You are on page 1of 4

2009 International Conference on Emerging Technologies

Service Guarantees in Wireless Sensor Networks


Hasan Tahir and Muhammad Younus Javed
College of Electrical and Mechanical Engineering National University of Sciences and Technology (NUST) Rawalpindi, Pakistan hasanmailbox@yahoo.co.uk, myjaved@ceme.edu.pk
AbstractWireless Sensor Networks possess unique characteristics because of their hostile deployment environment and their resource starved nature. Wireless Sensor Networks are now being deployed in a number of complex environments like battle fields, agriculture, hospitals and many more. Almost all the environments where wireless sensor networks are deployed need to sense events in their surroundings and report them to the collection points in an efficient and timely manner. Both of these requirements require a Quality of Service (QoS) based view of the entire network. In this paper we have explained why sensor networks need QoS and what are the QoS concerns in wireless sensor networks. We have also studied a number of architectures that contribute to the QoS requirements. A tradeoff regarding QoS has also been presented to explain the gains and losses of any architecture that tries to address the quality of service in wireless sensor networks. Keywords - quality of service; QoS concerns; service guarantee

TABLE I.

QUALITY OF SERVICE FACTORS Technical Factors Reliability Expandability Effectiveness Maintainability of System Transmission Quality

Human Factors Stability of Quality Service Waiting Times Fault Clearance Times Stability of Operations Availability of Network

I.

INTRODUCTION

Over the last few years Wireless Sensor Networks (WSN) have gained much importance keeping in view their benefits and low cost. Currently WSN are providing services in the areas of battle field analysis, health monitoring, and agricultural / environmental monitoring. Since WSN were basically not designed to support large amount of traffic that requires battery and processing power, due to this reason Quality of Service (QoS) in WSN remains largely unexplored. WSNs are very different as compared to conventional wireless networks. Some of the features that distinguish WSN from conventional networks are the limited battery power, limited processing capabilities, limited memory processing capabilities. In the near future we can expect applications of WSN in domains that require QoS guarantees. We have analyzed the QoS requirements of WSN and also discussed various protocols/ architectures related to QoS in WSN. In the end we have tried to perform a trade-off between QoS and the issue of a resource demand storm. II. QUALITY OF SERVICE Quality of Service is a largely misunderstood term because it offers different meanings from different perspectives. QoS has been analyzed from two different perspectives, namely human and technical [1]. Human factors are those that form the basis of our request/ expectations from the network. On the other hand technical factors are purely based on the network concerns. Both human and technical perspectives are illustrated in the following table [1].

The two factors on which QoS is based can be better illustrated in a QoS model in which we have two entities i.e. applications/users and network. The user/application is not concerned with how the network provides the quality service. The QoS requirements are provided to the network which in turn provides services by meeting the user requirements. As the diagram illustrates the requirements can be advertised using the application. After collecting the requirements the network provides services based on the expected requirements. The provision of QoS cannot be done without infrastructure support. Infrastructure support means the use of protocols specifically designed for QoS and employment of network technology that is both efficient and accurate.
Applications/Users

Services

Requirements

Network QoS Infrastructure Figure I. An ideal QoS model.

As Fig. I effectively highlight, an ideal QoS model is one that maps user requirements to the QoS services provided by the system. QoS can be defined as the ability to provide different priority to different applications, users, data flows, or to guarantee a certain level of performance to a data flow. In WSN there has been an evident shift in the type of traffic that the network accommodates. In the initial years WSN were only sensing locations and other related parameters but now they are handling data which is diverse in nature. As the

