A Average Response Time Prediction Method For Seasonal Non-Stationary Concurrency Based On Improved RBF Algorithm
Jun Guo, Jiayi Wang, Jina Wang, Aixuan Dong, Bin Zhang · 2021
In the cloud service platform, seasonal non-stationary concurrency is widespread, and the amount of concurrency peaks in the form of a cycle. As the seasonal non-stationary constants continue to change, the service performance of the cloud service platform will be affected. The predicted average response time tends to lag behind the real time load condition by the traditional load balancing strategy. This paper proposes an average response time prediction model based on the improved RBF(Radial Basis Function) algorithm for seasonal non-stationary concurrency. The model uses the RNN-LSTM(Recurrent Neural Network, Long Short Term Memory) algorithm to predict the concurrency. The average response time is predicted by the improved RBF algorithm. In the prediction of seasonal non-stationary concurrency, the mean relative error of RNN-LSTM are 0.0467. In the prediction of average response time, the mean relative error of improved RBF are 0.0125. Experimental results show that the method proposed in this paper has higher prediction accuracy and lower prediction error.