Network topology???

Acidburn I would have to agree with you on the star with many stipulations.

In a hub/sharing system everyone in the Collision Domain has access to the bandwidth at all times. Therefore since it is a first come first served basis a single computer on the network can bog down the entire network for everyone. This normally does not happen with standard data since it uses a connection oriented transmission control (TCP in TCP/IP). However many functions using a connectionless (UDP in the TCP/IP stack) protocol can flood the network and cause a DoS to other users. In fact this is the whole point between a DoS attack or a DDoS although these normally occur with incoming rather than generated data.

The most common cause of DoS on a shared star is a bad NIC. These often send runts and orphans or just flood the network. I have seen this on a number of occasions.

A switched Star does not fall prey to this if the switch is setup correctly. With a layer 3 switch data sent to a system on one port goes to that port only and does not affect traffic on the other ports. In this instance the only way to do a DoS to the entire Star would be to flood the backplane of the switch which generally has considerably higher bandwidth than any one port (ideally it would be equal to the max bandwidth of all ports, normally it is equal to the bandwith of more than 50% of the ports). A single port may be affected by bad NIC but not the entire switch.

Unfortunately since most DoS or DDoS attacks use UDP and broadcasts the switch is not as immune as one would like. Broadcasts are meant for every computer and therefore switches send the data to every port. This is where expensive switches with good ASICs (Application Specific Integrated Circuits) shine over cheaper switches. They can be set to block or forward certain types of Broadcast and Multicast signals.

I guess in short it is important to seperate the Topology from the capability and vulnerability. Topology is a layout scheme and currently the Star topology is King. Also don't forget that topologies can be mixed. I think the best overall topology set up is a Semi-Meshed topology for the backbone and a Star for the end users. So your connections between buildings, sites, etc. would not rely on one path but have a minimum of two paths so any one link could go down and no major location would be lost to the network. As you get down to the user level the added cost of meshing is not that critical. The time to remove a switch or fix an individual's problem is not as critical as that same time when fixing 100 people's problem.

So while you study topologies keep in mind you also have to think about implementation of that topology, mixing it with other topologies and implementations and then the protocols and methods used to access the physical topology (hierarchical/flat, Switched/Shared/Routed, Controlled Traffic/Collision Domain, etc). At first this may seem very confusing but think of it like putting lego's together: each lego is a piece of your network, a protocol, topology, or requirement. They can all (through the right equipment) connect to each other to build your network which is only limited by your imagination and what you seek to create.

Does any body even use Token Ring anymore?

Yes, there are a lot of Token Ring networks out there. They tend to be considered legacy now but they are still going. It follows the old, and wise, saying that if it isn't broken don't fix it.

Once a Token Ring network is set up and running they have proven to be very, very stable. So if you needs are fulfilled by a 4 or 16 Mbps network why spend thousands of dollars to switch to a 100 Mbps Ethernet setup which is going to really only effectively double or triple your real bandwidth. Most networks run great on 10 Mbps internally with a T-1 or less going out WAN. It is only in large networks (or in the dreams of us IT professionals) that the 100 Mbps+ networks are required.

I knew one person that wanted OC-12 (622 Mbps ATM) to his desktop. Why? There was no logical reason or requirement but that is what he wanted. Many networks are not designed to current, or even future, need but to current and future technology.

So, yes there are Token Ring networks still running. Lots of them. When you see one you will have to decide if it points to someone afraid to embrace technology (bad) or frugal and wise enough not rush to new technology for its own sake (good). However, the real question and one I cannot answer is: is anyone installing new Token Ring networks and can the equipment still be purchased?

A production comany in this area recently (in the last 3 years) installed a token ring in their plant. I don't know much of the details about it, but it was cheaper then a star/bus, and it works perfect. It allows the production macines to communticate properly, becuase they can not afford to have packet collisions which would mess up the timing on the machines, and it always works in a certain order (ie machine 1 always talkst to machine 2... and so on.