C.M.R. Almeida (Portugal)
Real-Time Group Communications, QoS Adaptability, Quasi Synchronous Systems, Fault-Tolerance
The response time variability that exists in systems that are not fully synchronous “forces” real-time applications developers to assume as “worst-case”, values that are not the real worst-case bounds, but closer to the normal case. This is done due to practical reasons, but implies the possibility of having timing failures. This tradeoff between tight timeliness and timing failures is acceptable to some classes of real-time applications provided that there exists a way of detecting and controlling the situations when timing failures do occur. In a previous work, we proposed the use of a timing failure detection service to deal with such a situation, that when combined with a hierarchy of groups allows to efficiently handle timing failures in an active replication scenario. There is, however, a non null probability of having runs where all replicas experience a timing failure. In such situations it may be advisable to have the application switch between different qualities-of-service (QoS) so as to have a graceful degradation. In order to preserve safety, this switching must be done in a timely and controlled way. In this paper a solution is presented to deal with the situation of having dynamic QoS adaptability in quasi-synchronous systems.
Important Links:
Go Back
