The Cloud-edge-based Dynamic Reconfiguration to Service Workflow for Mobile Ecommerce Environments
Honghao Gao,Wanqiu Huang,Yucong Duan
DOI: https://doi.org/10.1145/3391198
IF: 5.3
2021-02-01
ACM Transactions on Internet Technology
Abstract:The emergence of mobile service composition meets the current needs for real-time eCommerce. However, the requirements for eCommerce, such as safety and timeliness, are becoming increasingly strict. Thus, the cloud-edge hybrid computing model has been introduced to accelerate information processing, especially in a mobile scenario. However, the mobile environment is characterized by limited resource storage and users who frequently move, and these characteristics strongly affect the reliability of service composition running in this environment. Consequently, applications are likely to fail if inappropriate services are invoked. To ensure that the composite service can operate normally, traditional dynamic reconfiguration methods tend to focus on cloud services scheduling. Unfortunately, most of these approaches cannot support timely responses to dynamic changes. In this article, the cloud-edge based dynamic reconfiguration to service workflow for mobile eCommerce environments is proposed. First, the service quality concept is extended. Specifically, the value and cost attributes of a service are considered. The value attribute is used to assess the stability of the service for some time to come, and the cost attribute is the cost of a service invocation. Second, a long short-term memory (LSTM) neural network is used to predict the stability of services, which is related to the calculation of the value attribute. Then, in view of the limited available equipment resources, a method for calculating the cost of calling a service is introduced. Third, candidate services are selected by considering both service stability and the cost of service invocation, thus yielding a dynamic reconfiguration scheme that is more suitable for the cloud-edge environment. Finally, a series of comparative experiments were carried out, and the experimental results prove that the method proposed in this article offers higher stability, less energy consumption, and more accurate service prediction.
computer science, information systems, software engineering