Managing Population and Workload Imbalance in Structured Overlays

• Autor(a) principal: Leitão, Ricardo Miguel Gonçalves
• Data de Publicação: 2021
• Tipo de documento: Dissertação
• Idioma: eng
• Título da fonte: Repositório Científico de Acesso Aberto de Portugal (Repositórios Cientìficos)
• Texto Completo: http://hdl.handle.net/10362/151405
• topic: Structured Overlays; Distributed hash Tables; Contingency tables; Domínio/Área Científica::Engenharia e Tecnologia::Engenharia Eletrotécnica, Eletrónica e Informática

Every day the number of data produced by networked devices increases. The current paradigm is to offload the data produced to data centers to be processed. However as more and more devices are offloading their data do cloud centers, accessing data becomes increasingly more challenging. To combat this problem, systems are bringing data closer to the consumer and distributing network responsibilities among the end devices. We are witnessing a change in networking paradigm, where data storage and computation that was once only handled in the cloud, is being processed by Internet of Things (IoT) and mobile devices, thanks to the ever increasing technological capabilities of these devices. One approach, leverages devices into a structured overlay network. Structured Overlays are a common approach to address the organization and distri- bution of data in peer-to-peer distributed systems. Due to their nature, indexing and searching for elements of the system becomes trivial, thus structured overlays become ideal building blocks of resource location based applications. Such overlays assume that the data is distributed evenly over the peers, and that the popularity of those data items is also evenly balanced. However in many systems, due to many factors outside of the system domain, popularity may behave rather randomly, al- lowing for some nodes to spare more resources looking for the popular items than others. In this work we intend to exploit the properties of cluster-based structured overlays propose to address this problem by improving a structure overlay with the mechanisms to manage the population and workload imbalance and achieve more uniform use of resources. Our approach focus on implementing a Group-Based Distributed Hash Table (DHT) capable of dynamically changing its groups to accommodate the changes in churn in the network. With the conclusion of our work we believe that we have indeed created a network capable of withstanding high levels of churn, while ensuring fairness to all members of the network.