Remote tag provider performance

I have several IO servers + FrontEnd server
Each IO server connects to 2 devices
Each server has a Tag Provider
tagProviderIO1, tagProviderIO2, tagProviderIO3, tagProviderIO4, tagProviderIO5

1 of the 10 devices is just reference data and not operational data like the other 9 devices. I need the tags on all the IO servers to be able to reference that data that is currently connected to IO1

For Example: I need to be able to put in a shared UDT something like
[tagProviderIO1]reference data/folder/{parameter}/Name

which as is wont work because that tag provider is not available on all servers.

Gateway connections are only established between IO servers and the FrontEnd server

Option 1. Is to add IO1 as gateway netwrok connection to all other IO servers and add IO1 server as remote tag provider on all IO servers. (direct remote tag provider + new gateway network connections)

Option 2. Is to add IO1 as remote tag provider on all the IO servers pointed to the remote tag provider on the FrontEnd server (Remote tag provider of a remote tag provider)

Option 3. Make a new remote tag provider on the FrontEnd Server for all the common reference tags and put remote tag provider on each IO server (direct remote tag provider with no change to gateway network)

Thoughts on these different options? I am leaning towards option 3 as it doesn't require the additional gateway network connections and I am concerned with performance issues of using a remote tag provider of a remote tag provider. It will put the device load on the FrontEnd but it is connected to same physical network as PLCs and our front end load is really light so I don't think we should have performance issues there.

What kind of device is the reference? Could it be a static simulation in all of the servers?

Consider option 1+2. Make a remote tag provider in the other IO servers pointing at the gateway containing IO1, but let the traffic flow through your gateway network as is. (Unless all the IO gateways are in a single facility, separate from the front end, in which case I'd do option 1.)

all the devices are AB PLCs .

Static simulation?

Everything IO and Front end and PLCs are in the same facility connected to same physical network.

My EtherNet/IP module's Host device type looks like a Logix PLC, both to OPC connections on the gateway, and to external EtherNet/IP connections. It implements the PLC side of Rockwell's Logix Data Access manual, in addition to all of its other connectivity.

I'd probably put the reference information in the front end, then.