Performance Measurement in Wireless Sensor Networks



The use of Wireless Sensor Networks (WSNs) has opened the possibility of a new set of applications that are having a growing impact in personal and business activities. However, most of its application scenarios have been restricted to non-critical environments, with WSN being operated with no controlled performance. The aim to extend the flexibility and unique characteristics of these networks to a broader set of applications and scenarios, such as industrial and health care, poses new challenges that must be met with a new approach. In such environments, WSNs may have significant benefits over traditional networks, such as enabling a deeper control, lowering deployment and maintenance costs, and by offering simple reconfiguration and adaptation to changing business models.
To enable WSNs with controlled performance the first step is to be able to characterize and describe the requirements that the network needs to fulfil. The second step is to be able to translate those requirements into effective metrics. The metrics to be used must be adapted to the unique characteristics of WSNs, taking into account its processing, energy and storage restrictions. The next step is to monitor those metrics, and allow for their debugging when necessary, a procedure that involves their collection to a central base station where further treatment, with better resources, is possible and where an effective network monitoring can be achieved. The last step completes the cycle and corresponds to the ability to dynamically act in the network, based on the metrics received, either automatically (by each node or by a central monitoring tool connected to the sink) or through the Network Manager.
In this thesis, the performance control life cycle of a WSN is addressed, especially considering the performance needed in industrial facilities, one of the most demanding scenarios for these networks, requiring not only strict performance boundaries but also real-time monitoring of the network. Valuable insights of these environments were possible through the participation in project FP7 GINSENG. First, a new classification of WSN application scenarios, that also includes critical environments, and a proposal of a new taxonomic tree of WSNs Quality of Sensing (QoSensing) requirements, including WSNs with performance control, is presented. The objective is to characterize WSNs needs, both in the information as in the network planes, and to create a reference classification that lays the foundations for the creation of effective metrics that permit the evaluation and verification of each requirement. The taxonomy was applied to different types of GINSENG scenarios and also to well-known types of applications found in the literature for validation. Having as reference the taxonomy created, and the specificities of WSNs, a study of different types of metrics is presented and their characteristics and applicability to WSNs discussed. In this context, collective metrics are introduced as a useful type of metric to address the evaluation of the global network QoSensing, while using least resources than other types of metrics and hiding the normal fluctuation of values in networks subject to many hazards. Simulations showed that collective metrics are an efficient alternative to individual or aggregated metrics, in the assessing of the global QoSensing of a WSN. Next, the Network performance branch of the previously proposed taxonomy is analysed and a general set of metrics, adapted to each of the phases of the life cycle of a WSN, is proposed to address it. A reduced set of metrics specifically targeted to WSNs in industrial environments, focusing collective metrics for the operation phase of the network, is also proposed. The control and maintenance of levels of performance, based on a continued evaluation of specific metrics and in the dynamic actuation in the network was also addressed, with the participation in the creation of a new protocol that deals with interferences. Finally, a new protocol that collects performance data from the network is proposed. By using data fusion, the protocol presents an effective way to monitor the global performance of the network, while guaranteeing that if some error or problem occurs, an alert is generated and immediately sent to sink. The evaluation of this protocol, made by simulation, showed a decrease in the energy spent and in the interference generated by the number of packets sent, while providing for a global knowledge of the overall performance of the network. The thesis also contributed to project GINSENG, namely in the classification of the project scenarios, according to the taxonomy proposed, and in the specification of the performance metrics to be used.


Monitoring, performance evaluation, performance metrics, quality of service, taxonomy, wireless sensor networks.


Performance in Wireless Sensor Networks

PhD Thesis

Performance Measurement in Wireless Sensor Networks, February 2016

Cited by

No citations found