Fault Tolerance In The Internet Of Things

The world has seen a great increase in natural disasters with time and this has not only affected the human life but also the cyber life. And this affect is more deadly when it hits a urban and remote areas. This calls for the need of Fault Tolerance and the way to overcome this issue is to either use wireless wireless sensor networks (WSNs) for collecting data and machine learning which foreseeing and predicting natural disasters before they occur. Lately an innovation has been made SENDI (System for detecting and forecasting Natural Disasters based on IoT). SENDI which is a fault-tolerant system based on IoT, ML and WSN for the detection and forecasting of natural disasters and the issuing of alerts. The system was modeled by means of ns-3 and data collected by a real-world WSN installed in the town of São Carlos - Brazil, which carries out the data collection from rivers in the region. It operates by adding intelligence to the nodes to carry out the data distribution and forecasting, even in extreme situations. A case study is also included for flash flood forecasting and this makes use of the ns-3 SENDI model and data collected by WSN.

Keywords: WSN, ML, SENDI, Fault ToleranceI.

In recent years’ damage to the environment in urban areas have led to certain changes in the local climate. Due to these changes the number and intensity of natural disasters such as landslides, flash floods and fires have increased so much that they have now become a global problem. As recently as 2016, 330 natural disasters were documented and about 96. 5 million people were affected, causing an estimated financial loss of US$ 118. 6 billion around the world. In the case of floods, every year about more than 102 million people are directly or indirectly affected and this number is constantly increasing. According to a survey, several applications areas for WSN, where we find energy systems, transport, health-care, gas, air, structures, and urban temperatures monitoring, among others.

Despite presenting some works that present forecasts solutions, and the forces of nature as being brutal and unpredictable, causing material damage and death of millions of people, leaving the lack of energy, food, water, and communication failures for the survivors. Regarding WSNs application in natural disasters, Rashid et al. [4] listed a series of related works to surveillance, where WSNs were used for detection of transportation and smuggling routes of radioactive material as a long-term solution to nuclear material-based terrorism, and tsunamis detection and response, where WSNs were used for the prediction of tsunamis and floods, using prediction based on web services and neural networks. In addition, WSNs were used as systems to mitigate tsunami and floods effects, most of which, like ours, used water pressure sensors for data One possible course of action is to follow the new trend and adopt sensor networks based on IP by using emerging standards, such as 6LoWPAN/IPv6, which allow the most diverse objects in IoT to be connected with each other [5].

The use of these standards makes communication possible between the WSN, the nodes of a WSN and the Internet, thus enabling the sensor nodes to be viewed as smart objects that form a link between the physical world and online systems [6]. In the management and forecasting of natural disasters, the data collected by means of the sensors can be analyzed together with data available on the Internet such as satellite-based forecasting or types of variables which the sensor nodes are unable to obtain. Moreover, a scenario in which the sensor nodes are connected to the Internet, makes it possible to use technologies such as Cloud computing and social networks to help in the forecasting and issue online warnings. In addition, this scenario also allows the sensor nodes to communicate with nearby devices which share the same technology and propagate the information and forecasts. The need for fault tolerance approaches is another important issue when dealing with natural disasters.

This issue is partially addressed, where a Cognitive Radio Based Internet Access Framework for Disaster Response for inhospitable environments is presented. |The term Cognitive for devices is used that can modify their parameters on-the-fly, being able to adapt to environment changes and, thus, restore the communication between partially damaged networks. In the city of São Carlos, Brazil, there is a WSN deployed by the Institute of Mathematical Sciences and Computing (ICMC), University of São Paulo (USP), called REDE (WSN for monitoring urban rivers), the purpose of which is to monitor, analyze the data, detect flash floods, and present online information about rivers in urban areas. The REDE system collects data on the rivers and rainfall in the area and uses two different kinds of technology to communicate, ZigBee and 3G. However, the REDE system does not envisage that this data will be used to forecast flash floods or that this information which is collected in the real world can be integrated with other information available in online environments.

Additionally, the REDE system does not have any fault-tolerance mechanism, which is necessary in hazardous environments where the loss of nodes and communication is a common phenomenon. With regard to this paper, the IoT paradigm allows the REDE system to be extended by introducing the concept of integration between devices and remote services. These services help in the distribution of information and data, as well as in the forecasting and decision-making and act in the real world by sensing and providing information about the monitored environment. This means that the nodes of a WSN (such as the REDE system) can be seen as smart objects that have to meet certain requirements such as the following: (a) diversity in data collection from different types of sensors; (b) the possible use of these data via the Internet and (c) collective intelligence among all/some nodes, based on information about multiple nodes and not only about the situation and data of a single one.

Thus, it can be said that the management and forecasting of natural disasters takes place in scenarios where the integration between the real and the online worlds occurs spontaneously. In light of this, the aim of this paper is to propose, design, evaluate and assess the performance of a fault-tolerance system for forecasting and issuing warnings of natural disasters called SENDI (System for detecting and forecasting Natural Disasters based on IoT), with a case study about flash flooding. SENDI was implemented by using ns-3, emerging IoT standards and Machine Learning (ML), and by applying real data collected by means of the REDE system. The system is designed for natural disasters in a hostile environment, which is susceptible to failures of communication and loss of nodes (either by physical destruction or a lack of energy).

To this end, we also proposed and evaluated a node cooperation mechanism for cluster formation, based on the energy remaining in each node. This mechanism is designed to make better use of the remaining energy in each node, by extending the lifespan of the system and ensuring it is able to keep making forecasts, even in the case of node losses. The forecasting scheme of SENDI relied on embedded models created by means of ML techniques and real data collected by the REDE system within the nodes.