Characterization and Analysis of a Nested Genetic Algorithm for Distributed Database Design

Abstract

Distributed database design is a difficult and complex task involving two

interdependent problems : data allocation and operation allocation. First,

data must be allocated to nodes in the network. Second, given such an allocation,

data must be efficiently retrieved, processed, and possibly communicated

to meet the retrieval and update requirements of the users. The

problem is characterized by integer variables, a discontinuous and extremely

complex cost function, and numerous constraints. A nested genetic algorithm

naturally fits this problem formulation with the outer algorithm addressing

data allocation and the inner algorithm addressing operation allocation. We

present and characterize such an algorithm according to its gene structure

and control parameters. We experimentally analyze the effects of poolsize

and crossover operator on the performance of our algorithm.

Index Terms - Genetic algorithms, performance modeling and analysis, experimental

analysis of algorithms, distributed database design