Database Systems: Implementation of a Distributed Database Management Systemto Support Logical Subnetworks

Implementation of a Distributed Database Management System to Support Logical Subnetworks By D. COHEN (Manuscript received November 3, 1981) This paper describes the design and implementation of a specialpurpose distributed data management system. The design and implementation were parts of a study to evaluate the database management needs of software-defined logical subnetworks. The paper describes the authorization model used to define logical subnetworks and the related subnetwork management transactions. The definition of the subnetworks and their service capabilities are stored in a distributed database. The distributed database architecture and the implemented software architecture are described. The requirement to design and implement within a specific time frame has kept the design simple, but the nature of the application dictated that we consider many aspects of the more general distributed data-management problem. The database management issues that are addressed in this paper, in the context of transaction processing, include multicopy updates, concurrency control, and crash recovery. A version of the primary node concept for multicopy updates was adopted. Data inconsistencies, created by premature termination of transaction processing (e.g., system crash), are detected and removed by the software. I. INTRODUCTION The advantages of distributed data processing in general, and distributed data management in particular, have been presented in many publications. In spite of wide interest in its potential cost benefits, 2459 distributed data management has met little acceptance in the field because of its potential impact on the way organizations are accustomed to managing their data-processing facilities.1,2 Issues such as who is responsible for purchasing of equipment, or who is responsible for availability of services have to be revisited in the distributed processing environment.