The changes in customer network requirements—in combination with the problems with collision, bandwidth, and broadcasts—have necessitated a new network campus design. Higher user demands and complex applications force the network designers to think more about traffic patterns instead of solving a typical isolated department issue. We can no longer just think about creating subnets and putting different departments into each subnet. We need to create a network that makes everyone capable of reaching all network services easily. Server farms, where all enterprise servers are located in one physical location, really take a toll on the existing network infrastructure and make the way we used to design networks obsolete. We must pay attention to traffic patterns and how to solve bandwidth issues. This can be accomplished with higher-end routing and switching techniques.
Because of the new bandwidth-intensive applications, video and audio being delivered to the desktop, as well as more and more work being performed on the Internet, the new campus model must be able to provide the following:
Fast convergence When a network change takes place, the network must be able to adapt very quickly to the change and keep data moving swiftly.
Deterministic paths Users must be able to gain access to a certain area of the network without fail.
Deterministic failover The network design must have provisions that make sure the network stays up and running even if a link fails.
Scalable size and throughput As users and new devices are added to the network, the network infrastructure must be able to handle the new increase in traffic.
Centralized applications Enterprise applications accessed by all users must be available to support all users on the internetwork.
The new 20/80 rule Instead of 80 percent of the users’ traffic staying on the local network, 80 percent of the traffic now crosses the backbone and only 20 percent stays on the local network.
Multiprotocol support Campus networks must support multiple protocols, both routed and routing protocols. Routed protocols are used to send user data through the internetwork (for example, IP or IPX). Routing protocols are used to send network updates between routers, which will in turn update their routing tables. Examples of routing protocols include RIP, Enhanced Interior Gateway Routing Protocol (EIGRP), and Open Shortest Path First (OSPF).
Multicasting Multicasting is sending a broadcast to a defined subnet or group of users. Users can be placed in multicast groups, for example, for videoconferencing.
QoS We need to be able to prioritize different traffic types.
The new campus model provides remote services quickly and easily to all users. The users have no idea where the resources are located in the internetwork, nor should they care. There are three types of network services, which are created and defined by the administrator and should appear to the users as local services:
Local services are network services that are located on the same subnet or network as the users accessing them. Users do not cross layer 3 devices, and the network services are in the same broadcast domain as the users. This type of traffic never crosses the backbone.
Remote services are close to users but not on the same network or subnet as the users. The users would have to cross a layer 3 device to communicate with the network services. However, they might not have to cross the backbone.
Enterprise services are defined as services that are provided to all users on the internetwork. Layer 3 switches or routers are required in this scenario because an enterprise service must be close to the core and would probably be based in its own subnet. Examples of these services include Internet access, e-mail, and possibly videoconferencing. When servers that host enterprise services are placed close to the backbone, all users would be the same distance from the servers, but all user data would have to cross the backbone to get to the services.