SmartFogLB: balanceamento de carga na computação em névoa

• Ano de defesa: 2020
• Autor(a) principal: Pereira, Éder Paulo
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Federal de Santa Maria; Brasil; Ciência da Computação; UFSM; Programa de Pós-Graduação em Ciência da Computação; Centro de Tecnologia
• Palavras-chave em Português: Computação em névoa; Internet das coisas; Balanceamento de carga; Fog computing; Internet of things; Load balancing; CNPQ::CIENCIAS EXATAS E DA TERRA::CIENCIA DA COMPUTACAO
• Link de acesso: http://repositorio.ufsm.br/handle/1/22269

Fog Computing is characterized as an extension of Cloud Computing to the edge of the network. Such a paradigm, therefore, does not exclude the Cloud, but complements it, filling gaps such as lower response time and also less use of internet links. This paradigm meets the needs imposed by the Internet of Things applications, which often have restrictions on low processing times, privacy, priority, bandwidth, among others. Considering the growing and diverse demand for Internet of Things applications, the nodes that compose the Fog Computing tend to be overloaded, given the large number of smart things requiring computational capabilities, such as processing, storage, networking, among others. Consequently, overloaded computational nodes compromise the response times of IoT applications that have restrictions for the shortest possible time. In this sense, the main challenge to provide the shortest response time for such applications is the distribution of tasks between the fog nodes. However, the availability of computational resources in the fog must be considered since it is characterized as a dynamic environment to perform load balancing in this new computing paradigm. To alleviate the response time problem, this work presents a load balancing approach that aims to reduce the processing time of the tasks in the fog nodes. The distribution of tasks between the Nodes of the Fog was carried out through dynamic load balancing in real time, whose contribution is therefore the load balancing algorithm that takes into account the dynamics and computational heterogeneity of the environment, as well as the sudden changes in the indexes use of computational resources, which associates tasks more appropriately. To prove the effectiveness of the proposed solution, a simulation environment was organized, where this work was compared with some load balancing approaches, such as RoundRobin and also without a balancer. The results show that high priority tasks consume the shortest possible response time in the environment, either in processing or in the queue, which brings out the effectiveness of the proposed solution. The priority-based queuing mechanism proved to be an important component of the solution, which analyzes and reorganizes the task queue based on its priorities.