I am in the planning stages of a wireless ISP in the Pacific Islands and it would be great to get some advice.
My idea was for each base station site to run it’s own PPPoE concentrator, with MPLS/BGP used over the backhaul. I would allocate all clients with real world IP’s on a /31 subnet via PPPoE then route these via RFC1918 addresses on the base stations using BGP over the MPLS so all the other base stations on the network know how to reach the /31. In theory (inside my head) this should work.
My thoughts are that this should provide much more efficient use of the backhaul bandwidth as inter-customer traffic will go direct between the relevant base stations rather than via a central PPPoE concentrator.
What are your opinions of this design ? Pros/Cons and if you were to set up a wisp from scratch how would you design it?
My network works almost the same. Each base station has own PPPoE concentrator, all base stations are bridged with ospf. There are 2 separate gateways which serves internet connection and everything works like a charm
That is great to hear. We are waiting to see what the RB616 is, and then we are going to build up a couple of base station setups to do some testing with. We are hoping to use MPLS from the start so I will let you know how we get on.
we have the same type of config, however we allocate a /32 to the customer, all customers point to the same gateway since PPP does not require the gateway to be on the same subnet.
we assign the local address in the ppp profile (becomes the clients gateway when they connect), and the remote address (aka the client’s public IP) is assigned via a ip pool using the radius framed pool attribute to determine which ip pool to use.
additionally each access point radio is bridged to a unique vlan, and we run the PPPoE server on the matching VLAN. this allows you to see which client is connected to which AP by way of the matching vlan, and keeps any broadcast traffic that gets through at a minimum…
I have just ordered about $10k worth of aerials for the “proof of concept” and will be using the /32 method mentioned by Forepoint. The main ADSL provider here does exactly this over their PPPoA connections, I have seen this cause issues with numerous routers as they will not accept a default route outside their interface’s subnet but we know how to work around this so it should be fine.
I am really excited about having a play with MPLS on something other than JunOS.
You mention your use of a VLAN per AP. This confused me a little, are you terminating the PPPoE on the AP closest to the client, or are you bridging the wireless interface they are connected to onto a VLAN that is terminated at a central concentrator ?
Also, are you using MPLS/BGP or OSPF for your backbone routing ?
wlan1 on AP bridged to a VLAN on ether1, goes into a switch along with many other APs each bridged to their own VLAN. then a MikroTik router (usually a x86 box) has an interface on that switch, and VLAN interfaces to match each AP. PPPoE is ran on the VLAN interface on the x86 box plugged into the switch.
We are rather limited space wise at the high sites. I may just run the small repeaters (30 users) in bridge mode and then terminate the PPPoE on the distribution nodes as there is no point on the smaller ones if they only have one uplink.
I am looking forward to all the equipment arriving so I can start having an in-depth play with it.
that works great for smaller sites (I’ve done the same in some areas). just remember to use station-psudobridge on the CPE, and disable forwarding at the AP. Also make sure you aren’t bridgeing that AP and another upstream where the PPPoE is terminated (use VLANs at the upstream location in the same config as I described for the regular AP) as this will prevent users from communicating with each other via layer 2 or layer 3 within in the realm of your bridge.
If I enable PPPoE server on Wlan1,
than in the profile I give the local address 10.1.1.1 (eth1 address)
than in the remote ip pool I give real IPs of 212.1xx.x.x
Will it work ? Real IPs will be routed through local addresses ?
But it’s not working for me.
What I ask is if I can give my PPPoE user a public IP in the same address range as my WAN IP address.
For example:
WAN IP: 62.111.111.1/24
PPPoE user remote address (in LAN): 62.111.111.2
You will need a block of public IP’s that are routed to your Mikrotik router via your upstream provider to be able to do this. If you have a single IP from your upstream provider, or the IP’s are not routed to the external IP of your router you will not be able to do this.
e.g. for this to work you would need:
Stub (aka /30) between you and upstream provider 203.75.30.4/30, you are 203.75.30.5 upstream provider is 203.75.30.6
Block of public IP’s 203.75.34.0/24 routed to 203.75.30.5 by upstream provider
Default Route to upstream providers router (203.75.30.6)
You will now be able to allocate IP’s in the 203.75.34.0/24 range to your PPPoE clients, with a RFC1918 gateway address e.g. 10.1.1.1
