达梦嵌入式数据库恢复机制研究

发布时间：2018-03-19 20:29

本文选题：嵌入式数据库　切入点：恢复　出处：《华中科技大学》2012年硕士论文　论文类型：学位论文

【摘要】：随着嵌入式技术的不断发展，嵌入式数据库应用得越来越广泛。嵌入式环境有许多特点和限制，如移动性、网络不稳定性以及电源能力等，对嵌入式数据库可靠性和性能提出了更高的要求和挑战，一个强健的恢复子系统是嵌入式数据库必不可少的组成部分。通过系统深入地研究数据库恢复技术，对达梦(DM)嵌入式数据库恢复子系统进行了改进，以提升系统可靠性和可用性。为适应嵌入式环境，使系统设计遵循简约性和微内核特性，对数据库回滚段进行了设计与改进，将回滚段集成到数据文件中，进行系统恢复时仅搜索预留回滚段，节省了恢复时间；同时将日志文件设计成单日志形式，并采用事务一致性检查点缩短恢复时需要扫描的日志量。在系统故障恢复时，为避免对所有Redo日志进行重做，设计并实现了选择性重做方案。对于那些已经写入到数据库文件中的记录，根据页面的日志序列号(LogSequence Number, LSN)和日志记录的LSN进行判断后再决定是否进行重做，而不是盲目地重做，分析并给出了使用该方法的前提条件。该方法节省了系统缓冲区资源，减少磁盘I/O，体现了嵌入式数据库微内核的要求。为充分利用操作系统多道程序设计思想以及达梦嵌入式数据库特殊的事务模型，系统恢复时采用多线程方式重做日志加速恢复。使用一个日志分发线程、多个重做线程的模型，线程间同步协作完成重现历史操作，达到快速重建系统崩溃时系统状态的目的。经过实验测试表明，改进后的恢复子系统能正确处理系统遇到的事务故障、系统故障和介质故障，且能缩短系统故障恢复使用的时间。
[Abstract]:With the continuous development of embedded technology, embedded database is more and more widely used. Embedded environment has many characteristics and limitations, such as mobility, network instability and power supply ability, etc. This paper puts forward higher requirements and challenges to the reliability and performance of embedded database. A robust recovery subsystem is an essential part of embedded database. In order to improve the reliability and availability of the Darmenton (DMN) embedded database recovery subsystem, the system is improved. In order to adapt to the embedded environment and make the system design follow the characteristics of minimization and microkernel, the database rollback segment is designed and improved. The rollback segment is integrated into the data file, and only the reserved rollback segment is searched for system recovery. The recovery time is saved and the log file is designed as a single log form and the amount of logs that need to be scanned is shortened by using transaction consistency checkpoint. In order to avoid redoing all Redo logs during system recovery, a selective redo scheme is designed and implemented. According to the log sequence number of the page, LogSequence number (LSNs) and the LSN of the log record, we can decide whether to redo it or not instead of blindly redoing it. The precondition of using this method is analyzed and given. This method saves the system buffer resource. Reducing disk I / O reflects the requirement of embedded database microkernel. In order to make full use of the multiprogramming idea of operating system and the special transaction model of Dameng embedded database, the multithread redo log is used to speed up the recovery of the system, and a log distributing thread is used. In the model of multiple redo threads, the synchronization and cooperation between threads complete the reproducing history operation, so as to quickly reconstruct the system state when the system crashes. The experimental results show that the improved recovery subsystem can correctly deal with the transaction faults, system failures and medium failures encountered by the system, and can shorten the time of system fault recovery.
【学位授予单位】：华中科技大学
【学位级别】：硕士
【学位授予年份】：2012
【分类号】：TP368.1;TP311.52

【参考文献】