Reliability-based optimal task-allocation in distributed-database management systems

A Distributed Database Management System (DDBMS) manages a collection of data stored at various processing nodes of a computer network. Global queries are processed by DDBMS, and tasks (Join operations in these queries) are assigned to various nodes in the network. Optimal allocation of these tasks are discussed, with reliability as the objective function and communication-cost as the constraint function. A model for reliability-based task allocation is developed. It is converted into a State-Space Search-Tree and is implemented using a branch-and-bound algorithm. The query is considered as a multiple Join problem (MJP). Reliable task-allocations for all permutations of MJP are found. The permutation which has minimum cost of data transmission is calculated. A method is suggested for calculating the reliability of the path that a DDBMS network follows to deliver a message to the receiver, given the number of spoolers