WebSphere is made up of multiple layers. A clean install of WebSphere will not establish any of the layers. A clean install of WebSphere is a sort of empty vessel. In fact, the ${was_install_root}/profiles directory will be empty, which is how you know that all you have is the cell but no profile, no node, no application server, and no applications.

A cell is a sort of container that contains all of the layers, such as the profiles, nodes, and servers. The creation of a profile is what starts the process of creating the layers. A profile will always contain at least one node. Then, an application server can be created in a node, and finally, an application can be deployed to an application server.

While this might seem completely theoretical at first, this actually becomes quite pratical. When doing certain things in the WebSphere admin console, there will be a drop-down where you are to select a scope. The scope you select will define if an application is or is not able to access the resource in question. Here is an example of a scopes drop-down.

Or, you may get a tree view like this.

Now it's important to recongize that each parent layer can have multiple child layers. Hopefully this image tells the story. If you think this image paints a scarry picture, imagine being the administrator of a multi-cell domain, or worse yet, troubleshooting an issue! Welcome to the wonder world of engineering.

Let's thing about scopes again. Now, let's say you setup WebSphere so that a certain JDBC (database) connection is in the "node a" scope. Only applications that are in the application servers in "node a" will be able to use the JDBC connection. If an application in any other node attempts to use the JDBC connection, some problem will occur. This can come in quite handy actually, so that you have a way to limit or silo off certain resources to certain applications, which comes in very handy from an architecture perspective.