Sérgio Gorender,  Raimundo J. A. Macêdo,  Matheus Cunha. 

Fault tolerance and adaptability are important issues for modern distributed systems. The capability of dynamically adapting to distinct run-time conditions is especially relevant for environments where negotiated QoS conditions cannot always be delivered between communicating processes. In this paper, we present an implementation of a framework for communication and resource management, using a network of LINUX workstations, which were equipped with a traffic control packet. We also implemented a fault tolerant mechanism, composed of a failure detector and a consensus protocol, which dynamic adapts to distinct QoS levels. The consensus protocol has the property that in the same consensus distinct processes can have different QoS, which characterize a hybrid fault tolerant model. We also measure the execution of the consensus, and present the result of these experiments.

http://www.lbd.dcc.ufmg.br:8080/colecoes/wtf/2004/001.pdf

Processando consulta.