each direction (20 Mbit/sec total) between an attached device and the switch. For fast Ethernet, the full duplex
speed is 100 Mbit/sec in each direction (200 Mbit/sec total). Like traditional bridges, switches build and maintain
internal tables that map Ethernet addresses to ports. A packet received on one port is rapidly "switched" to the
appropriate output port, typically within microseconds. Advanced switches support a virtual LAN (VLAN) feature
that allows users to configure the switch so that ports are subdivided into groups such that all packets received on
one port of a group will only be transmitted to another port within the group. The receiving port and the group of
transmitting ports constitute a VLAN. VLANs may typically be overlapped within a switch, such that any one port
may appear on multiple VLANs. This feature allows the user a great deal of flexibility over partitioning the ports on
a switch into multiple overlapping collision domains. Switches are capable of handling a greater throughput than
repeater hubs without experiencing the collision-induced delays that can be experienced by devices on a repeater
hub as network traffic increases. This makes them a good choice for replacing repeater hubs on loaded networks
that are experiencing an unacceptable level of collision-induced delays. Although switches are currently more
expensive than repeater hubs, their cost is dropping and will soon be low enough that switches will likely replace
repeater hubs as the network concentrator of choice for all Ethernet networks, not just those used for control. It is
important to note that switches do have some performance limitations that may affect some applications. If a
switch experiences internal congestion due to message packets on multiple input ports contending for
transmission on the same output port, the switch may simply drop packets. Or it may force a collision back to the
transmitting devices, so they back off long enough for the congestion to clear. The approach that is taken depends
upon the implementation chosen by the switch vendor. In either case, a variable latency is inserted into the
message stream, which is generally not a problem for office applications but may have profound impact on
industrial automation applications. Although switches isolate separate collision domains on each port, they do not
create separate broadcast domains. However, each VLAN is a separate broadcast domain, if this feature is
enabled on the switch. An Ethernet broadcast message that is received on any port will be re-transmitted on all
switch ports to all attached devices. This means that switches do not eliminate the problem of excessive broadcast
traffic that can cause severe performance degradation across an entire Ethernet network when a damaged or
improperly configured device is attached to the network. Some switch vendors are working on proprietary methods
for suppressing excessive broadcast messages in their switches, but this is not universal. Broadcast messages
are common on Ethernet networks that carry the TCP/IP protocol because Ethernet broadcast messages are used
by TCP/IP for address resolution. However, broadcast traffic represents a small percentage of network traffic on a
network that is properly configured and operating normally. Also, switches and repeater hubs are active devices,
containing complex digital circuitry and requiring power (AC in most cases) to operate. The failure of a switch or
hub will effectively cause a communication failure for all of the devices attached to that device's ports, including
other hubs or switches that may be attached to one or more ports of the failed device. The devices attached to the
failed hub or switch will be unable to communicate with the rest of the plant network until the switch is replaced or
repaired. Furthermore, most Ethernet media components have been designed for use in an office or light industrial
environment. They have not been designed and tested for compliance to the rigorous environmental standards
typical of industrial control devices (i.e., extended temperature range, industrial CE mark, shock and vibration,
etc.). This may become an issue as the mission for Ethernet on the plant floor is expanded into new areas.
C. The Evolution of Ethernet Performance More recent developments in Ethernet technology include Fast Ethernet
and Gigabit Ethernet. Fast Ethernet is defined and documented in IEEE specification 802.3u. Fast Ethernet is
basically Ethernet running at 100 Mbits/sec. Fast Ethernet and 10 Mbit Ethernet use the same frame structure,
addressing scheme, and CSMA/CD access method. However, all network timing parameters must be scaled by a
factor of 10 when configuring a Fast Ethernet network. This tends to reduce the distances between nodes in some
configurations when compared to a 10 Mbit network. Fast Ethernet provides a wire speed that is 10 times as fast
as traditional Ethernet, which tends to benefit bandwidth hungry applications, such as video and audio
transmission, as well as the transfer of large data files over the network. However, most applications will not enjoy
a substantial performance increase due to increased wire speed alone. In particular, a plant-floor network of small
microprocessor-based intelligent I/O blocks, sensors, actuators, drives and other device interfaces are likely to
consume and produce small amounts of data encapsulated in 64 byte Ethernet frames (the smallest frame size
supported by Ethernet). The performance of these devices is more likely to be limited by the speed of their
microprocessor and embedded firmware than the wire speed. It is unlikely a network of such devices would fully
utilize the full 10 Mbit/sec Ethernet bandwidth, unless an inefficient application layer protocol was utilized that
repeatedly polled the devices in point-to-point fashion. One area of performance wherein 100 Mbit Ethernet may
show noticeable improvement over 10 Mbit Ethernet is in the area of collision recovery. As mentioned earlier, the
backoff times for 100 Mbit Ethernet are one tenth of those for 10 Mbit Ethernet. On a loaded network where
collisions are an issue, 100 Mbit Ethernet may show noticeably better performance than 10 Mbit Ethernet.
Additionally, it would be expected that a 100 Mbit Ethernet network would be able to handle a larger offered load
than a 10 Mbit Ethernet network before collisions became an issue. If the application requires the use of multiple
switches, the links between the switches may benefit from the higher speed. However, if loading and collisions are
not already an issue on a 10 Mbit Ethernet network, simply upgrading to 100 Mbit Ethernet may not show
sufficient improvement to justify the investment.
D. Implicit (I/O) Messaging over EtherNet/IP In section 4.5
, we discussed the communication path and the use of
explicit messages and unconnected messages to exchange point-to-point messages between nodes. The
second type of messaging, implicit messaging is used where the exchange of data between nodes is transparent
to the user and takes place within the application layer of the protocol, with both producing and consuming nodes
aware of the content of the message before transmission. While commonly used for I/O messages, these make