The implementation of a multi-backend database system (MDBS)

Description

The multi-backend database system (MDBS) uses one minicomputer as the master or controller, and a varying number of minicomputers and their disks as slaves or backends. MDBS is primarily designed to provide for database growth and performance enhancement by the addition of identical backends. No special hardware is required. The backends are configured in a parallel manner. A new backend may be added by replicating the existing software on the new backend. No new programming or reprogramming is required. A prototype MDBS is being completed in order to carry out the design verification and performance evaluation. This report is the fourth in a series which describes the MDBS implementation. The processes in the MDBS controller (request preparation, insert information generation and post processing) and the processes in the MDBS backends (directory management, record processing and concurrency control) have been described in the previous reports. The concurrency control process, formerly used to control access to just user data, is modified to control access to directory data as well. The directory management process is also modified to improve the execution of update requests. Finally, a directory management is modified for the storage of directory data on the secondary storage. Next, the report describes the revised definitions of inter-process messages (messages between processes within a minicomputer) and intercomputer messages (messages between processes in different minicomputers). Finally, we conclude this series of reports dealing with the implementation of MDBS.

Subjects