Quality Of Service Management In Mobile Real-time Database Systems

With the extending of mobile real-time database systems applications in e-commerce and traffic controlling, the timing constraints of data objects in these applications are growing higher and higher. And these applications require database systems to reflect the current data objects state using fresh data value before the objects status in physical world change. At the same time the transactions in mobile real-time database systems have strict timing constraints and must be committed in their deadlines, otherwise it will lead to a heavy damage even catastrophe. The objective of a mobile real-time database system is not system throughout but the ratio of transactions that meet their timing constraint. Thus we must resolve the academic and technical problems about the data models, transaction models, and priority assignment and schedule strategies of the mobile real-time transactions. Conventional scheduling methods of transaction in database systems are not suitable to mobile real-time applications because they don't consider the timing constraints of data objects and transactions.The unpredictable system workload and even possible overloading, the limited system resources are the problems which must be resolved in mobile real-time applications. It is impossible to meet the various requirements of performance because of the limit constraints of system resources. Many scholars proposed improving data freshness to resolve the problems, but they only focus on improving the single performance. Therefore, quality of service management is proposed to process data objects and improve system performance. And a novel performance metric is presented to ensure system quality of service in mobile real-time database systems for example the deadline miss ratio of transactions, data freshness, CPU utilization and system energy consumption and so on. It is hard to meeting same standard for all performance parameters simultaneously because they conflict possibly each other. So a novel complicated performance of quality of service is proposed to balance all the performance parameters, and the system employed feedback control architecture to realize it. At the same time QDDU policy, the management policy of quality of service based on service differentiation and imprecise technology are proposed to ensure the system performance. The relationship among all data objects should be included when managing quality of service in mobile real-time database systems. The data objects are divided into derived data objects and bases data objects, and a directed acyclic graph is presented to describe the relationship among them in this paper. It is not necessary to update the data object when its derivation in value domain can be neglected. The reason is that it cannot improve value of database systems and the workload is upgraded. Therefore, the validity in value domain is proposed to describe its derivation of a data object which can be neglected.The requests on the database systems of transactions which have different deadlines are distinct. The system performance will be downgrade heavily when hard update transactions miss their deadlines, and it is acceptable for soft update transactions when they miss their deadlines. Service differentiation is employed to schedule the transactions with different value and to decrease competing for system resources. At the same time imprecise computation technology is applied to manage quality of service in mobile real-time database systems. A complicated and nested transaction is classified into a mandatory part and many optional parts. The system prefers executing the only mandatory part to process when it will take too much resource to process a whole transaction. Moreover, conventional concurrency control and scheduling methods for update transactions are not suitable in that mobile real-time transactions are complicated and nested. RODDFTS and RODBFTS algorithms based on trigger mechanisms are proposed to schedule update transactions in quality of service management, and feedback control architecture will be applied to realize the two algorithms. They are realized based on traversal a directed acyclic graph of data objects, and they are recursive algorithms. The algorithms proposed in this dissertation can guarantee the quality of service performance parameters (steady and transient) will not be beyond the references by database administrators through simulation experiments, so the complicated performance can be upgraded.