The Internet is undergoing substantial changes from a communication and browsing infrastructure to a medium for conducting business and services. These changes require that Web servers should support the preferential services such as those in E-commerce and also achieve good fairness among different types of requests. In this paper, some integrated schemes of request dispatching and selecting for Web server clusters is provided, with the goal of achieving load balance and meeting the requirements of both WebQoS and fairness. The stochastic high level Petri net (SHLPN) models for them are given. To cope with the well-known state space explosion problem, this paper a novel approximate analysis technique is proposed, which can significantly reduce the complexity of the model solution. Moreover, the numerical results of the performance analysis for those schemes are presented, and the best one among them is recommended that is suited to the preferential services and can achieve high system performance for Web server clusters.