Industrial Ethernet network design

Question

looking for some advice on network architecture for an industrial switchroom with several constraints.

Modbus TCP devices are located in motor control centre cells, these need to connect to the control system located in the same switchroom.

Constraints:

• Control system addressing of downstream device addresses can only be an IP address, no ports or hostnames.

• All downstream device addresses are on network 192.168.1.x (unless DHCP is used - more info below).

• Any equipment needs to be suitable for industrial use and easy to maintain by technicians.

• Up to three separate control system controllers need to access all devices.

Equipment in MCC tiers have dip switch set last octet of address, network is 192.168.1.x. To maintain some sort of order these are set as XXY where XX is the tier number and Y is the cell in that tier. Eg 192.168.1.132 would be the second cell in tier 13.

Now the difficulty comes in that there may be up to 38 tiers, results in insufficient IP addresses if we want to maintain the numbering scheme.

A possible solution to this is to split the MCC into separate networks, eg front 19 tiers and rear 19 tiers. This would result in duplicate IP addresses of course which could be dealt with a few ways.

1. Installing a separate control system network module for each network. This will allow us to access all devices but becomes messy and expensive. (eq, in one switchroom there would be three MCC groups and three control systems, so would require 9 x network interfaces at $4.5k each).

2. A router doing some sort of address translation, where requests for 192.168.10.132 would be translated to 192.168.1.132 on port 1, 192.168.20.132 would be translated to 192.168.1.132 on port 2.

Another completely different solution is to use DHCP on the downstream devices to move away from the 192.168.1.x constraint. This would require an DHCP server suitable for industrial environments, downside is that this is a single point of failure and difficult for technicians to deal with in breakdown situations.

Appreciate any advice on the most elegant method of making this work.

Cheers,

Answer 1

• don't use duplicate IP addresses ever unless you're asking for trouble
• DHCP servers can be set up in redundant/failover mode, that is what you want
• reserve IP addresses by MAC, so they don't change
• use a layer-3 switch for routing in between subnets (low price, high performance)
• depending on budget and requirements, consider a redundant switch layer in between the end switches, connect DHCP servers in a redundant fashion as well
• don't forget that the 172.16.0.0/12 and 10.0.0.0/8 private IP ranges allow for a more generous numbering scheme than 192.168.0.0/16

Answer 2

You could just use some sort of industrial Modbus/TCP gateway to poll all your devices and make the data available on your main network at an IP address of your choice. The addressing scheme of the original Modbus devices would then not matter as that network would be physically isolated from your main network. So long as these devices support multiple clients, you could duplicate the gateway to provide redundancy at all times with only two addresses required.