Protocol                          Dissemination scheme

LEACH [27], [42]   --Set up sensors into cluster nodes.
                   --Elect periodically a cluster head for each
                     cluster nodes.
                   --Sensor nodes send their data directly to their
                     cluster head using TDMA schedule
                   --Cluster head aggregates and sends the data to
                     the sink directly or via a super cluster head
                     using multi hop communication

CAEE [82]          --Set up sensors into cluster nodes.
                   --Choose some cluster heads as data collector
                     nodes (DC), each DC node creates its mini sink
                     (MS) nodes group.
                   --Sensor node sends the collected data to the
                     nearest DC node which stores it in one of the
                     buffer nodes of its MS nodes

SMPDC [83], [84]   --Sink broadcasts periodically a beacon message.
                     Each sensor node receives this message replays
                     by transmitting its data to the sink.
                   --Sink periodically broadcasts a beacon message
                     and indicates the depth of the trees or the data
                     dissemination path.

EAP [49], [50]     --Set up sensors into cluster nodes and elect
                     periodically a cluster head.
                   --Cluster head elected based on its weight
                     calculated based on its energy available and the
                     signal strength for the signal broadcasted by
                     the base station.
                   --Cluster heads constitute the data dissemination
                     path, next cluster head selected based on its
                     weight.

SEER [48]          - Event based.
                   --Sensor node sends its data to one next
                     neighbor.
                   --Next neighbor is selected based on the hop count
                     (number of hop for this neighbor to the sink)
                     and the available energy.
                   --Hop count and energy of each neighbor are
                     learned during the initialization phase when the
                     sink broadcast the path construction packet.

DCAST [87], [88]   --Network area is partitioned in equal square
                     regions.
                   --Sensor node sends its storage data in passive
                     manner.

DD [53], [54]      --Sink broadcasts its interest to establish the
                     data dissemination path.
                   --The node that has the interest sends back an
                     event to the sink.
                   --The sink reinforces the path from which the
                     event is received
                   --The interested data are disseminated using the
                     reinforced path.

PEGASIS [47]       --Create a communication chain.
                   --Sensor node sends its data to its nearest
                     neighbor.
                   --Only one node (chain head) sends data to the
                     sink.
                   --Chain head selected randomly and equitably
                     rotated among the nodes of the chain.

DEEP [85]          --Sensed data is disseminated based on density
                     sensitive probabilistic forwarding with
                     deterministic corrective measures

TEEN [45]          --Three communication levels, sensor node sends
                     its data to its cluster head, cluster head sends
                     the data to the sink via a second cluster head.

APTEEN [46]        --Use the same model as TEEN.
                   --Support both periodic data collection and
                     time-critical situations.

Protocol           Number of   Sink type
                     sink      Static /
                   Single /     Mobile
                   Multiple

LEACH [27], [42]      One       Static

CAEE [82]             One       Mobile

SMPDC [83], [84]      One       Mobile

EAP [49], [50]        One       Static

SEER [48]             One       Static

DCAST [87], [88]      One       Mobile

DD [53], [54]         One       Static

PEGASIS [47]          One       Static

DEEP [85]             One       Mobile

TEEN [45]             One       Static

APTEEN [46]           One       Static

Protocol                Sink mobility scheme

LEACH [27], [42]                  /

CAEE [82]          --The sink moves along the
                     periphery of the sensor
                     field.
                   --In the first trip, the
                     sink selects the data
                     collector nodes
                   --The mobile sink periodically
                     visits each MS and
                     retrieves the stored data.

SMPDC [83], [84]   --Three mobility schemes:
                   --Sink moves randomly
                     towards a chosen
                     direction with constant
                     speed.
                   --Sink visits only the
                     centre nodes, the next
                     centre node chosen
                     randomly or,
                   --Based on The visiting
                     frequency of the region
                     and the number of
                     sensor nodes in the
                     region.

EAP [49], [50]                    /

SEER [48]                         /

DCAST [87], [88]   --Two mobility schemes:
                   --sink visits regions from
                     left to right and vice
                     versa in deterministic
                     walk.
                   --sink selects The less
                     frequently visited
                     regions among the
                     adjacent regions.

DD [53], [54]                     /

PEGASIS [47]                      /

DEEP [85]          --sink moves towards any
                     destination at any time
                     and in any way.

TEEN [45]                         /

APTEEN [46]                       /

Table 2. Data dissemination protocols with multiple sinks.

