Upgrading and Repairing Networks

Multi-Tiered Network Topology

By combining what you have learned so far about network topologies, you can develop a design that can meet most of the objectives mentioned in the preceding sections concerning the campus network design. Several powerful features of the new campus design appear in Figure 2.14.

Figure 2.14. The partial mesh topology can be used to provide scalability, fault tolerance, and redundancy.

[View full size image]

Note that the design is a partial mesh star topology. In the next few sections we'll look at how this topology can be used to provide for scalability, redundancy, and fault tolerance for your network.

Scalability

With the advent of a distribution layer, you have the ability to offload some of the network capacity requirements at a lower level of the network. For example, if you placed your file servers at the distribution layer, the backbone layer would be freed of significant burden and could focus its resources on cross-building and data-center level traffic requirements. The distribution layer would turn around about 80% of all the network traffic and respond directly to the client layer, speeding network access. By offloading file and print servers at a lower level in the network, you would have a great deal of scalability.

Note the mesh between the distribution layer and the backbone layer. You could use a high-speed network technology such as gigabit Ethernet to deliver the aggregate bandwidth you needed at the top layers of the network to process whatever requests were coming out of the client layers. Having separated this mesh from the client layer switches means you could slip higher-speed technologies into the mesh as they came along without costly downtime to your clients and expensive rewiring of your buildings.

Finally, by using VLANs from the distribution layer down into the client layer, you can ensure smaller broadcast domains to further drive down network use.

Redundancy

Significant gains in redundancy have been added by inserting a distribution layer of switching technology between client layer switches and backbone layer switches. This design would not have been possible with hub technology because the redundant connectivity would cause loops in the network, something a switch controls with a process called the Spanning Tree Protocol.

A failure at the distribution layer causes the client layer switches to fail-over to a secondary distribution switch. Likewise, if a failure occurs at the backbone layer, the traffic reroutes to the secondary backbone switch. Another nice feature is that because the distribution and backbone have a primary and secondary switch device, you can dual-home your most critical servers at these layers and still have network connectivity if you have a NIC card failure on your server.

Fault Tolerance

Fault tolerance is the aspect of your network that defines how resilient it is to the various problems that will inevitably crop up. Keeping the fault tolerance of the design in mind, consider just a few questions on this topic:

• Could a problem with a single workstation negatively impact the entire campus LAN?

• How would the LAN environment react to a broadcast storm?

• Do you have any isolation features to keep problems from propagating throughout the campus environment?

By using VLAN technologies, you can control the spread of broadcast, which provides an extra layer of control and fault tolerance to your network. Also, by incorporating some layer-3 switching technology, you can provide further isolation for robustness. Recall that routing does not forward segment broadcasts, so by using layer-3 switching, you also provide a means of isolating problems to a particular segment.

As you can see, a combination of old and new topologies and technologies provides a solid foundation that can be scaled to great size without sacrificing stability and performance.