Abstract
Applications are increasingly characterized by manipulating large amounts of data and by time constraints to which are submitted data and treatments. RTDBMSs (Real-Time DataBase Management Systems) are an appropriate formalism to handle such applications. However, a RTDBMS often goes through overload periods following the unexpected arrival of user transactions. During such periods, transactions are more likely to miss their deadlines and that directly affects the QoS (Quality of Service) provided to users. Thus, our work is to propose a new scheduling protocol to optimize the execution of transactions without exceeding their deadlines. It consists on assigning priorities to transactions based both on their deadlines, their arrival dates and their priority levels defined by users. Also, we show that our approach can maximize the number of successful transactions, in particular those classified as critical for users. The obtained results are compared with conventional scheduling approaches.