The Star topology has, for the most part, replaced the ring topology. However let's look at each in a little depth and see the evolution of where they came from.

First, let's not confuse our physical layer protocol with our network protocol. So I can use TCP/IP and at the physical layer it can be transported on Ethernet which is a collision topology, ATM, FDDI, or many other physical layer protocols.

At one time (late 80's/early 90's) the star topology was a weak topology because the hierarchy was not advanced enough to support larger networks. Routers sucked and switching was not available so if you had 20 computers on a network anytime one of them sent a message out the connection from all of them were blocked for that short period of time. The more computers the greater the chance of collision and the need for retransmission was. The accepted max on a network was around 120 while the suggested was no more than 24. The main method for segmenting a network was the bridge (see also brouter for an interesting historical term) which for star systems was fairly expensive. Routers were available but most didn't work well or were very unstable.

At this same time the Token Ring topology came out. Because it did not use a CSMAC/CD (Carrier Sensing Media Access Control/Collision Detection) system like Ethernet does it could use a higher percentage of the bandwidth. For comparison Ethernet can use up to about 60% of its rated bandwidth before collisions start reducing the effectiveness of the network while Token Ring can effectively use 95% of its rated bandwidth. This is sustained usage not burst or spike.

The way Token Ring operates is that each system when it enters the network sends out a token announcing it presence and it locates its down stream neighbor and its upstream neighbor. Then it becomes part of the ring. Assume the ring has 3 computers on it A,B,C. The ring is established so A sends a message and at the end of the message passes the token along. B gets the message, uses it if needed, retransmits it, finds the token and adds its own message. C receives A's, reads, retransmits, receives B's, reads, retransmits, sees token and adds its own message. A receives its own removes it from the ring, receives B's, reads, retransmits, receives C's, reads, retransmits, sees token and can add its own packet. This continues. No collisions ever happen because the retransmit is always to the next computer up the chain and everything runs like a train.

If two networks have to connect then one of the devices must operate as a Bridge or Gateway (bridge if between like systems, gateway if between dis-similar systems). With this scenario if Ring A wants to communicate with Ring B it sends the message around which is checked and retransmitted on Ring A until it reaches the bridge. The bridge then retransmits it on A and B however on B it makes itself the sending agent. On ring A the packet is removed when it returns to its owner on Ring B it is removed when it returns to the bridge. In this way every device that could possibly want the packet gets it on both rings.

In this manner the original 4 Mbps Token ring could operate as fast or faster than a 10 Mbps Ethernet. When the 16 Mbps Token ring came out it blew the Ethernet out of the water for speed and capability. The drawbacks, however, were that any time a machine was added or removed the ring had to be broken. Ring restoration is fairly quick but this meant that if one system went down the ring went down until it was removed or replaced. However, Token Ring proved to be extremely reliable once set up properly. I have never set one up but what I was told by a few experts is that it was difficult to get them working exactly right but once they were they seemed to work forever or until messed with.

In an attempt to fix some of the inherent problems with the ring the dual ring topology was implemented. FDDI is the best example of the dual ring and while I have worked with FDDI and it seems viable I loathe it with a passion. The theory here is that there is a primary and secondary ring. Data flows on the primary only while the secondary merely maintains itself with beacon packets going in the opposite direction of the Primary ring. If a station goes down then the neighbors to that station loop the primary to the secondary and a single ring is created to re-establish a path. This way hardware can be removed or added and the ring will self heal itself and re-establish full operation automatically. In practice where I worked FDDI proved to be unstable at times and very difficult to troubleshoot. Also its speed (100 Mbps) and range (total ring distance of 100 Km) were overwhelmed by ATM and Gigabit Ethernet (Thank God).

Now with the advent of cheap hubs, switches, routers, and the ASIC (Application Specific Integrated Circuits) the Star topology is much better than the ring. Devices can be added and removed without affecting the topology. Collision domains can be easily created to limit network interference. Troubleshooting becomes much simpler because you can trace from router to switch to port and finally to device. The Star topology can also be built with easily created redundant pathways that lead to a more fault tolerant network. This has only been really available though in the last 5-8 years. Before that is when the real battle between Star vs Ring occurred.