This model depicts a user making a request, which ends up at an ISP, which then forwards the request to the server. This model leaves out many critical components such as routers, switches, and the nature of the request (TCP/IP stack), but is an excellent representation of the basic communication that occurs when a user requests information from a server. The information could be a web page, DNS query, or any other type of request that the server could handle, but the key idea is that a user initiates a connection that is received by a server and communication commences. The methodology behind the actual transmission and reception of the request is beyond the scope of this article and is tied to the nature of the request.

Though many hardware specialists consider there to be one main server model, with variations, I believe that a key to properly administering a server is to understand the many ways a server and client could communicate. Another such model is where there is more than one server for one client, as in the following diagram from http://www.infomotions.com/:

This image depicts a client making a request to a server, and the server in turns makes a request to another server containing a database. There are many advantages and disadvantages to this type of approach over the simpler single client-server model above. The request server is free to handle more requests than the simpler client-server model because data requests are handled by the database server; likewise, the database server is able to handle more database requests than the single client-server model because initial requests (which may not require a database request) are handled by the main server. However, this setup requires more hardware and introduces more points of failure into the client-server model. The same model could be expanded such that more servers are introduced to further decrease the overhead of an individual server, where each individual server processes a certain type of request passed along from the main server. This is typically referred to as a mainframe setup, in which the work load is shared amongst many servers. The system could be further complicated by the fact that multiple servers could perform the same task to prevent an overload of an individual server or to provide a backup in case of hardware failure on an individual server. Again, however, cost and points of failure are directly proportional to the number of servers in the model.

A server administrator must take all of these factors, cost, reliability, and setup, into account when designing the server environment. What becomes evident is that the server environment is as much, if not more, about hardware as it is about software. In the next section, we go into detail about the types of hardware with which a server adminstrator should be familiar.