Um modelo hÃbrido e adaptativo para sistemas distribuÃdos tolerantes a falhas

AUTOR(ES)

SÃrgio Gorender

DATA DE PUBLICAÇÃO

2005

RESUMO

A distributed system can be defined as a set of processes that communicate via communication channels. Models for such systems characterize its aspects, mainly in the temporal and reliability domains. Partially synchronous models have been successfully developed to circumvent the impossibility of achieving fault tolerant consensus in pure, time free, asynchronous model. On the other hand, the new QoS architectures have been developed to provide communication services with distinct levels of service quality, and therefore give rise to hybrid execution environments - where it is possible to provide communication channels with synchronous or asynchronous characteristics. Besides, such a QoS environment may change its configuration at run time, requiring support for dynamic adaptation in the applications running over it. The new challenge is then to develop a distributed system model to execute over an environment equipped with such QoS architectures. This thesis tacked this challenge by presenting a new hybrid and adaptive model for fault tolerant partial synchronous distributed systems. The proposed model is hybrid in the sense that it allows for the coexistence of synchronous and asynchronous processes and communication links. In addition, it allows that processes change their behaviour during execution without affecting the consistency of the upper layer applications. The proposed model has been implemented atop a QoS communication infrastructure. This implementation is based on a new mechanism, also proposed in this thesis, called the QoS Provider (QoSP), whose main goal is to provide a standard interface between our model and existing QoS architectures. A new adaptive consensus protocol developed to execute with the proposed model is also presented in this thesis. This protocol yields optimal results it terms of its resilience and complexity. It tolerates f

ASSUNTO(S)

consensus sistemas distribuÃdos ciencia da computacao tolerÃncia a falhas failure detectors fault tolerance distributed systems detectores de defeitos consenso

Documentos Relacionados

• JEDPI: um ambiente para executar programas distribuÃdos na Internet.
• Barramento de serviÃos federados para integraÃÃo federativa de sistemas distribuÃdos
• Um ambiente para geraÃÃo automÃtica de biblioteca de componentes de comunicaÃÃo em sistemas embarcados distribuÃdos
• Software livre: proposta de boas prÃticas para um modelo hÃbrido de migraÃÃo
• AvaliaÃÃo de dependabilidade de sistemas com mecanismos tolerantes a falha: desenvolvimento de um mÃtodo hÃbrido baseado em EDSPN e diagrama de blocos