Most examples given thus far show only small EtherNet/IP networks. They do so for simplicity. Larger EtherNet/IP networks are possible as well. This topic discusses a few issues that come to the forefront with large systems. The terms "large" and "small" are ambiguous, but similarly, the concepts really apply to both large and small networks. An example of a small systems is one to four RMCs connected to a single 1756-ENET or 1756-ENBT. An example of a large system is 40 RMCs connected to one or, ideally, more 1756-ENBT modules.
Reduce Bandwidth Usage to Actual Requirements.
Suppose you need to control 40 RMCs from a single ControlLogix 1756-L1. If you use a single 1756-ENBT for this task, the bandwidth required for this system at a 5 ms RPI is calculated as follows:
Frames/Second |
= |
(2 x connections) / RPI |
|
= |
(2 x 40) / 0.005 s |
|
= |
16,000 |
This is well over the allowed 4500 frames/second on the 1756-ENBT. However, if the system can get by with a slower RPI, the bandwidth drops dramatically. For example, increasing the RPI from 5 to 20 ms reduces the bandwidth requirement as follows:
Frames/Second |
= |
(2 x connections) / RPI |
|
= |
(2 x 40) / 0.020 s |
|
= |
4,000 |
In many cases, the reduction in bandwidth comes at no cost to the system performance because it may be likely that the ControlLogix is unable to scan its ladder logic more frequently than every 30 ms, in which case most 5 ms updates are ignored.
Notice that in some applications, the RPI required for different RMCs may be different. For example, suppose that three of this group of forty require a 5 ms RPI, but the rest can get by with a 25 ms RPI. Since the RPIs are set independently for each RMC, the bandwidth can be further reduced as follows:
Frames/Second |
= |
(2 x connections) / RPI + (2 x connections) / RPI |
|
= |
(2 x 3) / 0.005 s + (2 x 37) / 0.025 s |
|
= |
1,200 + 2,960 |
|
= |
4,160 |
Divide the Network.
You can further improve the stability and determinism of your system by dividing up the system into multiple networks. For example, we could reduce the near-maximum load of 4,000 frames/second to a comfortable 2,000 frames/second by adding a second 1756-ENBT module. By keeping the two networks separate (that is, use a separate switch for each), the collisions are also reduced drastically. For a large Ethernet system, the additional cost of a second, third, or even fourth 1756-ENBT module (around $1000 each as of this writing) is often insignificant compared to the total cost of the system and gives much higher reliability.
Upgrade to Smarter Switches.
EtherNet/IP utilizes IP multicasting, and as such uses a protocol called IGMP (Internet Group Management Protocol). Most low-cost switches do not utilize IGMP to control which ports care about the multicast packets, but instead broadcast multicast packets to all ports. This increases the load on all segments of the network, which will increase collisions on half-duplex segments and increase the possibility of overruns. By choosing switches that support IGMP (often called IGMP Snooping), you can reduce the load on your network components and thereby increase your determinism. However, notice that most of these switches require a router to be present in the network for the IGMP snooping to work. Therefore, if your system will not always have a router present, then you should purchase a switch that supports IGMP snooping without a router present.
Copyright (c) 1997-2015 by Delta Computer Systems, Inc.