Licensing requirements

Having a valid per seat mode CAL only guarantees access to a server configured in the per seat mode. It does not guarantee access to a server that is licensed in the per server mode. A valid per seat mode CAL also consumes one of the licenses assigned to the pool of available per-server licenses assigned to the server. Therefore, the client can connect only if there are fewer connections than the limit allowed on the server.

The per seat option is often the most economical one for networks in which clients tend to connect to more than one server concurrently. For example, suppose that you have several servers that users must connect to during the day. On one server you have Microsoft SQL Server installed, on another server you have Microsoft Exchange 2000 installed, and on a third server you have Microsoft SNA Server installed. Users often access these servers simultaneously.
If you used per server licensing, you would have to buy one license for each user for each server. If you had three users, you would have to buy three licenses for each server, for a total of nine licenses. With per seat licensing, you would need to buy only one license for each user, or a total of three licenses. Now you can see why per seat licensing is much more cost effective when users need to access multiple servers in an organization.

With per server licensing, you must have at least as many CALs dedicated to a server product as the maximum number of client computers that will connect to that product concurrently. For example, if you are logged on to a workstation and you connect to

\\server\apps

and

\\server\public

from that workstation, your actions constitute a single connection and require only one CAL. However, if you log on to two different workstations using the same username and connect to the server from both, it is considered two connections and requires two per server CALs. If a network has multiple servers, each server licensed in per server mode must have at least as many CALs dedicated to it as the maximum number of clients that will connect to it at any one time. After the limit is reached on a server, the server does not allow additional connections. Clients attempting to connect to the server receive an error message. Connections made by administrators are counted in the total number of concurrent connections, but after the limit is reached administrators are still allowed to connect. This permits them to manage a lockout situation. Other users can sconnect only after enough clients (including administrators) have disconnected to move below the limit.