Researches On Recovery Techniques Of Embedded Real-Time Database Systems

The recovery strategy suitable for different demands of applications is critical to the embedded real-time database systems which can be running reliably and effectively in the embedded real-time environment. To achieve the robustness and the high availability of the embedded real-time database systems, the failure recovery strategy based on real-time logs, checkpoint strategy, reload strategy and the applications recovery in mobile environment are researched on the basis of analyzing the characteristics of data and transactions of the embedded real-time database systems.According to the special needs of embedded applications and the timing constraints of data and transactions in embedded real-time databases, the architecture and the system composition of the embedded real-time database management system which use the main memory databases as the underlying support are presented. Furthermore, the architecture of the recovery system is also presented.The characteristics and the consistency of data and transactions in embedded real-time database systems are analyzed, moreover, the recovery characteristics and demands are proposed based on the analysis.In order to guarantee the time consistency, the predictability of recovery time, fast and timely recovery, as well as the low recovery overhead in system failures, the failure recovery strategy based on real-time logs is discussed.A real-time fuzzy checkpoint strategy supporting the embedded real-time database systems is designed to reduce the number of the transactions which miss their deadlines and get a rapid recovery after a failure on the basis of embedded real-time databases ARTs-EDB. Besides, the strategy to determine the redo point in failure recovery is discussed to ensure the durability of the committed transactions during failure recovery and decrease the recovery time. For the sake of meeting the data and transactions timing constraints, as well as reducing the number of transactions missing their deadlines and recovery time in the embedded real-time main memory database systems, a partition checkpoint strategy SPPFCS based on the priority of data segments is presented, and what's more, a partition reload strategy SPPRS in failure recovery is discussed accordingly. SPPFCS partitions the database in accordance with the priority of data segments, and sets the corresponding checkpoint frequency for each partition to execute the partition checkpoint operations, which reduces the ratio of the transactions missing their deadlines and the recovery time. SPPRS reload the database in the light of partition checkpoint priority which reduces the system halt servicing time and improves the system availability during the recovery for considering the requests of the transactions to data.The robustness and the high availability of the embedded real-time database systems are also reflected in system applications recovery in mobile environment. A mobile agent-based framework for supporting applications recovery in a mobile environment is presented to improve the availability of the applications. A log management scheme on the basis of this mobile agent-based framework is also proposed. Log unification strategy and sending notification message strategy in log management are not only reduce the recovery time, but also have the better and consistent system performance on handoff log management costs, log retrieval costs and failure costs.