Traditionally, with big installation sizes and long startup times, application servers have been used to deploy several, if not dozens of, enterprise applications. A server instance was shared among several teams, sometimes a whole company. This comes with certain inflexibility, similar to shared application models. Teams cannot simply choose newer JDK or server versions, or restart or reconfigure the application server without coordinating with other teams. This naturally inhibits fast and productive processes, and complicates Continuous Delivery.

In terms of team working methods, and project and application life cycles, the easiest approach therefore is to deploy an application on a dedicated application server. The DevOps team has full control over its version, configuration, and life cycle. This approach simplifies processes and avoids potential issues such as collisions with other teams and used technology. Issues with hierarchical classloading that deploying several applications could introduce are avoided as well.

An application server certainly represents quite a construct for just a single application. However, as we have seen previously, the installation sizes of application servers have already decreased compared to the past. Apart from that, developers should care more about the sizes of the deployment artifacts, since these are the moving parts of the development workflow. In a Continuous Delivery approach, the application is potentially built and packaged many times a day. The longer the time spent on the project build and transmitting artifacts, the longer the turnaround times. This affects each and every build and can add up to a lot of overhead during a day. The application server is not installed and shipped that often. Therefore, it is advisable to deploy an application to a single, dedicated Java EE server:

In the next chapter, we will see how container technologies such as Docker support this approach. Shipping the application, including the whole stack down to the operating system as a container, encourages the approach of one application per application server.