How to manage real-time transactions in non-strict context

Real-time database management systems (DBMSs) attempt to reach the double objective of maintaining both database logical and temporal consistency while attempting to enhance concurrency in transaction execution. This objective being difficult to reach and since some real-time applications need not respect this objective, some proposed work attempts to bring new concepts for relaxing transactions and/or data properties in a real-time context, that might be respected in traditional DBMSs. According to data fluctuations and/or deadline fluctuations, there are certain kinds of applications. In these applications, the key problem is to efficiently schedule real-time transactions that allow to manipulate inaccurate data and/or that have deadline overtaking. We show that multimedia applications are the best candidates to use these notions which are tightly connected to QoS of these applications. We propose transaction concurrency control and scheduling protocols that allow to bound data imprecision while allowing some transactions to execute beyond their deadlines by a bounded time quantity. These algorithms are based on the extended earliest deadline algorithm which cooperates with the transaction manager in order to deal with transaction concurrency control