1

I have set up a network in a non-standard way, and would like to know if what I did is technically correct, or if I just got lucky.

The setup works fine with my $20 5-port Linksys switch, but not with my $1200 Allied Telesys Fiber SFP/GigE switch :|

We have dedicated Internet access through AT&T via Ethernet, and have been assigned an IP block like: 12.12.12.224/27*

32 bit IP address minus 27 bit mask = 5 bits, so we have 5 bits worth of IPs in our subnet, 2^5 = 32 addresses, making our subnet mask is 255.255.255.224.

This means that our 'network' is 12.12.12.224, and we have 30 IP addresses, from 12.12.12.225 through 12.12.12.254, with 12.12.12.255 being the 'broadcast' address for our network.

We were told that the AT&T gateway is 12.90.119.213, and our gateway should be 12.90.119.214.

Normally to provide access to the Internet you would have a router with the IP address 12.90.119.214 on the ISP side, and 12.12.12.225 on our side, then use 12.12.12.226 - .254 for local devices or other routers.

The router is what allows communication between devices on the 12.12.12.x subnet and the 12.90.119.213 gateway, which is clearly a different subnet. (This is the extent of my networking knowledge; past this, I don't have a clear understanding.)

However, since we're using multiple NAT routers on our side of the connection, each serving a different location with Internet access, I am bypassing the "main" router entirely.

We plug the AT&T Ethernet connection directly into a switch, then the WAN interface of the router for each location plugs into the same switch. Each router uses the AT&T gateway of 12.90.119.213 and a subnet mask of 255.255.255.224 with an IP from our range of 12.12.12.226 - .254 for its WAN interface.

My understanding is that this should still work fine even though the IP of our router is 12.12.12.227 and the IP of AT&T's gateway is 12.90.119.213, and they are on separate subnets, because they are on the same switch, and an ARP request would find that they can communicate with each other directly, without a router...

...and it does work, when the AT&T Ethernet and the Ethernet from each of our routers are all plugged into my cheapo Linksys 5-port desktop switch. But when I try to plug them in through my expensive Allied Telesyn (layer-2) switch, it works "fine" for 4 hours, then our routers lose communication with the Internet and have to be reset. The routers don't freeze or hard-lock, you can still log in to them locally, you just can't reach them from the Internet, or reach the Internet from behind them - not even pings.

Does anyone have an academic answer for why this may be?

Is what I've done completely ridiculous? Or is there just some setting I'm missing. The expensive switch is just a layer-2 switch, so I'm not sure what "settings" there could be on it. Maybe it's too smart, and doesn't like my hack-y configuration.

Should this type of set-up theoretically work?

Interested in your thoughts!

*some IPs have been changed to protect the innocent

2
  • Is DHCP snooping or gratuitous ARP protection turned on in the Allied Telesyn switch, possibly by default? If you can turn up the verbosity on that switches logs, you might see some interesting details? Mar 12, 2018 at 5:36
  • Thanks for your suggestion. I had to decommission this switch a few years ago due to a power supply failure, but I'll see if I can hook it up and check for those settings.
    – waldo22
    Mar 14, 2018 at 2:15

1 Answer 1

1

That setup isn't correct (and I don't know how it's even working). You're supposed to have a router on your side which has a point-to-point link to a router at the ISP. The network should look something like this:

(clients) 12.12.12.226-254/27 <-switch-> 12.12.12.225/27 (your router) 12.90.119.214/30 <-AT&T link-> 12.90.119.213/30 (ISP's router) <---> The Internet

The clients should have an address of 12.12.12.226/27 with a gateway of 12.12.12.225 (which is attached to your router). Your router then takes care of forwarding those packets to the ISP's router, as well as forwarding the responses back to the clients.

2
  • Thanks for your reply (sorry I just noticed). I did specify that I understand the way it is normally done. Do you think this works because the connection from AT&T is via Ethernet, and it's able to use ARP to find the gateway on the same switch?
    – waldo22
    May 7, 2012 at 21:18
  • That may explain how the packets get out (although your machines must also have a route for 12.90.119.212/30 on the local network for them to try to route directly to .213), but I don't know how they get back in, because the AT&T side router is going to try route via the router on your side at .214, which doesn't exist.
    – mgorven
    May 8, 2012 at 23:19

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .