Some businesses have a higher demand for the deployment plan because of the need to keep up with the bigger volume of traffic. The natural answer to this challenge would be and was to add servers with the same .ear or .war file deployed and join all the servers into a cluster. So, one failed server could be automatically replaced with another one from the cluster, and the site user would never experience disconnect of the service. The database that backed all the clustered servers could be clustered too. A connection to each of the clusters went through a load balancer, making sure that none of the cluster members worked more than the others.
The web server and database clustering help but only to a degree, because as the code base grows, its structure can create one or several bottlenecks unless such and similar issues are addressed with a scalable design. One of the ways to do it is to split the code into tiers: front end (or web tier), middle tier (or app tier) and back...