Looking at the controller and I/O segment a little closer...
micky said:
...Communications between PLC and remote racks is crucial...
If we are talking crucial as in I/O traffic...
Considerations need to be made if using unmanaged switches between a controller and I/O under its control...
As a general rule for unmanaged switches, make sure of the following:
Your application does not contain I/O traffic
or
If your application does have I/O traffic, then make sure the following is true:
The network is not directly connected to the IT network
All nodes on the network are Rockwell Automation devices
There is no potential to overload a device with traffic
If the above can be satisfied, then...
Traffic-wise, the Linear setup you've outlaid will "most likely" work fine.
Resiliency-wise, the middle 1783-US5T switch going down will take down its local I/O, and also the second 1783-US5T switch and its local I/O.
If ye can live with that, it "should" be ok, but remember, each hop is adding latency to the packets on the wire. If you were getting any CRC errors or timeouts, the Topology would be one of the first places to look.
Also, the advantage of having embedded switch technology in the AENTRs and the CompactLogix is being negated, to a degree, by the external use of Stratix 2000 unmanaged switches (1783-US5T) local to each of those nodes. The Stratix 2000 switches are adding an extra hop at each node. They are also devoid of any diagnostics features normally present in a managed switch. They cannot manage or constrain Multicast data. The embedded switches can cut "Through Forwarding" times to reduce latency. Also, they use "Broadcast rate limiting" for DLR devices when the Broadcast traffic is excessive. This feature prevents end devices from becoming overwhelmed by network noise. Another useful feature is the filtering of incoming Unicast and Multicast frames to the DLR device. This feature prevents data that is not directed to the end device, but is passing through the embedded switch, from being processed by the device.
What degree the unmanaged switches reduce the capabilities of these embedded features is hard to say, but the DLR devices were specifically designed to wire in and wire out for optimum performance. Of course, having local port access is often a necessity, but I would advise managed switches here if needs must. The Stratix 5700 managed switches would probably be the best fit here.
However, all said, I would still prefer a Star Topology as a minimum, for the least latency and for better resiliency. You could go one better and Ring the DLR devices (controller and AENTRs) and still go single back to the Layer 2 Switch from the controller as shown.
Taking the Line 3 outlay as proposed by the Engineer...
As a worst case scenario, if, now or in the future, the controller in Line 3 Wrapper needs to communicate with the second 1769-AENTR and its I/O, this would be a six hop transaction each way.
At the least, Line 3 Wrapper is four hops from the Server, over a greater distance. It may or may not be an issue, but you won't know that until it's too late, if you install it as outlaid.
And yes, the hubs, at a minimum, have to go!
Switches, not hubs or repeaters, should be used at the plant floor because they have the following benefits:
Directed traffic (non-real time, explicit messaging) is seen by only 2 ports
This is true with full-duplex or with half-duplex configuration
Full-duplex eliminates collisions
This reduces the range of worst-case response time, effectively making network or system response time predictable. The effect of both of these is increased system performance.
The Star outlay for Lines 1 and 2 are how you should proceed, at a minimum, in my opinion. If the switch supports inter-VLAN routing, I would make each Line a VLAN from the main switch. You can then communicate between Lines, etc., in a more managed way, if needs be.
But ideally, I would place a Layer 2 Switch at each Line, say at the Bagger, and Star out, making each node on the Line a VLAN off the Line Layer 2 Switch. The three Line Layer 2 Switches would go back to a Main Layer 3 Switch, and then to the Server, and beyond if needs be later.
(btw - thumbs up for the Ignition Server)
Pushing that concept out further, I would also place a Layer 2 Switch at the Main PLC, and make it and the AENTRs VLANs on that Switch, and then back to the Layer 3 Switch.
There is so much involved in getting these networks installed correctly, and every project is different. You nearly need a degree in this stuff alone by todays standards. If the Project Engineer does not fully understand the implications of the choices they are making, other than saving on costs, then someone else that does know needs to politely "take" that decision away from them, if at all possible.
It took a bit of persuading some time back, but we finally got our Project Engineer around to our way of thinking. Now we purchase Stratix 8000 switches, and the like, as and when we need them, no questions asked.
Regards,
George