Computer Science Speaking Skills Talk

Friday, April 29, 2016 - 12:00pm to 1:00pm


Traffic21 Classroom 6501 Gates & Hillman Centers



Replication is a fundamental technique to guarantee service availability and consistency. However, when a system component may misbehave, for instance due to a software intrusion, coordinating its replicas requires the adoption of complex distributed fault-tolerant protocols. In fact, such protocols require several executing replicas and many communication steps to work correctly. This results in redundant service executions that are inefficient and typically force the service to be deterministic so to achieve consistency (i.e., if correct replicas process the same requests in order, they produce the same output). In this seminar we introduce the concept of hardware-based computation verification in the area of fault-tolerant replication. It is shown how the technique allows to deal with malicious replicas while linearly (in the number of replicas) reducing the computational effort to execute a fault-tolerant distributed service. In particular, we design a fully passive replicated system that dramatically reduces the fault-tolerance complexity, in terms of protocols, number of messages, processing effort and allows non-deterministic operations. Also, we implement a system prototype based on commodity hardware and on recent research work in Trusted Computing (TC). Finally, we show experimentally that our system delivers good performance, compared to optimized state-of-the-art systems, despite its main overhead is due to TC technology. Presented in Partial Fulfillment of the CSD Speaking Skills Requirement.

For More Information, Contact:


Speaking Skills