Research On Distributed Website Monitoring And Performance Analysis

Since the invention of Internet technology, it has gone through40years, and its rapid development and enormous scale make it applied in various industries widely. One of the largest application of the Internet is website, which has penetrated every corner of our country, but many websites have the problem of poor accessibility, user satisfaction is not high and website has been down but no one knows. Administrator can’t know the operating conditions usability issues of the website.Previous research to determine the website whether it is good by single point measurement data, the accuracy rate is low and lack of website prediction function. Combing previous studies’drawbacks, to ensure the website healthy and stable operation, improving the predictive accuracy by distributed monitoring website, in this paper, firstly we introduce performance parameters of the website, simulation tools for building distributed website OPNET, information granulation.Secondly, it introduces the website monitoring classification and evaluation criteria of website performance status, refer to the research results at home and abroad, designed a distributed monitoring system for the website, analyzed the relationship between web performance parameters by simulation data of distributed website monitoring. Finally, use two ways that a single point and distributed monitoring to monitor the actual site. Use random verification, cross-selection, genetic algorithm and particle swarm optimization to state of the website classification and comparison instructions. Predict response time for a period of time based on support vector machine and information granulation way, the relative to the minimum prediction accuracy rate can reach95%, the mean relative accuracy rate of96.2%, the maximum relative accuracy of87%, it can predict the response time range of the website more accurately, ensuring the website healthy and stable operation, improving website usability and user satisfaction. Successfully predicted the issue has important significance and application value to the increasingly serious network security problems.