Installation guide

Network Infrastructure for EtherNet/IP™
6-66
Infrastructure Application Scenarios
Within each of these four network types, recommendations can be made regarding the types of active
infrastructure components (switches, routers, etc.) and the key functions needed to balance real-time
performance with data connectivity requirements. Appendix A Recommendations for Ethernet Switches in
EtherNet/IP Systems complements this chapter by summarizing the recommended switch functions for
each of the four system types.
6.1 General Application Considerations
There are several key concerns that vary based on the type of system desired.
6.1.1 Multicast Traffic Management within a Control System
Data producers on an EtherNet/IP network can generate a significant amount of multicast traffic,
requiring that particular care be taken to manage it. Managing multicast traffic is the most important
concern in the design of a high-performance EtherNet/IP network. With EtherNet/IP, implicit messages
(I/O control and controller-to-controller communication) use unicast (one-to-one) packets to communicate
output information from a controller to I/O or other control devices. Input information, sent from these
devices to the controller, is usually transferred using multicast (one-to-many) packets. As noted in 4.7
Switching Technologies, switches normally retransmit multicast packets to all switch ports (flooding), so
that all devices see all the multicast traffic. Without multicast control functions, such as IGMP snooping,
the end devices’ operation is impaired as its CPU becomes overloaded due to discarding extraneous
multicast traffic.
Requested Packet Interval
One of the major controller or scanner setup parameters is the Requested Packet Interval (RPI).
This value is the minimum rate at which control devices will produce the input information needed
by the controller. The larger the application (and the number of interconnected control devices) and
the faster the response time needed (lower RPI), the greater the need for IGMP snooping.
The tradeoff here is that managed switches (those typically with web/SNMP diagnostics and message
filtering functions, such as IGMP snooping, VLAN, and others) cost more than simple unmanaged
switches, which do not have message filtering functions, such as IGMP snooping. Systems with the
highest performance requirements and the fewest application timing problems will use all managed
switches with IGMP snooping.
6.1.2 Control Device Sensitivity to Extraneous Traffic
Another consideration related to multicast message control relates to each control device’s sensitivity to
unwanted multicast and broadcast traffic. The impact of unwanted traffic is analogous to sorting the “junk
mail” from the important mail.
6.1.2.1 Potential for Traffic Overload
Every time a control device receives an Ethernet frame, it takes a tiny bit of its microprocessor/Ethernet
chip resource to look at the address, decide if the message is for it, and discard messages addressed to
others. The rate of incoming unwanted messages (expressed in packets per second) can reach a point
where so much time is spent processing and throwing out junk mail that it can not produce or receive
control messages at the configured RPI value. This will cause an error condition that shuts down
communications and causes the system to go into its fault mode.