Open The Web Load The Device Model

Companying with the maturation of the Internet, Web-based applications become more and more popular. This trend makes it a very important and pressing task to develop those technologies enhancing the service quality of web applications. Web benchmark plays an important role in web service and web server performance enhancing, and it is helpful to the website layout. Some widely used web benchmark tools are analyzed in this thesis. The results suggest that the web workload generator is in the kernel of the web benchmark fields. Two main works, focusing on construct a better model of web workload generator, which we mean has more user equality, are discussed in the thesis. The first is a web visiting data collection mechanism, which collects data from those clients using a special source-open web browser. The second is E-SURGE, an open model of web workload generator. This model is gained on the base of analyzing some existing web workload generators. The stream of web workload generated by E-SURGE can be accurately equivalent to those workloads in the real world.The structure and principle of WebStone, a widely used web benchmark tool, is introduced in this thesis. It is used to test a mini web site and some test results are gained. Maybe these results seem to be apparent, or have been gained by other tests, even maybe incorrect, the test itself can enhance our understanding of the web benchmark process and web tools' structure.To construct an user equivalent web accessing model, the first thing should be done is lo collect a pile of user accessing data records, which can be used to analyze the characters of user accessing pattern to the web. A data collection mechanism, which is based on Web data mining technology and an open-source browser, is provided in the thesis. The data collector collects data records from web clients, it can also pre-process those data collected, record them in a simple and efficient way.SURGE, an important web workload generator model, analyzed the characters of web workload comprehensively. Some distribution models of workload characters introduced by SURGE, and the SURGE model itself, are widely accepted by other researchers, they provide us the basis of constructing our own web workload generator model. SURGE, though is important, uses single character model to characterize millions of web clients, this maybe inaccurate and incomplete, and more seriously, SURGE cannot reflect the variation of web clients' characters For variation, we think it is the primary character of web clients' accessing pattern. Accordingly, an open web workload generator model, E-SURGE, is provided in the thesis E-SURGE expand the SURGE model in two respects: first, it divides all webclients into some definable user groups, these user groups can be characterized separately. Second, user groups' characters can be adjusted during the test according to some stimulation. We believe that workloads produced by E-SURGE can be accurately equivalent to those real workloads.Limited by time and testing environments, the data collector and the E-SURGE model have not been used in the real environments. Surely, they are inaccurate and incomplete. Some important research works related to this thesis are still on the head for us.