(m, k)-Firm Constraints and Derived Data Management for the QoS Enhancement in Distributed Real-Time DBMS

摘要

Distributed Real-Time DBMS (DRTDBMS) is a collection of Real-Time DataBase Management Systems (RTDBMS) running on sites connected together via communication's networks for transaction processing. This system is characterized by the data distribution and the unpredictable transactions. In addition, in this system, the presence of several sites raises the problem of the unbalanced load between those nodes. In order to enhance the performance of DRTDBMS with taking into account those problems, Quality of Service (QoS) based approaches are the most appropriate. Distributed Feedback Scheduling Control Architecture (DFCSA) is proposed for managing the QoS in this system. In DRTDBMS, the results produced in time with less precision are sometimes preferable than exact results obtained in delay, inaccuracy can be tolerated. In order to take this assertion, in this paper, we extend the DFCSA by using the (m, k)-firm constraints, which take into account the imprecise results, using three data replication policies. The obtained architecture is called (m, k)-firm-User-DFCS. The second contribution consists of taking into account the real-time derived data on (m, k)-firm-User-DFCS architecture, always, using three data replication policies. The obtained architecture is called Derived Data Management (DDM)-(m, k)-firm-User-DFCS. Then, we are focusing on two ways of service optimization in DRTDBMS. We are interested in (1) the space optimization in which we propose to apply three replication data policies, and (2) the QoS optimization in which we propose to take into account the real-time derived data and (m, k)-firm constraints.