Big services are collections of interrelated web services across virtual and physical domains, processing Big Data. Existing service selection and composition algorithms fail to achieve the global optimum solution in a reasonable time. In this paper, we design an efficient quality of service-aware big service composition methodology using a distributed co-evolutionary algorithm. In our proposed model, we develop a distributed NSGA-III for finding the optimal Pareto front and a distributed multi-objective Jaya algorithm for enhancing the diversity of solutions. The distributed co-evolutionary algorithm finds the near-optimal solution in a fast and scalable way.
K E Y W O R D Sbig services, distributed co-evolutionary algorithm, distributed Jaya, distributed NSGA-III, quality of service, service composition
INTRODUCTIONWith the onset of cloud computing, Internet of Things (IoT), and machine learning/ deep learning, manufacturing industries are undergoing a drastic change. Modern industries are incorporating these new technologies to achieve better efficiency in manufacturing products and services, with an improvement in quality control and a reduction in production costs. The Fourth Industrial Revolution (or Industry 4.0) is the ongoing automation of traditional manufacturing and industrial practices, using these modern smart technologies. Big services are collections of interrelated web services across the virtual and physical domains of Industry 4.0, processing Big Data. 1,2 Big services are integrated from various domains to develop a composite service that addresses the requirements of a customer. 3 Consider a fictitious company that manufactures electric vehicles. The company uses IoT, cloud computing, and machine learning/ deep learning to test for and predict possible failures and for tweaking or designing new engine parts, suspension units, and many more. Let us consider that the company has two divisions: design analysis unit and production unit. The design analysis unit analyzes the performances of the electric vehicles and suggests new design models. The lessons learned from the design analysis unit are incorporated in the design and manufacturing of electric vehicles by the production unit. Assume that both the design analysis unit and the production unit use a similar workflow. We can imagine two separate sets of service composition requirements: The design analysis unit requires high performance in terms of response time and throughput, high security, and with no restrictions on cost. The production unit has to keep the cost low and maintain the desirable quality of service (QoS) parameters and changes with respect to supply and demand throughout the year.In this scenario, we need to execute the service composition process repeatedly whenever there is a change in the requirements, with certain changes in the fitness function. This becomes computationally expensive as the number of services grows, being classified as NP-hard. 4,5 Co-evolutionary approaches offer promising solutions in this case and man...
