In an industrial automation application that relies heavily on the health of the Ethernet network
that attaches all the controllers and computers together, a concern exists about what would
happen if the network fails. If the result is loss of production or loss of a processed batch
or the endangerment of people or equipment, redundancy schemes are examined. Since
cable failure is the most likely mishap, cable redundancy is suggested by configuring the
network in either a ring or by carrying parallel branches. If one of the segments is lost,
then communication will continue down a parallel path or around the unbroken portion of the
ring. The problem with these approaches is that Ethernet supports neither of these topologies
without special equipment. However, this issue is addressed in an IEEE standard numbered
802.1D that covers bridges, and in this standard the concept of the Spanning Tree Protocol
(STP) is introduced.
The redundancy issue is addressed by ANSI/IEEE Std 802.1D, 1998 edition: Information technology
Telecommunications and information exchange between systems Local and metropolitan area
networks common specifications Part 3: Media Access Control (MAC) Bridges. The equipment
covered by the standard is called a bridge. A bridge is used to connect two or more local area
networks (LANs) at the MAC level, which is layer two in the ISO Reference Model. Generally the MAC
type can be different on each LAN, but usually Ethernet LANs are on either side of a bridge.
Interconnecting LANs by using bridges creates a Bridged LAN whereby end stations located on
different LANs can communicate as if the bridges were not present.
Cable redundancy introduces loops in the topology and, as we will see, these loops must be disabled. An industrial
automation user may want loops to guard against a primary cable failure while an office automation user may want to guard against an inadvertent loop. The 802.1D standard
addresses both situations.
If you understand how an Ethernet switch works, you know how a bridge operates. However, all the requirements of a
bridge (e.g., STP) are not always present in a switch. A bridge needs to relay and filter frames and it must make independent decisions about when to do this.
Look at Figure 1. In a two-port Ethernet bridge, each port has an Ethernet-type MAC
port connected to a separate LAN and a filtering database (memory) shared by
both ports. Within each LAN is a collection of end stations, repeating hubs
and simple plug-and-play switches. Each end station has a unique MAC address.
For simplicity, we will assume ordinary integers although true Ethernet MAC
addresses are 48 bits long. In our example, three numbered end stations are
present in each LAN. Assume Bridge 1 has recently been powered and its memory
cleared (Bridge 2 will be added later). Station 1 sends a message to station 11
followed by Station 2 sending a message to Station 11. These messages will
traverse the bridge from one LAN to the other. This process is called relaying
or forwarding. The database in the bridge will note the source addresses of
Stations 1 and 2 as arriving on Port A. This process is called learning. When
Station 11 responds to either Station 1 or 2, the database will note that
Station 11 is on Port B. If Station 1 sends a message to Station 2, the bridge
will do nothing since it realizes that because Stations 1 and 2 are on the same
LAN their message does not need to be shared with other LANs. This process is
called filtering. If Station 1 ceases to initiate messages for a period of
time, the bridge will erase Station 1 from its database requiring the
location of Station 1 to be relearned. This is called aging.
Figure 1 The addition of Bridge 2 creates a loop.
The above examples are all directed or unicast messages meaning that one station
is sending a message to another station. With multicast (one station to many
stations) or broadcast messages (one station to all other stations), the bridge
will forward messages to all stations since it may not know the actual location of
all stations. This process is called flooding.
Looking at the same Figure 1, Bridge 2 is now added to parallel Bridge 1. This gives us a
redundant path, but it also creates a loop with the following adverse results. When
Station 1 initiates a message to Station 11, this message is forwarded by Bridge 1
and appears on LAN B. Bridge 2 interprets this message as originating on LAN B so it
forwards the message to LAN A while incorrectly noting that Station 1 is located on
LAN B. When Station 1 initiates a second message, Bridge 2 interprets this action as
if Station 1 has now moved to LAN A from LAN B and resets its filtering database
accordingly. Now assume that Station 1 sends out a broadcast message. Bridge 1 will
forward the message to LAN B. Bridge 2 will observe the message on LAN B and forward
it to LAN A. Bridge 1 will observe this message on LAN A as a new message and forward
it to LAN B again initiating an endless cycle, totally consuming the bandwidth
of both bridges and rendering both LANs useless. To maintain the integrity of our
network, we must guard against the formation of loops.
(No part of this article may be reproduced without the written consent of the Industrial Ethernet University.)