Designers in many fields (architecture, software, and engineering, for example) rarely approach a design problem by re-inventing the wheel. Instead, designers topically apply design patterns. Design patterns are typical solutions to common problems in design. A pattern is a standard blueprint that can be used to solve a particular design problem. However, since design problems are rarely identical, the initial design pattern may be customized and modified to solve the particular problem at hand.
There are several patterns in network design. The most common is the hierarchical model that divides a network into two or three discrete tiers. Each tier, provides specific functions within the overall network, such as connectivity to end nodes or bulk data transport. The hierarchal model apply mostly to LANs, but it often (ab)used to solve other design problems due to its popularization by the vendor that created it.
A spine-leaf pattern is used in data centre designs. The pattern includes two (or more) tiers of network devices forming a spine and leaf. The leaf devices connect to end nodes (servers) to aggregate and pass traffic to the spine devices. Although the spine-leaf topology may resemble the hierarchical model, there many notable differences, including avoiding bandwidth oversubscription by ensuring equal bandwidth at all tiers and avoiding the use of the Spanning-Tree Protocol (STP) to facilitate multi-path forwarding of data traffic.
Other patterns may not be as common due to their specialized use. Service providers’ networks are constructed as a partial mesh topologies connecting several Point-of-Presence (POP) locations (which could be data centers). A centralized tree topology is used often to deliver broadband network services to residential and business customers. The tree is constructed by connecting customers to an aggregation point (e.g. curb node), which in turn is connected to another aggregation point until all traffic is aggregated at the service provider’s head end.
Read about other Network Design Principles.