The massive growth of the Internet paired with the rise in dynamic website content has increased the need for scalable network architectures. While these various network architectures should be transparent to the client, the speed, reliability, and maintainability of the system depends on the particular architecture that is implemented. This paper will discuss the findings from a case study that tests the speed of two software architectures that are often implemented to build scalable web application systems. The first architecture is server clustering with shared resources. Server clustering can be defined as a group of servers that directly share resources and actively partitions work based on work loads. Thus client traffic to the cluster can be distributed across several physical machines each running an instance of the application server. The other architecture is a shared nothing design, where application servers do not share resources, except for a dispatcher (load balancer). This paper addresses the question, what is the performance overhead of adding application servers into a tiered system.