978-1-4244-5632-1/09/$26.00 2009 IEEE

433

networks get physically deployed they are now handling realtime traffic that requires certain guarantees from the network. For example in battle field monitoring the detection of an attack event has to be delivered to its appropriate destination as quick as possible with enough details so that a counter attack can be launched. As discussed earlier QoS becomes even more important in situations where the criticality of data/detected event is very high. III. QoS IN CONVENTIONAL NETWORKS In conventional networks many mechanisms for QoS provisioning are used. One of the earliest and easiest technique was to use over provisioning of resources and traffic engineering. When we over provision resources our basic focus is on providing excess resources to the network so there is no pileup of traffic which results in better services for multimedia hungry applications. It is worth pointing out here that in such a case the situation can get out of hand since there is no control over who has access to the privileged services (there are no service classes). While discussing traffic engineering there are two approaches of providing QoS i.e. reservation based approach and reservation less approach. In reservation based approach we employ bandwidth management policies to assign resources based on an applications QoS request. Asynchronous Transfer Mode (ATM) uses resource reservation mechanisms to provide QoS. In reservation less approach there is no reservation of resources. This approach provides services by formulating classes of services, enforcing admission control using policy managers in combination with queuing techniques. An admission control mechanism is formulated to enforce if a node can access the network. The Policy managers make sure that once the nodes gain access they are provided the services they are eligible for. The class of service concept divides network traffic into different classes and provides class-dependent service to each packet depending on what class it belongs to [2] In accordance with the class of service, packets can be admitted and dropped from a queue that is being governed by a queuing mechanism. IV. QoS CONCERNS IN WSN Wireless sensor networks possess unique characteristics because of their network capabilities and their deployment complexities. QoS concerns in WSN are very similar to those of conventional wireless and wired networks. Below we highlight some of the concerns that make QoS implementations in WSN a complex task. A. Bandwidth Limitations WSN suffer severely from bandwidth limitations. The issue becomes even more critical in situations where there is limited battery and little processing capabilities. WSN have to handle real time and non real time traffic at the same time. If traffic classes are used to allocate resources to a particular flow then

the network may starve because there is not much bandwidth that can be shared. A better solution for sharing network load is to route traffic through different nodes. Making routing decisions can also be a cumbersome task because WSN are not designed for doing complex calculations. B. Data Redundancy Data redundancy is a major issue in WSN. Although redundancy does not affect the reliability of data it only causes a loss of energy which is already very limited in WSN. Data redundancy can be reduced by using data fusion and data aggregation. Using such mechanisms causes an increase in the QoS design. C. Energy Distribution To ensure the long life of a WSN energy consumption should be distributed over the entire network. Otherwise nodes tend to die out randomly due to energy depletion. This phenomenon becomes a black hole in the coverage of the terrain and in severe cases it can cause segmentation of the network. Routing protocols [3] have been proposed that distribute the energy consumption over the entire network instead of putting stress on a few nodes. D. Accuracy While considering sensing operations in WSN the level of accuracy is influenced by the amount of received data. Larger amounts of data means that redundant and erroneous data can be excluded and the sampling set is fairly large. E. Accommodating Heterogeneous Sensors A truly intelligent WSN should be able to accommodate heterogeneous sensors that have been designed to handle different types of traffic. For example some sensors may have the ability of sensing seismic vibrations while others can only sense wind speeds/ atmospheric pressure. As two or more heterogeneous sensors coexist, separate classes of traffic may have to be formed to accommodate data that is real time. For example while sensing earth quakes data about vibrations can be classified under real time while data regarding the wind speed may be classified under the normal data category. F. Buffer Size Limits WSN are equipped with very limited sized buffers so that the size and complexity of the device remains low. Having a small sized buffer also means less memory for maintaining the temporary memory. Having a small sized buffer also has its drawbacks. A limited sized buffer can greatly affect the routing procedures. In multihop routing incoming packets need to be stored if they have to be transmitted to the next hop. As discussed earlier data fusion and data aggregation require multiple packets which need to be accommodated no matter how small the buffer space is. Thus buffer size limitations can greatly effect how QoS requirements are met. G. QoS Based Queuing To provide QoS support packets may have to be queued while high priority data is being serviced. It is worth pointing out here that while designing a QoS mechanism we need to ensure

434

that a balance is kept between the real time traffic and the regular traffic. By efficient utilization of links we can avoid problems like memory overflow and node starvation. Fig. II effectively illustrates the process of queuing and queue processing in wireless sensor networks.

OSI layer stack with reduced complexity/ functionality. For an application to make efficient use of the protocols the application may have to control the low level features directly even though the services of middleware have also been employed. The middleware will control data flows based on QoS requirements/ specifications.
Application Application Middleware Routing MAC Physical Routing MAC Physical Middleware Routing MAC Physical Application

(a) Figure II. An ideal queuing model for a sensor node [4].

(b)

(c)

H. Timeliness Timeliness is a property which can be very critical in a WSN. If the network is sensing critical data then timely delivery of the data is very important. Data collected properly becomes useless if it is not delivered in time in the correct format. For example in the domain of battle field monitoring a sensor may correctly sense the presence of an enemy but if a warning is not sent to the user within a defined time limit then the data may be of no use to the user because by the time the message is received it may be to late to launch a counter attack. I. Network Centric Reliability Operations In WSN sensing operations using a single node is next to irrelevant because of the small coverage area and limited features of a single node. While providing QoS it should be ensured that coverage is upon the whole network. A single node in WSN cannot convey enough to give meaningful results. Hence reliability is an issue which should span the entire network rather than only a few nodes. Network reliability can be viewed from the aspect of loss and corruption rate. Proportion of total data which does not arrive as sent, e.g. network error rate [5] V. ARCHITECTURES FOR QoS SUPPORT As highlighted previously WSN suffer from severe shortage of resources, which forces us to reconsider the implementation of the conventional OSI protocol stack on a WSN. Hence there is need for proposing new protocols that provide QoS support while managing sensors and network resources. Many of the proposed architectures have been derived/ conceived from our regular OSI protocol stack. Fig. III shows three different architectures that promise QoS support for WSN. A major contribution of these architectures is that they are based on the

