As stated above, the group should contain all resources that must be on the same node for processing to occur. In that sense, the resources of the group are interdependent. More pragmatically, often one resource must be online before another resource can configure itself and start correctly. The second resource is said to "depend" on the other. The Cluster Administrator allows you to designate one resource as depending on another. If this is done, the Cluster service will not start the resource until the one it depends upon is started. If a resource fails, the resources that depend on it will be taken offline, and, if it restarts, the resources will be returned to an online state.
Many of the resource types that ship with MSCS have requirements for dependencies. The only resource types that do not usually depend on another resource are the Physical Disk, and IP Address. The Network Name resource must depend upon an IP address so that it can register the name and address pair with WINS. The other resources have similar requirements.
The cluster service will start resources in the order of their dependencies. In the above example, a typical group going online would start the Disk and IP Address resources first, because they have no dependencies. The Network Name resource would be started next, and after that starts, the File Share resource would start, because both the Disk and Network Name resources are online.
Note that it is not possible to have resources in different groups depend on one another. Because different groups can reside on different nodes, there is no way to guarantee that the dependencies can be satisfied.
You can also define dependencies where none are required to satisfy logical dependencies that you create. For instance, you may not want your Internet Information Server (IIS) virtual web page describing the shares to which your users have access to be online if the SMB shares themselves are not available.