Method of metadata synchronization in parallel distributed databases

Metadata in a parallel distributed database contains information about table fragmentation, physical storage distribution, and access control policies. It is essential to query processing and security policy enforcement. Metadata in a parallel distributed database is usually stored in the form of system tables, it can be replicated through out every site in the system in order to improve performance and availability. Changes to the metadata should be aware by every site with acceptable delay and consistency. In this paper, a novel method is proposed to deal with the problem of metadata synchronization in parallel distributed databases. A locking mechanism is devised to coordinate different database operations, it can postpone data moving operations and carries out propagation only when necessary. Compared with existing methods, it dramatically reduces the communication between different sites, such that situations can be avoided in which all sites require updates simultaneously and a large portion of network bandwidth is consumed at a burst. Experiment results show the superiority of the proposed method over the others.