CoordenaÃÃo de AperfeiÃoamento de Pessoal de NÃvel Superior / Na computaÃÃo em nuvem, a elasticidade e a capacidade de isolamento de carga permitem que seus recursos sejam aprovisionados e liberados em resposta a cargas de trabalho dinÃmicas com baixo tempo de inatividade. Tais caracterÃsticas sÃo tÃpicas de clusters hospedados em nuvem (cluster virtual - CV), de tal forma que estes recursos precisam ser gerenciados a fim de se garantir a minimizaÃÃo do desperdÃcio de recursos nos provedores e garantir que o desempenho dos recursos nÃo seja afetado negativamente. Este trabalho propÃe uma polÃtica para adaptaÃÃo dinÃmica de clusters virtuais (CVs) a fim de reduzir o nÃmero de recursos ociosos sem comprometer o desempenho dos serviÃos. O algoritmo proposto baseado em heurÃstica, realiza instanciaÃÃo e desligamento de mÃquinas virtuais nos CVs conforme variaÃÃo na demanda por recursos de aplicaÃÃes do tipo saco de tarefas (Bag-of-Tasks, BoT). O algoritmo foi especificado, verificado e validado atravÃs de simulaÃÃes em Redes de Petri (RdP). O desempenho da proposta à avaliado em trÃs cenÃrios distintos a partir das mÃtricas: quantidade total de mÃquinas iniciadas, quantidade de mÃquinas ociosas reutilizadas, tempo total de execuÃÃo da aplicaÃÃo e quantidade mÃdia de clusters iniciados. Os resultados mostraram que a polÃtica de adaptaÃÃo proposta à capaz reduzir a ociosidade e a sobrecarga de um CV e, consequentemente, melhorar o consumo de energia. / In cloud computing, elasticity and capacity of load isolation allow their resources to be provisioned and released in response to dynamic workloads with reduced downtime. These characteristics are typical of clusters hosted in a cloud (virtual cluster - VC), so that these resources need to be managed in order to minimize its waste in cloud providers and ensure that resource performance is not adversely affected. This work proposes a policy for dynamic adaptation of virtual clusters (VCs) to reduce the number of idle resources without compromising their performance of resources. The proposed algorithm, based on heuristics, performs instantiation/shut-down of the virtual machines (VMs) in virtual clusters according variation in demand for resources of applications of type Bag-of-Tasks, BoT. The algorithm has been specified, verified and validated using Petri Nets formalism. The selected metrics to evaluate the proposal performance in three different scenarios are the total amount of machine started, amount of idle machines reused, total time of execution the application and average number of clusters started. The results showed that the adaptation policy proposal is able to reduce idle and the overhead of a virtual cluster and thus improve power consumption.

Autonomic computing aims at making computing systems self-managing by using autonomic managers in order to reduce obstacles caused by management complexity. This thesis presents results of research on self-management for large-scale distributed systems. This research was motivated by the increasing complexity of computing systems and their management. In the first part, we present our platform, called Niche, for programming self-managing component-based distributed applications. In our work on Niche, we have faced and addressed the following four challenges in achieving self-management in a dynamic environment characterized by volatile resources and high churn: resource discovery, robust and efficient sensing and actuation, management bottleneck, and scale. We present results of our research on addressing the above challenges. Niche implements the autonomic computing architecture, proposed by IBM, in a fully decentralized way. Niche supports a network-transparent view of the system architecture simplifying the design of distributed self-management. Niche provides a concise and expressive API for self-management. The implementation of the platform relies on the scalability and robustness of structured overlay networks. We proceed by presenting a methodology for designing the management part of a distributed self-managing application. We define design steps that include partitioning of management functions and orchestration of multiple autonomic managers. In the second part, we discuss robustness of management and data consistency, which are necessary in a distributed system. Dealing with the effect of churn on management increases the complexity of the management logic and thus makes its development time consuming and error prone. We propose the abstraction of Robust Management Elements, which are able to heal themselves under continuous churn. Our approach is based on replicating a management element using finite state machine replication with a reconfigurable replica set. Our algorithm automates the reconfiguration (migration) of the replica set in order to tolerate continuous churn. For data consistency, we propose a majority-based distributed key-value store supporting multiple consistency levels that is based on a peer-to-peer network. The store enables the tradeoff between high availability and data consistency. Using majority allows avoiding potential drawbacks of a master-based consistency control, namely, a single-point of failure and a potential performance bottleneck. In the third part, we investigate self-management for Cloud-based storage systems with the focus on elasticity control using elements of control theory and machine learning. We have conducted research on a number of different designs of an elasticity controller, including a State-Space feedback controller and a controller that combines feedback and feedforward control. We describe our experience in designing an elasticity controller for a Cloud-based key-value store using state-space model that enables to trade-off performance for cost. We describe the steps in designing an elasticity controller. We continue by presenting the design and evaluation of ElastMan, an elasticity controller for Cloud-based elastic key-value stores that combines feedforward and feedback control. / <p>QC 20120831</p>