An improved genetic algorithm for cost-effective data-intensive service composition

The explosion of digital data and the dependence on data-intensive services have been recognized as the most significant characteristics of IT trends in the current decade. Designing workflow of data-intensive services requires data analysis from multiple sources to get required composite services. Composing such services requires effective transfer of large data. Thus many new challenges are posed to control the cost and revenue of the whole composition. This paper addresses the data-intensive service composition and presents an innovative data-intensive service selection algorithm based on a modified genetic algorithm. The performance of this new algorithm is also tested by simulations and compared against other traditional approaches, such as mix integer programming. The contributions of this paper are three folds: 1) An economical model for data-intensive service provision is proposed, 2) An extensible QoS model is also proposed to calculate the QoS values of data-intensive services, 3) Finally, a modified genetic algorithm-based approach is introduced to compose data-intensive services. A local selection method with modifications of crossover and mutation operators is adopted for this algorithm. The results of experiments will demonstrate the scalability and effectiveness of our proposed algorithm.