QoS and EtherNet/IP Control Traffic

QoS and EtherNet/IP Control Traffic
ABSTRACT
Industrial Automation devices, conforming to the EtherNet/IP model for communicating data via
the CIP protocol, transmit their data using well known TCP/UDP port numbers. Because of this, it
is relatively ease to create a QoS model that will elevate CIP based messages over other noncritical
Data. Placing Control traffic in the high priority expedite queue ensures that it will be
transmitted even in times of congestion. These QoS policies should to be added to the front line
access layer switches and need to be consistent with the remaining network equipment.
BACKGROUND
Industrial Automation devices, conforming to the Ethernet/IP model for communicating data via
CIP. The CIP protocol is layered onto of the standard TCP/IP network stack. It occupies Layer 4
or the Application layer in the OSI model. EtherNet/IP uses UDP/TCP port numbers which are
registered and often referred to as “well known”.
Application
Transport
Network
Data Link
Physical
FTP HTT
OSPF
Explicit IGRP
Messaging
TCP
Figure 1: EtherNet/IP Stack Model
SMT CIP SN
ICMP
UDP
IGMP
ARP IP RARP
IEEE 802.3
EtherNet/IP provides for several different types of communications, as outlined below:
BOOTP
DHCP
1. Unconnected – sent over TCP. Used for setting up connections (see below) and
communications such as getting status of devices.
2. Explicit Connections – sent over TCP. End-to-end CIP application connection,
established using Fwd_Open service (which is an unconnected message). Used for
application messaging, program upload/download, etc.
3. Implicit Connections – sent over UDP. End-to-end CIP application connection,
established using Fwd_Open service (which is an unconnected message). Used for
critical I/O communications.
IP
Real-time
I/O

The following port numbers are used as the Destination Port when a device sends EtherNet/IP
messages to another device:
TCP port 44818 – used for Unconnected and Explicit messages
UDP port 44818 – used for network browsing commands
UDP port 2222 – used for Implicit messages
The above ports may also be used as the Source Port. It is highly recommended that device
vendors use the registered ports. However, depending on how a device has implemented
EtherNet/IP, non-registered ports may be used. These ports are generally referred to as
“ephemeral” ports. For more information on registered ports see the IANA website at:
http://www.iana.org/assignments/port-numbers
Quality of Service Overview
It is very important in Factory networks to insure above traffic takes precedent over non critical
traffic such as web browsing or file transfer. To priority traffic, Intelligent Ethernet networks use
the Quality of Service (QoS) tags in the packet header. QoS can be marked in either the MAC
layer (Layer 2) or the IP (Layer 3) header.
For the Layer 2 header, this is referred as the Class of Service (CoS) bits. Three bits are used to
identify 8 levels of service. The bits are carried in the 802.1q Trunk Headers along with the VLAN
ID. When traffic crosses a Layer 2 boundary (i.e. going through a router), these bits may be
converted to the Layer 3 values below and then remarked on the next Layer 2 segment.
For Layer 3, they are called the Differentiated Services Code Point (DSCP) values. Six bits are
used to identify 64 levels of service. Generally, the upper 3 bits are mapped to the three bits in
the Layer 2 CoS tag and the lower three bits are used for more granular services.
Once traffic has been marked (or classified), it can then be routed correctly. Consider the
network to be a multilane highway. Traffic flows in each lane and different speeds and with
different priority. This fine when the traffic on the highway is light. However, what happens when
an ambulance or firetruck needs to make its way down a congested section of freeway. The
same is true here, a mechanism must exist to elevate and insure that high priority traffic makes it
way through the network, even if the network becomes congested.
Typically, Intelligent Ethernet devices contain multiple output queues where each queue is
assigned a certain priority level. By servicing the higher priority queues first, the important traffic
gets transmitted first. This is done by mapping the QoS values to the particular queues.

QoS example using the Catalyst 2955 Industrial Ethernet Switch
The Catalyst 2955 switch has four output queues. Each queue is linked to two CoS and 8 DSCP
Values:
Queue CoS Values DSCP Values
1 0,1 0-15
2 2,3 16-31
3 4,5 32-47
4 6,7 48-63
CoS values 6 and 7 are generally reserved for network control messages. Therefore, the highest
value for user data is CoS value 5 (DSCP value 40). In this example, the CIP traffic will be
marked as CoS value 5.
For CIP based messaging, Implicit messages using UDP port number 2222 have a higher
importance than either the explicit or network browsing commands using TCP/UDP port number
44818. In both cases, these messages should still have a higher priority over other network
traffic. To accomplish this, two classifications need to be made using standard Cisco Access
Control Lists (ACLs). Access list 101 will define the Implicit messages and access list 102 will
define all other CIP based messages
access-list 101 permit udp any any eq 2222
access-list 102 permit udp any any eq 44818
acesss-list 102 permit tcp any any eq 44818
Once the ACLs are created, a policer needs to be created to match these ACL and mark the QoS
values to assign the appropriate priorities.
class-map match-all CIP-IMPLICIT
match access-group 101
class-map match all CIP-Other
match access-group
policy-map CIP-Traffic
class CIP-IMPLICIT
set ip dscp 40
class CIP-Other
set ip dscp 32
The above set of commands tell the switch that any traffic that is matched by the ACL group 101,
will have its priority value (DSCP value) set to 40 and traffic matched by ACL group 102 will have
the value of 32. Since the switch will also map the Layer 3 DSCP value back to the Layer 2 CoS
value, then that traffic will also have CoS value of 5. Lastly, the policer needs to be applied to all
access ports that are connected to CIP based controllers. For each port, the follow command
needs to be added to the configuration.
Summary
service-policy input CIP-Traffic
Once these two policers have been created, all CIP traffic will be marked High Priority and will be
transmitted throughout the network using the High Priority Queues. Implicit messaging will have
the highest priority followed by the remaining CIP messages. All other traffic will remain at their
existing priority levels. In this example, Implicit messages will get assigned to the highest priority

queue, other CIP messages to the next highest, and all remaining network traffic can use the two
lower priority queues.
Authors
Kenneth Coley,
Technical Marketing Engineer for Industrial Ethernet Products,
Cisco Systems, Inc.
Brian Batke,
Senior Project Engineer for Embedded Software
Rockwell Automation