Protocol                      Dissemination scheme

SPEED [48]        --Paths are built using least cost
                    algorithms.
                  --Next hop is selected based on the data
                    transmission speed and miss ratio.
                  --If the required node is not found, the
                    message has to be sent back to the
                    source nodes

MMSPEED [59]      --Each message is associated with delivery
                    deadline.
                  --provides multiple delivery speed options
                  --Uses multiple paths to transmit data.

CODE [89]         --The sensor network field is divided into
                    grids.
                  --Source node sends data announcement
                    message.
                  --The interested sink sends query to the
                    source node and creates the data
                    dissemination path.
                  --Source node sends the data to the sink
                    using the data dissemination path.

SAR [62]          --Multi Hop, Trees are constructed either
                    from node to sink or sink to node.
                  --Data dissemination path is chosen based
                    on the energy resource and a weighted
                    QoS metric

TTDD [77]         --A virtual grid has to be created at any
                    new sensed data by the source node.
                  --Source node sends a data announcement
                    message to four selected neighbors.
                  --The interested sink builds the two-tier
                    grid structure and sink broadcasts its
                    query.
                  --When the query reaches the source, the
                    data will be forwarded to the sink using
                    the data dissemination path created
                    during the query forward.

HMPR [63]         --Muli hop hierarchical tree is constructed,
                  --Sensor node keeps only the nearest next
                    hop to its sink.
                  --Multipath data forwarding path is used
                    in case of several next hop nodes.

PDDD [93]         --When sink sends its query, a Totally
                    Ordered Graph (TOG) using pseudodistance
                    is created.
                  --Sensor node uses this TOG to
                    disseminate the requested data.

SAFE [64], [65]   --Sink node initiates the data
                    dissemination path on demand by
                    sending its data query.
                  --Each sensor node which has the
                    requested data replies to the sink.
                  --Sink node chooses the best data
                    dissemination path.

TRDD [94]         --A query region has to be created on the
                    network perimeter which contains the
                    border nodes and their one-hop
                    neighbors.
                  --The interior nodes send their sensed data
                    toward the query region.

Protocol          Number of   Sink type     Sink mobility scheme
                    sink      Static /
                  Single /     Mobile
                  Multiple

SPEED [48]        Multiple     Static                /

MMSPEED [59]      Multiple     Static                /

CODE [89]         Multiple     Mobile     --Sinks move randomly.
                                          --Sink checks
                                            periodically its
                                            position.
                                          --If sink changes its
                                            position, the previous
                                            data dissemination
                                            path has to be removed
                                            and new one has to be
                                            set up.

SAR [62]          Multiple     Static                /

TTDD [77]         Multiple     Mobile     --Sinks move randomly.

HMPR [63]         Multiple     Static                /

PDDD [93]         Multiple     Mobile     --Sink sends periodical
                                            heartbeat messages to
                                            permit its direct
                                            neighbors detect its
                                            mobility

SAFE [64], [65]   Multiple     Static                /

TRDD [94]         Multiple     Mobile     --Sinks move within
                                            the query region.

Figure 15. Data dissemination protocols and application
needs.

                PBDS     QBDS    EBDS    HBDS

* APTEEN                                  9%
* TRDD

* LEACH
* CAEE          9%

* DD                      43%
* SAFE
* SMPDC
* DEEP]
* DCAST
* CODE
* TTDD
* EGDD
* PDDD

* TEEN                             39%
* PEGASIS
* SEER
* EAP
* SPEED
* MMSPEED
* SAR
* HMRP

Note: Table made from pie chart.
COPYRIGHT 2013 Kohat University of Science and Technology
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 2013 Gale, Cengage Learning. All rights reserved.

Article Details
Printer friendly Cite/link Email Feedback
Author:Guerroumi, Mohamed; Pathan, Sakib Khan Al-; Badache, Nadjib; Moussaoui, Samira
Publication:International Journal of Communication Networks and Information Security (IJCNIS)
Article Type:Report
Date:Dec 1, 2013
Words:16822
Previous Article:Link quality and MAC-overhead aware predictive preemptive multipath routing protocol for mobile ad hoc networks.
Next Article:QoS-VNS-CS: QoS constraints core selection algorithm based on variable neighborhood search algorithm.
Topics:

Terms of use | Privacy policy | Copyright © 2019 Farlex, Inc. | Feedback | For webmasters