• Selecting an interconnection pattern or layout for the links between the various locations
• Selecting the technologies for those links to meet the enterprise requirements at an acceptable cost

Many WANs use a star topology. As the enterprise grows and new branches are added, the branches are connected back to the head office, producing a traditional star topology. Star end-points are sometimes cross-connected, creating a mesh or partial mesh topology. This provides for many possible combinations for interconnections. When designing, re-evaluating, or modifying a WAN, a topology that meets the design requirements must be selected.

In selecting a layout, there are several factors to consider. More links will increase the cost of the network services, and having multiple paths between destinations increases reliability. Adding more network devices to the data path will increase latency and decrease reliability. Generally, each packet must be completely received at one node before it can be passed to the next. A range of dedicated technologies with different features is available for the data links.

Technologies that require the establishment of a connection before data can be transmitted, such as basic telephone, ISDN, or X.25, are not suitable for WANs that require rapid response time or low latency. Once established, ISDN and other dialup services are low latency, low jitter circuits. ISDN is often the application of choice for connecting a small office or home office (SOHO) network to the enterprise network, providing reliable connectivity and adaptable bandwidth. Unlike cable and DSL, ISDN is an option wherever modern telephone service is available. ISDN is also useful as a backup link for primary connections and for providing bandwidth-on-demand connections in parallel with a primary connection. A feature of these technologies is that the enterprise is only charged a fee when the circuit is in use.

The different parts of the enterprise may be directly connected with leased lines, or they may be connected with an access link to the nearest point-of-presence (POP) of a shared network. X.25, Frame Relay, and ATM are examples of shared networks. Leased lines will generally be much longer and therefore more expensive than access links, but are available at virtually any bandwidth. They provide very low latency and jitter.

ATM, Frame Relay, and X.25 networks carry traffic from several customers over the same internal links. The enterprise has no control over the number of links or hops that data must traverse in the shared network. It cannot control the time data must wait at each node before moving to the next link. This uncertainty in latency and jitter makes these technologies unsuitable for some types of network traffic. However, the disadvantages of a shared network may often be outweighed by the reduced cost. Because several customers are sharing the link, the cost to each will generally be less than the cost of a direct link of the same capacity.

Although ATM is a shared network, it has been designed to produce minimal latency and jitter through the use of high-speed internal links sending easily manageable units of data, called cells. ATM cells have a fixed length of 53 bytes, 48 for data and 5 for the header. ATM is widely used for carrying delay-sensitive traffic. Frame Relay may also be used for delay-sensitive traffic, often using QoS mechanisms to give priority to the more sensitive data.

A typical WAN uses a combination of technologies that are usually chosen based on traffic type and volume.  ISDN, DSL, Frame Relay, or leased lines are used to connect individual branches into an area. Frame Relay, ATM, or leased lines are used to connect external areas back to the backbone. ATM or leased lines form the WAN backbone.