Crash Faults Identification in Wireless Sensor Networks
西柏坡村
Stefano Chessa1 and Paolo Santi2*
1 Dipartimento di Informatica, Univ. of Pisa, Italy
2 Istituto di Matematica Computazionale del CNR, Pisa, Italy.
Abstract – In this paper we consider the problem of identifying faulty (crashed) nodes in a wireless nsor network. This problem is of fundamental importance in tho applicative scenarios of wireless nsor networks in which battery replacement is feasible.The diagnostic information gathered by operational nsors can be ud by an external operator for the sake of network reconfiguration and/or repair, thus extending network lifetime. A fault diagnosis protocol specifically designed for wireless nsor networks is introduced and analyzed. The protocol is proved to be optimal and energy efficient under certain assumptions.
Keywords – wireless networks, nsor networks, network lifetime, fault diagnosis, energy-efficiency.
1. Introduction
Wireless nsor networks are receiving increasing interest due to their ability of monitoring a wide variety of different environments: they can be ud to monitor remote geographical regions as well as industrial plants, office buildings or toxic urban locations. Examples of possible scenarios where wireless nsor networks can be ud are described in [9,11,14,15,16,17,18,19].
A wireless nsor network is usually compod by a large number of nsors (depending on the application, hundreds or thousands) equipped with nsing, computation, and wireless communication devices, which coordinate themlves in a distributed fashion in order to collect information on the surrounding environment. During the network lifetime, the information collected by the nsors is periodically transmitted to sink nodes, which could be either mobile or * Part of this work was done when the author was with the Istituto di Elaborazione dell’Informazione del CNR, Pisa, Italy, and with the Dept. of Elec. and Comp. Eng., Georgia Institute of Technology, Atlanta GA.
1
ba stations. Sink nodes are ud by external operators to retrieve the information gathered by the network.
Since the nsor nodes have extremely compact form factors and are wireless, they are highly ener
gy constrained. Furthermore, replenishing energy by replacing batteries is unfeasible in many applicative scenarios. Hence, one of the key challenges in unlocking the potential of such data gathering nsor networks is conrving energy so as to maximize their post-deployment active lifetime.
It has been obrved that any effort aimed at extending network lifetime must be two-fold: on one hand, the nsor itlf must be made as energy efficient as possible; on the other hand, the collaborative strategy which coordinates nodes in the nsing task must be energy efficient. It has been shown that dynamic voltage scaling techniques [12] and puld battery discharge schemes [8] can be ud to increa the nsor lifetime, while many energy efficient cooperative strategies have been recently propod [5,6,9,10,11,14,20]. However, in the applicative scenarios in which battery replacement is unfeasible, the network lifetime cannot be extended beyond a certain time, which depends mainly on the initial capacity of the batteries equipping the nodes. An upper bound on the lifetime of a wireless nsor network that collects data from a specified region using a certain number of energy-constrained nodes has been recently derived in [2].
While increasing attention has been paid to the problem of extending network lifetime when batteries cannot be replaced, few have been done in the ca in which battery replacement is feasible. As an 英语的动词
example of the latter scenario, consider the following application in which a wireless nsor network is ud to help rangers to monitor a vast natural park. A number of nsors, which have limited energy supplies and are placed in strategic positions (e.g. on the top of a hill or in locations with wide view), are ud to collect information about the prence of animals, tourists, fire, flooding and so on. Sensors periodically exchange and disminate this
2
information throughout the network, so that a “global view” of the park is available at each node. Rangers, who are equipped with mobile stations, move around the park for control and maintenance. Mobile stations are connected to the nsor network through the nearest nsor. This way, rangers can be alerted of abnormal events, and they can quickly intervene where needed. If nsors maintenance is out of the picture, eventually they deplete their batteries and crash. As a conquence, the number of non-operating nsors in the system increas until the network gets disconnected, thus impairing its functionality. In order to prevent such an event, rangers can replace the depleted batteries, thus allowing the network to operate for a virtually unlimited time.
Obrve that since nsors in general are subject to uneven workloads, they are expected to deplet
e their batteries at different times. Hence, the problem of identifying the crashed nsors at a given time aris. Returning to the example above, nsors could provide diagnostic information (i.e., the status – operational/crashed – of each node) along with nsor data, thus enabling the rangers to maintain network functionality by replacing the depleted batteries.
