I am implementing a WISP network.Usually I would use separate physical ports for internet breakout/wireless backhaul. But I have changed this and now only using VLANs via a managed switch.
Is this a better way? Are there any potential disadvantages of this design?
I would like to hear what sort of architecture real large WISPs are implementing and would love your thoughts on my designs.
Please refer to Fig.1 showing the design and Fig.2 showing the route from the client to the internet via both possible designs.
So, with my current design, traffic flows into the switch, into the router via ether7 and out again via ether7 and back to the switch to reach the internet.
Would there be any performance issues using this topology?
We are serving 100mbps plans so client traffic will reach around 700mbps - 950mbps.

With router on a stick concept every packet passes the switch-router connection twice, once in each direction. In both directions there's mix of ISP-bound packets and clients-bound packets. If uplink/downlink throughputs were symmetrical, this cocept would effectively halve the link capacity. So if your subscription plans are symmetrical (100/100) and all clients were using them fully, the 1Gbps interconnect limits you to serve 5 clients.
In the case where router uses separate ethernet port for upstream connection the limit is at 10 clients because ISP-bound and clients-bound packets use separate physical connections.

I would go for the second topology (with red dotted flows) so use indeed separate interfaces towards ISP and clients.
Is this equipment located close to each other ? Or does "ether7" run across some provider L2-link to some remote location where the EdgeSwitch is located ?

Having designed and built hundreds of WISPs, i've got a few things to share with you that may be helpful.

We typically deploy a "switch-centric" design where all connections go into a switch because it lowers operational overhead. This is similar to your first design drawing. We've scaled this design to tens of thousands of subscribers and a tower count over 1000.

This is an example of that topology at the core




Here is an example at the tower




Then take a look at one of the routing architectures from this article to put on top of that L1/L2 design using VLANs

https://stubarea51.net/2020/03/03/start ... hitecture/

With router on a stick concept every packet passes the switch-router connection twice, once in each direction. In both directions there's mix of ISP-bound packets and clients-bound packets. If uplink/downlink throughputs were symmetrical, this cocept would effectively halve the link capacity. So if your subscription plans are symmetrical (100/100) and all clients were using them fully, the 1Gbps interconnect limits you to serve 5 clients.
In the case where router uses separate ethernet port for upstream connection the limit is at 10 clients because ISP-bound and clients-bound packets use separate physical connections.

Thank you so much for this, it did not occur to me.
Much appreciated

Having designed and built hundreds of WISPs, i've got a few things to share with you that may be helpful.

Thank you very much for this detailed information. Actually I am using your presentation: BGP as an IGP for Carrier/Enterprise Networks - MUM - MikroTik as a template for my network. By the way, how do I access the full configs shown in that paper, it says: Configure (example is PE1 – full configs/presentation will be posted on mum.iparchitechs.com) but when I visit that url, it says that site can't be reached, please advise.

I would go for the second topology (with red dotted flows) so use indeed separate interfaces towards ISP and clients.
Is this equipment located close to each other ? Or does "ether7" run across some provider L2-link to some remote location where the EdgeSwitch is located ?

The equipment is in the same rack. The reason I wanted to use a switch is that I want to move to a "switch-centric" sort of topology