Figure III. Different architectures that support QoS while managing sensors and network resources at the same time. (a) Architecture providing low layer application support. (b) Architecture proving middleware support to the application. (c) Architecture that provides cross-layer support by managing low layer protocol functions. [6]

Cross Layer Architecture Yanjun et al. have discussed cross layer architectures that provide interconnectivity between layers in a protocol stack. The architecture described in Fig. IV is based on the fact that parameters in the lower layer are reported to adjacent higher layers. This provides a mechanism in which higher layers can coordinate with lower layers. For example if the application has QoS requirements like a bound on end-to-end delay and service classes. These requirements will be communicated and priority scheduling will be used on the MAC layer and the routing layer. The physical layer will further calculate the amount of energy consumed in transmission. Another metric that will also be calculated will be the Link Quality Indication (LQI). The LQI reflects the reliability and robustness of the connection. When this is communicated to the MAC layer it will influence the active/idle/sleep states of the sensors. The whole protocol hence provides a mechanism of communicating parameters and merging of relevant protocol functions into one component.

435

Application layer QoS requirements

CONCLUSION

Network layer

MAC layer Tx/Rx energy, processing energy, etc

LQI Physical layer Figure IV.

Cross Layer Architecture for WSN [7]

Wireless sensor networks are unique in nature due to limitations like limited power and lack of computation capabilities. Even though these limitations exist WSN are being deployed in very challenging environments where the users need some sort of guarantees from the network. These guarantees can be in the form of bandwidth requirements, energy distribution, and removal of data redundancy. Many protocols/ architectures have been proposed to handle QoS issues in WSN. In this paper we have analyzed WSN from QoS perspective. Keeping in view the vastness of the WSN operations there is urgent need for developing mechanisms through which we can provide service guarantees. Most of the architectures have the ability of accommodating user demand but still there is a major issue that the user demands increase to such an extent that in the end the protocols end up consuming the entire system. REFERENCES
[1] Kauko Rahko, Stefan Herzberg, and Timo Rahko. Grade of service and human factors. In 6th International Symposium on Human Factors in Telecommunications, Stockholm, 1972. Peuhkuri M, IP quality of service, Helsinki University of Technology, Laboratory of Telecommunications Technology, 1999. Jae-Hwan and Leandros Tassiulas Chang, Maximum lifetime routing in wireless sensor network pp. 1-5. Mohamed Younis, Kemal Akkaya, Mohamed Eltoweissy, Ashraf Wadaa, On handling QoS traffic in wireless sensor networks pp. 1-6 Dan Chalmers and Morris Sloman, Service of quality of service in mobile computing environments Imperial College London Research Report 1999. pp. 1-27. Wendi B. Heinzelman, Amy L. Murphy and Mark A. Perillo. Enabling data- and event-centric communications. In Wireless Sensor Networks A system Perspective. Artech House, 2005. Yanjun Li, Chung Shue Chen Ye-Qiong Song, Zhi Wang, Real-time Qos support in wireless sensor networks: a survey. Paul Ferguson and Geoff Huston, Quality of Service: delivering QoS on the Internet and in corporate networks. John Wiley & Sons, 1998.

VI.

QoS TRADEOFFS

The architectures/protocols discussed and numerous others have been able to address the issue of QoS in WSN. Although every protocol has its pros and cons we have been able to come across a fact that still needs to be addressed. All QoS protocols suffer from a phenomenon which is a corollary of Moores law. Corollary of Moores Law: As you increase the capacity of any system to accommodate user demand, user demand will increase to consume system capacity [8, p. 10]. It is very obvious that no matter which protocol we use to provide QoS, our network runs out of power or computation capability. Suppose we are able to design a protocol which addresses all the issues of QoS; it is still a bitter fact that as user demands increase there is a sharp decrease in the network performance and network life especially in WSN. What is actually needed is a protocol that addresses issues related to QoS and limits the resource demand storm created by users.

[2] [3] [4] [5]

[6]

[7] [8]

436

You might also like