Repository logo
 
Loading...
Thumbnail Image
Publication

Virtual sensor networks: collaboration and resource sharing

Use this identifier to reference this record.
Name:Description:Size:Format: 
thesis.pdf2.82 MBAdobe PDF Download

Advisor(s)

Abstract(s)

This thesis contributes to the advancement of the Sensing as a Service (SeaaS), based on cloud infrastructures, through the development of models and algorithms that make an efficient use of both sensor and cloud resources while reducing the delay associated with the data flow between cloud and client sides, which results into a better quality of experience for users. The first models and algorithms developed are suitable for the case of mashups being managed at the client side, and then models and algorithms considering mashups managed at the cloud were developed. This requires solving multiple problems: i) clustering of compatible mashup elements; ii) allocation of devices to clusters, meaning that a device will serve multiple applications/mashups; iii) reduction of the amount of data flow between workplaces, and associated delay, which depends on clustering, device allocation and placement of workplaces. The developed strategies can be adopted by cloud service providers wishing to improve the performance of their clouds. Several steps towards an efficient Se-aaS business model were performed. A mathematical model was development to assess the impact (of resource allocations) on scalability, QoE and elasticity. Regarding the clustering of mashup elements, a first mathematical model was developed for the selection of the best pre-calculated clusters of mashup elements (virtual Things), and then a second model is proposed for the best virtual Things to be built (non pre-calculated clusters). Its evaluation is done through heuristic algorithms having such model as a basis. Such models and algorithms were first developed for the case of mashups managed at the client side, and after they were extended for the case of mashups being managed at the cloud. For the improvement of these last results, a mathematical programming optimization model was developed that allows optimal clustering and resource allocation solutions to be obtained. Although this is a computationally difficult approach, the added value of this process is that the problem is rigorously outlined, and such knowledge is used as a guide in the development of better a heuristic algorithm.
Esta tese contribui para o avanço tecnológico do modelo de Sensing as a Service (Se-aaS), baseado em infraestrutura cloud, através do desenvolvimento de modelos e algoritmos que resolvem o problema da alocação eficiente de recursos, melhorando os métodos e técnicas atuais e reduzindo os tempos associados `a transferência dos dados entre a cloud e os clientes, com o objetivo de melhorar a qualidade da experiência dos seus utilizadores. Os primeiros modelos e algoritmos desenvolvidos são adequados para o caso em que as mashups são geridas pela aplicação cliente, e posteriormente foram desenvolvidos modelos e algoritmos para o caso em que as mashups são geridas pela cloud. Isto implica ter de resolver múltiplos problemas: i) Construção de clusters de elementos de mashup compatíveis; ii) Atribuição de dispositivos físicos aos clusters, acabando um dispositivo físico por servir m´ múltiplas aplicações/mashups; iii) Redução da quantidade de transferência de dados entre os diversos locais da cloud, e consequentes atrasos, o que dependente dos clusters construídos, dos dispositivos atribuídos aos clusters e dos locais da cloud escolhidos para realizar o processamento necessário. As diferentes estratégias podem ser adotadas por fornecedores de serviço cloud que queiram melhorar o desempenho dos seus serviços.(…)

Description

Keywords

Internet das coisas Web das coisas Sensing as-a-service Cloud

Citation

Research Projects

Organizational Units

Journal Issue

Publisher

CC License