| Popis: |
Real-time database management systems (RT-DBMS) have the necessary characteristics for providing efficient support to develop applications in which both, data and transactions have temporal constraints. However, in the last decade, new applications were identified and are characterized by a large geographic distribution, high heterogeneity, lack of global control, partial failures and lack of safety. Besides, they need to manage large data volumes with real-time constraints. Scheduling algorithms should consider transactions with soft deadlines and the concurrency control protocols should allow conflicting transactions to execute in parallel. The last ones should be based in their requirements, which are specified through both quality of services functions and performance metrics. In this work, a method to model and develop applications that execute in open and unpredictable environments is proposed. Based on this model, it is possible to perform analysis and simulations of systems, to guide the decision making process and to identify solutions for improving it. For validating the model, a case study considering the application domain of sensors network is discussed. |
