Recently, distributed database (DDB) technologies have advanced rapidly and have been investigated extensively. The development and use of DDB systems has been promoted primarily by the growing dispersion of business operations and the rapid technological changes that make the local area network (LAN) possible and practical. The LAN-based DDB system is very flexible. It may serve the information needs of both small businesses and global businesses.
In designing a DDB, one of the major design issues is the file allocation problem (FAP). FAP determines the location of data files for the design of a DDB. In addition, the assignment of workload to computing servers is critical. This issue is known as WAP (Workload Allocation Problem). The majority of studies in the past for FAP or WAP considered the two problems separately. However, FAP and WAP are interdependent as the location of data files affects the assignment of workload and vice versa. Therefore, it is more effective and realistic to solve the two problems simultaneously. Also, previous FAP or WAP models on LANs considered nodal processing overhead separately because incorporating communication overhead is not a trivial task. However, communication overhead should not be ignored because of its significant bearing on communication protocol delay.
The emphasis of this dissertation is on the integrated file and workload allocation problems which simultaneously incorporate communication overhead and local processing overhead in the design of distributed databases via LANs. Four file and workload allocation problems are investigated with special consideration given to LAN types (one-level LAN/two-level LAN) and allocation policies (non-replication/replication) as follows.
The first problem considers a file and workload allocation problem in a one-level local LAN with broadcasting and multi-access capability. The allocation problem is presented in the form of a nonlinear zero-one integer programming problem. Th...