Efficient resource management for fog computing

The number of Internet of Things (IoT) devices is growing rapidly due to advancements in sensor and wireless technology. As a result, the volume and variety of data generated by these devices are increasing dramatically. Traditional Cloud computing cannot process this data in an appropriately small time due to its limitations, such as high latency. Hence, Fog computing has evolved as an extension to Cloud computing to bring the computation closer to the end-devices in which data is generated. In Fog computing, failure of resources and sudden spikes of user service requests could occur frequently. These failures and sudden spikes of the requests should not affect the performance of the system. Thus, providing efficient resource management services to handle the sudden spikes of the requests to meet the requirements of future services is an essential task. Dynamic resource allocation of the system plays a vital role in managing the resources in Fog computing environments. The efficient allocation of resources depends on accurate resource monitoring and the availability of information. Researchers have proposed different techniques for advanced services, such as scheduling, fault tolerance and migration, by assuming that basic services, such as resource monitoring and discovery, are available. Some researchers have considered traditional Cloud and distributed computing approaches in Fog environments but these may not be suitable due to their limited resources and being battery-powered. To tackle this problem, this thesis proposes efficient resource monitoring, resource availability and resource allocation techniques to minimise resource usage and improve the performance of Fog-based applications. The thesis advances this field by making the following key contributions: 1. An extensive survey on resource management techniques, mainly resource monitoring, availability and resource allocation in the Fog computing environments so as to gain a better understanding of the studied methods and techniques, as well as the limitations of existing algorithms. 2. An investigation of the impact of traditional monitoring approaches on the efficiency of the resource usage, finding that the existing resource monitoring techniques are not suitable for Fog computing environments. An efficient resource monitoring service for Fog computing environments that minimises the resource utilisation of Fog devices during resource monitoring. 3. A generic stochastic model for resource availability that enables the modelling of Fog devices in the Fog network which can predict the number of resources available at a particular location and duration of time. This helps to decrease the number of Service Level Agreement (SLA) violations in terms of availability. 4. A resource selection model for minimal disruption rates during the service execution in Fog computing environments which can increase the number of satisfied requests. 5. A framework for service deployment in Fog Computing environments, which maximises the number of satisfied requests, and meets the requirements of the on-demand application requests.