The exponential rise in the number of Web users has inspired the creation of a diversity of Web applications. Hence, Web Quality of Service (QoS) is an increasingly critical issue in Web services, such as e-commerce, Web hosting, etc. In the future, improved QoS will be linked to a fee for service. Customers expect their requests to be served with a quality proportional to the amount charged to their accounts. Because most Web servers currently process requests on a first-come, first-serve basis, they do not provide differentiated QoS. This paper presents two approaches to implement differentiated quality of Web service. In the user-level approach, the Web server is modified to include a classification process, priority queues and a scheduler. However, with this approach, it is difficult to achieve portability. In this paper, a new, portable user-level approach is presented. In the kernel-level approach, a real-time scheduler to support prioritized user requests has been added to the operating system kernel. Prototype implementations for two approaches have been developed and their performances are evaluated.
展开▼
