Research Of WEB Container Memory Allocation Techniques Optimization Based On JVM

Based on Java Web technology is developed with the development of Internet,the site is growing.Since the beginning of the Java language is provided in the memory is automatically assigned recycling technology and popular.The virtual machine is responsible for the recycling memory enables developers to get rid of the memory operations of heavy,also avoid a lot because of the wrong operation of memory and cause crashes happening,but at the same time also makes the performance of the system to a great extent,depends on the efficiency of memory allocation recovery mechanism.This paper studies the current advantages and disadvantages of main memory allocation mechanism,analyzed the traditional Java virtual machine in a Web environment memory allocation method has certain limitation;On the analysis of the prediction based on neural network technology to predict the feasibility of user access behavior,the focus is put forward based on the user access behavior under the Web environment of neural network prediction model,and presents a model of network structure design,the activation function and the selection of initial parameters;In view of the sample data of the "noise" point of problem,put forward a kind of input data preprocessing and anomaly detection approach;To predict network learning algorithm are presented,and insufficient in algorithm is proposed to improve the way.Based on the proposed user access behavior forecasting model and neural network theory is designed and developed based on Java virtual machine's memory allocation optimization program Web site;Give the detailed implementation of program main function parts;Design and the related experiment,proves that the prediction model is developed based on user access behavior memory allocation optimization program can to a certain extent,optimize the purpose of the virtual machine memory allocation,improve the performance of the Web system.