Research On Dynamic Adjustment And Quality Optimization Of IoT-aware Service Composition

With the rise of service-oriented architecture and the increase in various complex services,the demand for service composition has become increasingly important.Service composition refers to the process of compositing individual services to create new and more complex services that can meet specific needs of users or applications.Technology related to service composition is a necessary method to solve the challenges faced by modern service-oriented systems.In this paper,the combination of services based on business processes makes it a fully functional service to meet the individual needs of users.Combining IoT technology to make IoT-aware business processes spontaneously compose services in response to changes in the physical world,which makes business processes more intelligent.However,the IoT environment changes frequently and rapidly,each IoT environment corresponds to different service composition schemes based on business processes.If a complete business process is established in advance for different IoT environments,the process will often be bloated and bulky.Using rule-based statistical algorithms or traditional algorithms to generate service composition solution is rough and the accuracy is not high.Experts artificially analyze the IoT environment situation to model the service composition plan.Although the accuracy can be guaranteed,the timeliness is relatively poor,because the real-time value of IoT data is very precious,and there is a certain lag in making relevant decisions by experts.Once an unexpected event occurs,it is necessary to dynamically adjust the deployed service scheme in time to meet the needs of users.However,there is currently no systematic method that can analyze IoT real-time data to generate a service composition scheme based on business processes and continuously and dynamically adjust it Therefore,this paper uses machine to learn to replace experts to analyze IoT data and make decisions on service composition solutions based on business processes.To address the above issues,this paper studies the dynamic adjustment method for real-time perception of IoT data changes and the service quality optimization method.The following work is follows:(1)Proposing an IoT-aware service composition dynamic adjustment method.This method includes a model DNN(Deep Neural Networks)and BPM-GCN(Business Process Management-Graph Convolutional Networks)that perceives IoT to generate based service composition solution based on the business process.This model senses changes of IoT data in real time and generates corresponding service composition scheme;it includes a dynamic adjustment algorithm for comparing service schemes,which can quickly mark the difference between the service scheme generated by the model and the deployed service scheme.The algorithm outputs the adjustment operation of the business process level,and calculate the adjustment operation of service level through the quality optimization method of service composition.(2)A quality optimization method of service composition is proposed,which is an algorithm combination of BPM-HA(Business Process Management-Heuristic Algorithm)and NSGA-Ⅱ(Non-dominated Sorting Genetic Algorithm-II).BPM-HA performs fast scheduling while meeting the user’s Qo S(Quality of Service)requirements;NSGA-II performs global iterative optimization and load balancing when the system does not meet the user’s requirements for Qo S.Both algorithms optimize the Qo S with the response time and scheduling overhead as the main optimization objectives.(3)Desiging and implementing of a prototype system.The prototype system consists of three modules: a data processing module,a system support module,and an application module.The data processing module is responsible for collecting,processing,and storing data,providing data input for the application module.The application module is responsible for visualizing IoT data and model-generated service schemes.If users are not satisfied with the dynamically adjusted service solutions,they can adjust them.The system support module includes a front-end and back-end system,a business process execution engine,a deep learning model,and a real-time big data processing framework.