【作者】 孙祥鹏；

【导师】 胡吉明；

【作者基本信息】 河海大学 ， 计算机应用技术， 2005， 硕士

【摘要】 随着分布式数据库系统在业务领域中日益广泛的应用,提供高可用性的服务也就越来越重要。数据库资源高可用性是指,尽可能地保证数据服务的正常运行,并在数据服务出现故障时,由备用数据源在极短的时间内继续提供服务,转换过程相对用户是透明的。 并发控制领域的死锁是分布式系统所要解决的一个重要问题,也是导致数据库资源不可用的一个重要因素。现有的分布式数据库通常采用加锁实现并发控制,通过撤销事务来打破死锁环路,但是仍然存在基于多粒度锁模式下的粒度转换引起的死锁可能性。 在本文提出的基于日志的高可用模型设计中,包括有数据通信层、数据源配置层、监视层、数据库状态控制层、命令转换层、日志层等多个模块结构。本文的对日志层进行了研究,包括异构数据库环境下的日志类型、日志格式、日志空间、日志同步和日志恢复等问题。 基于数据冗余的日志恢复机制是实现分布式数据库系统资源高可用性的重要方式。系统中的所有共享数据资源,将根据用户的需求存在一个或多个副本,并且由系统自动保证这些副本的一致性。当系统中某个数据源出现故障后,对该数据源的访问将自动转移到作为副本的备用数据源。当故障数据源恢复服务后,日志系统将恢复其故障期间未完成的操作,以保证数据的一致性。因此,当系统中一个或几个节点的失效,不会影响系统的正常服务。而当系统正常运行时,日志系统通过LSN自动维护日志记录的一致性。更多还原

【Abstract】 With the popular application of distributed database systems in business sections, providing the high availability(HA) service became very important. Database resource’s HA could be defined that a main data resource provide durative service and a backup data resource could take over all tasks and services in a short time when the main resource failed. Therefore, the entire system may still be available and users would not be aware of switching of nodes.Deadlock in concurrency control was an important problem which should be resolved by distributed systems. Moreover, it was also an important factor of causing unavailability. Distributed databases in existence usually adopt lock mode to implement concurrency control, in via of canceling transaction to break through deadlock loop. However, some kind of deadlock may still appear due to the granularity switch based on multi-granularity mode.In the HA mode Design, including Data Communications Layer, Date Resource Configure Layer, Scout Layer, Database State Control Layer, Command Switch Layer, Log Layer, etc. In this dissertation, it focuses on the research of log style, log format, log space, log synchronization and log recovery in heterogeneous database system.The log recovery mechanisms based on data redundant is a important quomodo to implement distributed database resource’s HA. In such a system, all of share data resources will build one or more duplicates according to user’s need. Log could be used to maintain consistency among duplicates. So, Log became the most popular and reliable technique to maintain system High Availability.更多还原