Part of the traffic that must be considered during design is going to or coming from the Internet. Since the Internet probably exists everywhere that the enterprise has LANs, there are two principal ways that this traffic can be carried. Each LAN can have a connection to its local ISP, or there can be a single connection from one of the core routers to an ISP. The advantage of the first method is that traffic is carried on the Internet rather than on the enterprise network, possibly leading to smaller WAN links. The disadvantage of permitting multiple links, is that the whole enterprise WAN is open to Internet-based attacks. It is also difficult to monitor and secure the many connection points. A single connection point is more easily monitored and secured, even though the enterprise WAN will be carrying some traffic that would otherwise have been carried on the Internet.

If each LAN in the enterprise has a separate Internet connection, a further possibility is opened for the enterprise WAN. Where traffic volumes are relatively small, the Internet can be used as the enterprise WAN with all inter-branch traffic traversing the Internet. Securing the various LANs will be an issue, but the saving in WAN connections may pay for the security.

Servers should be placed closest to the locations that will access them most often. Replication of servers, with arrangement for off-peak inter-server updates, will reduce the required link capacity. Location of Internet-accessible services will depend on the nature of the service, anticipated traffic, and security issues. This is a specialized design topic beyond the scope of this curriculum.