Multi-versions data and epsilon-serializability for QoS enhancement in distributed RTDBMS

Distributed Real-Time Database Management Systems (DRTDBMS) are characterized by the large amount of distributed data, the unpredictable transactions and the unbalanced workload between different nodes. A lot of work dealing with Quality of Service (QoS) has been done to control the behaviour of DRTDBMS. They are based on feedback control scheduling theory. In this paper, we propose to apply both multi-versions data technique (MVD) and epsilon-serializability criterion (ESR) in distributed feedback control loop by using three data replication policies. The proposed architecture is called MVD-ESR-DFCS architecture. We also show, by a set of simulations, that our proposed architecture can significantly improve the QoS in DRTDBMS by increasing the number of transactions that meet their deadlines.