Understanding Multi-Tiered Network Design

This covers designing networks with multiple tiers and redundancy.

DESIGNING A NETWORK

When designing a network, uptime, resiliency, redundancy, performance and manageability are the main things to design into the network. 

If you build a network with lots of redunancy then that will prevent a single point of failure causing a network outage. Redundancy allows the network to continue to run even if there is a link or network device down. The network can withstand outages while allowing end users to continue to work and give the network team time to resolve the issue. This resiliency will ensure the uptime is good within the network and lowers the impact of issues.

Redundancy can also improve performance. Multiple links between network devices can be logically made to act as a single link which allows traffic to load balance across the links for higher throughput. If a link does go down then spanning tree won't need adjust the tree and root ports, allowing traffic to continue to flow.

SIMPLE NETWORK DESIGN

Below is an example of a simple network. There's a router, layer 3 switch and layer 2 switches for end devices. Now this network can work, but if a link were to go down, or the layer 3 switch goes down then the network won't be able to carry on running.

TIERED NETWORK DESIGN

Now with redundancy then the network could look like below where the network is tiered. End devices connect to layer 2 switches at the access layer, if traffic needs to flow further through the network it can go up to the distribution layer then down to other layer 2 switch or up to the routers for internet access.

If a link or layer 3 switch goes down then the redunadant link can be used to avoid disrupting the flow of traffic and uptime within the network.

Segmenting the network like this makes it easier to manage. It will be easier to understand the flow of traffic and can be easier controlled with ACLs as well as adding in controlled segments like DMZs. 

FURTHER REDUNDANCY

You can go even further with stuff like clustering, the access layer switches can be clustered to logically behave as a single switch, same with the layer 3 switches clustered into a single logical switch. The routers can then use redundancy protocols such as HSRP, VRRP, GLBP to allow a router to failover if one goes down.