Web Service Recommendations Based On Load Balancing

Service-oriented architecture(SOA)is a common framework based on standard protocols,coarse granularity,and loose coupling.It supports platform-independence and standard protocols,builds functional units into services that are accessed through neutral interfaces,and improves the efficiency of software development.It adds services in a modular way to improve the reusability of soffware resources as business requirements change.Due to the large increase in the number of current services,how to select specific services from a large number of services to meet user needs,service recommendation can automatically recommend services according to user preferences.Since the load resources of the server are certain,different services occupy different load resources of the server.In order to recommend appropriate services to the server,realize load balancing of the server and improve the load capacity of the server.For the problem that the traditional load balancing strategy cannot effectively measure the load characteristics and preferences of the server,this paper proposes a web service recommendation method based on load balancing.According to the load characteristics of different servers,implement service recommendation and integrate it into the load balancing strategy of the server.The main work of this paper is divided into the following three aspects.(])Web service prediction is implemented by probability matrix decomposition method,and load-related attributes(such as response time,throughput,etc.)of the service are analyzed and predicted by historical running state and real-time running state of the service,and the predicted results are used for static and dynamic recommendation of the service.(2)before the server runs,the static recommendation of services is implemented through collaborative filtering algorithm,the similarity between predicted services is calculated,similar services are recommended to the server,the recommended services are arranged on the server,and then the dynamic recommendation of services is carried out.(3)when the server runs,the collaborative filtering algorithm is used to implement the dynamic recommendation of services,calculate the similarity between the predicted services,recommend similar services to the server,and assign the service request of the recommendation service to the server.Service request allocation is implemented by polling algorithm and server load prediction method.The polling algorithm rotates service requests to internal servers.The probability matrix decomposition method is used to predict the server load.During the service request allocation process,if the server load is too high,the service requests are successively polled to the next server until the server load is normal.The Web service dynamic recommendation method can reasonably adjust the service request allocation of each server and improve the load capacity of the server.The outliers of the training data in the web service prediction model and the server load prediction model are processed by the method of sample weighting to improve the accuracy of web service prediction.The validity of web service prediction model and server load prediction model is verified by root mean square error(RMSE)and mean absolute error(MAE).The results show that the web service prediction model and the server load prediction model have better prediction effect.The validity of Web service static recommendation method and Web service dynamic recommendation method is verified by the accuracy and recall rate.The results show that Web service static recommendation method and Web service dynamic recommendation method have better recommendation effect.The fitting degree between the predicted value and the real value and the sample point distribution is judged by the residual curve.