Online reliability time series prediction via convolutional neural network and long short term memory for service-oriented systems.

With the development of Web service technology, more and more enterprises choose to publish their own services on the Internet. However, with the increasing demands of users, it is difficult for a single service to meet the complex user requirements. To address this challenge, multiple services can be integrated by leveraging the service-oriented architecture (SOA) to generate a value-added service, referred to a service composition, where the component services are loosely coupled. However, due to the dynamic running environment, the performance of each component service (including reliability) may fluctuate. This will introduce cascading effects, which could cause the entire service system to fail. Since component services run in a dynamic environment, the parameters used to conduct reliability prediction are difficult to obtain. Therefore, online reliability prediction that ensures the runtime quality poses a grand challenge. This paper analyzes the historical response time series and throughput time series of component services, and predicts the reliability in the near future. To guarantee the stable and continuous operation of a service system, we proposed an online reliability time series prediction method by combining a Convolutional Neural Network (CNN) and Long Short Term Memory (LSTM). The proposed approach, referred to as CL-ROP, is able to predict the reliability of a service system in the near future. We conducted a series of experiments over real service data and compared with other competitive approaches to demonstrate the effectiveness of the proposed approach.