Because of its cost-effectiveness in achieving high availability and incremental scalability, server clusters have been recognized as a powerful architecture for Internet services. Web switching technology has become an increasingly important building block for Web content providers, Web-oriented enterprises, and data centers. However, the Web standard protocol proposes the use of persistent connections to convey several HTTP requests to the server on a single TCP connection. This creates significant challenges in the web clusters, and may constrain the policies they can support. This paper discusses the performance problem caused by persistent connections in the web clusters. We also describe the design, implementation, and performance of an efficient approach to address this issue.