Can 1 process panel 800 be connected to a primary and 1 to a redundant 800 control network with 800m controllers
I have a network of three AC866 controllers. All three are connected to an Ethernet switch via CN1 and also connected on this switch are 10 process panel 845's communicating with the AC866 controllers. I know that process panel 845's do not have redundant Ethernet connection ability. So, I am interested in knowing if I add another Ethernet switch as a redundant switch and connect all three of the AC866's to this switch via CN2, can I move 5 of the process panel 845's to the redundant switch (adjusting IP addresses accordingly) will they operate correctly? The rationale behind this is that if one Ethernet switch fails, I still have 5 operational process panels in the field to operate machinery. I recently had the one switch fail on my non-redundant network and all process panels obviously stopped communicating and I was unable to control the process. Will my proposed topology work?
Voted best answer
A ring might be the most cost effective method here, as already suggested.
Rings however, are more difficult to fault trace than a flat network.
I recommend using a proprietary ring protocol such as Westermo FRNT, Moxa Turboring, Hirschmann HiperRing, etc over the commonly available spanning tree (STP/RSTP) protocols.
Proprietary redundancy protocols detect & react within a few millieseconds while a Rapid Spanning Tree will need a couple of seconds to detect & stabilize. Regular Spanning Tree may require much more time than so. Proprietary protocols often output diagnostic messages which can be quite handy while fault tracing, some switches, e.g. Westermo have dedicated LEDs indicating ring redundancy status.
An ABB PM8xx controller will try to detect/block loops & storms, but it may shut down if network traffic volume is too high (e.g. due to the "storm" a spanning tree often cause while stabilizing). The controller shutdown is per design.
When trying out the redundancy concept; regularly pull the controller log files and look for the keywords "NSP" (network storm protection) and "loop". The log will tell what was detected and how it was handled, e.g. temporary blocking an Ethernet port.
For high availability systems I recommend wiring up two physically independent networks between the PM866 controllers and use RNRP for redundant routing. As process panels only have singular Ethernet connection, they have in this case to be connected to a router node which in turn has redundant connection to the control networks. Normally RNRP is required for redundant connection, but I believe if the controllers and process panels all set the router node as "default gateway" they will be able to reach each other over both control networks. Controllers + router must run RNRP. The routing role can be held by a computer running RNRP or the newly released ABB NE87x Industrial Routers.
Of course, the third, "process panel network" is still singular point of failure.
I assume this later proposal will be rejected since it require a router node, two dedicated switches per controller location and one dedicated switch per process panel location. Costsome, but with high availability.