The Mobile Internet Analysis Based On DNS Log

The mobile Internet is consisting of mobile communications and the Internet network. With the rapid development of the broadband wireless access technology and mobile terminal technology, people are much eager to acquire information and services from the mobile internet anywhere and anytime, mobile internet comes into being as the result. As the performance smart phones and mobile internet are developing so fast, more and more subscribers start using mobile phones to access the internet, therefore the analysis of the mobile Internet becomes more and important.Domain Name System (DNS) is an internet fundamental service that which is a distributed database and associates various information with domain names assigned to IP address, can make it more convenient for people to access to the internet. Both the traditional PC Internet and the fast growing mobile internet depend on the IP-based networks to implement information and communication services, and such these services have to rely on the domain name to locate the appropriate network resources. Because the DNS accessing log is full of the mobile internet accessing information rich, the log can be used to analyze the mobile Internet properly, and furthermore, accessing patterns of the mobile internet can be studied and analyzed.In this paper, the mobile internet is analyzed based on the DNS log. And the main research work includes as follows:First of all, we describe how to use HDFS (Hadoop Distributed File System) and the Hadoop Distributed programming tools to store and precede the massive mobile Internet DNS logs.Secondly, the pre-processing data that is obtained from the DNS logs is used for the statistical analysis. The main object of analysis consists of the query domain, return domain, ISP (Internet Service Provider) which is extracted from the query name, the server IP, DNS query type, rCode etc. Some conclusions can be drawn that the mobile internet user behavior is different from the traditional PC Internet user behavior during the day; the query domain visits meets the Pareto principle; the total mobile user visits in a whole day shows the exponential distribution.Thirdly, with the use of graph theory, the matrix multiplication ideas, we try to solve the1DNS domain full connectivity issue in a distributed paralleling way. In this section, we use3kinds of matrix multiplication ideas to implement the full connectivity solution.Finally, the improved classical clustering method is used to analyze the DNS data. First, the simulated annealing algorithm and distributed parallel is combined together to implement the more efficient clustering algorithm, with the improved algorithm we can analyze the DNS data after pre-processing analysis, and one result that there does exist some different domain querying patterns.