Ring Topology and the Dreaded Loop Condition

It would appear that a ring topology would be the logical choice for redundancy since a break at any point along the ring would still leave all stations connected. However, Ethernet does not inherently support this type of topology since it would result in an endless loop condition.

Switches store-and-forward frames. A frame received on one port is forwarded to the port indicated in its address table. However, a broadcast frame that is intended to all other stations is flooded to all ports on the switch. The next switch in line will do the same and eventually the broadcast will return to the originating switch repeating the process. This situation will continue endlessly, completely consuming bandwidth, until one of the ring segments is broken. This situation needs to be avoided if ring topology is to be employed.

Proprietary Ring

Several vendors have developed their own implementation of a ring topology that continues to function when one segment within the ring is broken. Typically, all switches are connected to one another through designated ring ports. One of the switches is assigned as master and blocks the port that would result in a loop condition. Each switch monitors the condition of its ring ports and reports to the master if one of its ring ports has failed. Upon receiving a report of a port failure, the master enables its backup port and instructs all switches to clear their address table. This rapid aging of each address table is to force the switches to relearn the new station-port assignments since there has been a topology change. Once the fault condition is corrected, the master is informed accordingly. The master then informs all switches to again clear their address table; the master disables the backup port and instructs the switches to enable their primary ring ports.

The disadvantage is that the schemes are proprietary requiring that all equipment come from the same vendor. Some implementations require a separate redundancy manager. This scheme only operates as a ring and sometimes ring topologies are not convenient to implement in a factory.

Spanning Tree and Rapid Spanning Tree Protocols

The standardized method of supporting alternate cabling paths is explained in IEEE 802.1D Media Access Control (MAC) Bridges. Two protocols have resulted from this standard. The original was the Spanning Tree Protocol and the second is the Rapid Spanning Tree Protocol. STP is based upon timers and will operate with link segments as well as coaxial cable segments. RSTP provides much faster recovery times because its protocol examines the link status of ports and it has reduced the number of states within the spanning tree algorithm to just three: forwarding, learning, and discarding. It will only operate with link segments. Both protocols are very complex, but they will operate with ring or mesh topologies.

The two protocols will interoperate. Bridges communicate to one another by sending out Bridge Protocol Data Units (BPDUs). The revision number of the BPDU identifies either STP or RSTP. Within the BPDU, there is information about the topology of the network. Bridges determine the best path to the root bridge based upon path cost which is reported by the various bridges. One bridge is elected the root bridge (based upon the lowest value Bridge ID) while all other bridges revert to designated bridge status. Alternate paths are identified with one of the connected ports reverting into the discarding state so as not to create a loop. Upon a topology change due to a failed connection, the discarding port reverts to forwarding and the alternate path is enabled.

The advantages of STP and RSTP are that they are part of an IEEE standard that is well supported by the various vendors allowing for the mixing of vendor equipment in the field. Both protocols are not limited to just ring topology. A mesh topology will work as well. The protocols could also be used in a star topology to guard against cabling mistakes that can result in a loop. RSTP has a much faster recovery time then STP since it is not solely based upon timers.

The disadvantages to either STP or RSTP is that both protocols are very complex to understand and difficult to configure. Parameters may need to be "tuned" in the field in order to yield acceptable recovery times. STP has a very slow recovery time and represents the "old" standard. Depending upon the application, the STP recovery time may be unacceptable.