The problem of identifying faulty (crashed) nodes in a distributed system has been extensively studied in the literature [1,3,13] in the framework of wired computer networks bad on the one-to-one communication paradigm, where energy consumption is not an issue. Contrary to the ca of wired networks, the natural communication paradigm in wireless networks is one-to-many (when a node transmits, all the nsors within its transmitting range may in principle receive the message). Furthermore, the protocol itlf should be as energy efficient as possible (even if batteries can be replaced, their replacement is an expensive and possibly difficult operation; hence, energy consumption is still an issue). For this reason, the protocols prented in [1,3,13] result either unfeasible or extremely energy consuming when applied to wireless nsor networks.
3中秋节的别称
In this paper, we introduce a distributed fault diagnosis protocol specifically designed for wireless
nsor networks. We are not aware of any other diagnosis protocol designed for wireless nsor networks, except for the protocol propod in [7]. However, the focus in [7] is on soft faults (a soft faulted node continues to operate, although with altered behavior) in the more general framework of wireless ad hoc networks.
The protocol introduced in this paper, which is called WSNDiag, is proved to provide correct diagnos if the number of crashed nodes is below a certain threshold, namely, the network connectivity1. Furthermore, it is proved that under the assumptions to be stated in Section 2 WSNDiag exchanges the minimal number of bits needed to perform diagnosis. To a certain extent, this proves that the protocol is optimal from the point of view of energy consumption.
2. System model
北上的列车
The system is compod by n nsors, also called nodes, which communicate via radio transceivers. Nodes are homogeneous and equipped with a limited energy supply. Any node in the system is a potential sink, i.e. it can be ud by an external operator to access the information gathered by the network. Sensors can be in one of two states: faulty or fault-free. Faulty nodes are unable to communicate with the rest of the system, either due to a crash or to battery depletion. This means that faults are permanent, i.e. nodes remain faulty until they are repaired and/or replaced.
We assume that all transmissions from any node u are omni-directional. Thus, any message nt by u can be received by any node in its neighborhood, i.e. within its transmitting range. The neighborhood of node u is denoted N(u). The topology of the system can be described as a di-
1The connectivity of a graph is the minimum number of nodes who removal results in a disconnected or trivial graph [4].
4
graph G=(V,E), called the communication graph, where V is the t of nodes and E is the t of edges connecting nodes. Given any u,v∈V, directed edge (u,v)∈E if and only if v∈N(u). The t of faulty nsors is denoted F, with |F|=f≥0.
In this paper we assume the following:
A1. each node has a unique identifier which can be encoded using log n bits.
A2. each node knows its identifier and the identifiers of its neighbors. This is the only knowledge of the system graph that a node has.巴沙鱼图片
A3. no new faults occur during the execution of the diagnosis protocol.
A4. the network topology is static during the execution of the diagnosis protocol.
A5. there exists a link-level protocol providing the following rvices:
A5.1.a MAC protocol is executed to solve contentions.
A5.2.the protocol provides two communication primitives: 1-hop Broadcast (1_hB) and Selective Send (SS). Primitive 1_hB(m) delivers message m to all the nodes in the惠美饺子
nder’s neighborhood, while primitive SS(v,m) delivers m only to neighbor node
v.
A6. the communication graph is connected and symmetric.
Assumption A1 is typical in distributed systems and derves no further comment. Assumption A2 is crucial in proving the lower bound stated in Section 3 and is realistic in this tting. In fact, even if the network topology does not vary with time, storing information regarding the communication graph at the nodes would imply non-trivial processing power and memory capacity of the nsors, which, in general, cannot be guaranteed. Even if this would be the ca, managing large data structures would be extremely energy consuming.
5
Assumption A3 is realistic whenever the mean time between failures of the network is relatively long as compared to the protocol execution time, as it is the ca in most situations.
Obrve that assumption A4 does not mean that the network is static, rather that its topology does not change during diagnosis: nsors are allowed to migrate, but they cannot migrate out of their neighbors’ transmitting ranges. Although rather stringent, this assumption is realistic in many applicative scenarios, such as the natural park monitoring described in the Introduction. Assumption A5 ensures that the messages nt by a node are correctly received by neighbor nodes.
Assumption A6 ensures that whenever u∈N(v), we have that v∈N(u). Nodes u and v are referred to as adjacent. This assumption implies that the communication graph is undirected. A nsor network and the corresponding communication graph are reported in Figure 1.
The assumptions stated above, although imposing constraints on the nature of the system, allow the derivation of a non-trivial lower bound on the total number of bits to be exchanged for the purpo of diagnosis, and the definition of the optimal diagnosis protocol prented in Section 4. However, some of the assumptions can be weakened or relead at the expen of straightforward modifica
tions of the diagnosis protocol, as discusd in Section 6.
a)b)
尊师故事6少先队手抄报内容
