16.1 Network Infrastructure Concepts
16.1.7 Network topologies
As previously mentioned, a topology is the physical way the components in a network are arranged or laid out. The primary topologies that exist are:

Bus - Early Ethernet LANs used a bus topology. A bus is a cable (coaxial) that is run through an area, and workstations and servers tap into the cable to get access to the network. The bus is a single length of cable that is terminated at each end. The bus design was prone to problems since a break anywhere in the bus could bring the entire network down. Although there are still many bus-based networks in service, new installation almost always use the extended star topology. Figure shows an example of a bus type topology.

Star - The star topology is the most common and is the one used with Ethernet hubs and switches. The hub or switch is the center of the star, and cables radiate out to nodes like spokes on a wheel. The biggest advantage of the star over the bus is that if a cabling connection to any node fails it does not effect the others. Two or more switches can be connected using backbone cable (usually fiber optic) which creates an extended star. The extended star is the most common topology used in today's networks. Extended stars are also used to create WANs.

Ring - The Token Ring and Fiber Distributed Data Interface (FDDI) network architectures use the ring topology. Token ring is commonly implemented as a star with a central concentration device known as a multi-station access unit or MAU and is technically a star ring. FDDI is a true ring used primarily to interconnect other types of LANs. FDDI is actually a dual ring with one acting as the primary and one as backup in case the primary fails. The ring topology is also found in WANs. FDDI can be used to create a WAN, although is it not commonly done. Another very widely used WAN technology that is based on the ring topology is Synchronous Optical Network or SONET. Figure shows a ring topology.

Mesh - The mesh is typically a WAN topology used to interconnect multiple LANs for redundancy of communications paths. They can also be used in LANs to provide redundant links between routers and switches. With a "full mesh", each site has a link to every other site. With a partial mesh each site has a link to more than one site but not to every site. Routers typically interconnect sites. Figure shows a partial mesh topology between LANs at four different locations or sites. Note that every site is not directly connected to every other site but every site has at least two connections to other sites in order to provide redundancy. This partial mesh uses four links between three sites so every site has an alternate route to get to another site if a WAN link goes down. The minimum number of links would be three: where LANs B, C, and D could all link back to LAN A, which would create a star topology. The maximum full mesh would require 6 links.