Diversity in Automatic Cloud Computing Resource Selection

• Autor(es): Cogo, Vinicius Vielmo
• Data: 2011
• Identificador Persistente: http://hdl.handle.net/10451/13887
• Origem: Repositório da Universidade de Lisboa
• Assunto(s): Diversity; Resource Management; Cloud Computing; Intrusion Tolerance
• Tipo de Documento: Dissertação de mestrado
• Idioma: Inglês
• Orientador(es): Pasin, Marcelo
• Contribuidor(es): Repositório da Universidade de Lisboa

Descrição

Obtaining correct results and behaviour on computing is a long-standing concern. Such guarantee can be obtained through fault and intrusion tolerance mechanisms, which aim to tolerate crash and arbitrary faults. Byzantine fault tolerant replication, when combined with proactive recovery techniques can tolerate any number of arbitrary faults during entire system life time. However, common vulnerabilities shared between replicas can compromise such tolerance, rendering diversity as a complementary mechanism. Diversity is a mechanism that consists in providing and combining diverse resources to increase vulnerability independence between system components. Obtaining diversity automatically is a process that can be decomposed into two phases: creation and selection. The first phase consists in providing enough diverse resources to be considered, combined and selected in second phase. In this thesis we present the DiversityAgent, a Java library for selecting cloud resources considering multiple diversity properties. Its clients only need to register available resources, then the DiversityAgent assumes the responsibility of selecting appropriate cloud computing resource combination for each server deployment. In order to design the DiversityAgent, we review taxonomies for diversity on computer systems and analyse several diversity group properties supported by cloud providers or tools, and opportunities for cloud computing players contribute with diversity management area. This document contains a review on basic fault and intrusion tolerance mechanisms, followed by an extensive diversity analysis in cloud computing environments and by the DiversityAgent development. We also present an integration of our component with two use cases foreseen by CloudFIT project, as well as present the results of correctness and performance evaluations. At the end there are the final remarks about this work and possible future work, besides three appendices regarding DiversityAgent public interfaces,usage and customising tutorials.