BGP on a Cisco 7200

[Kookieman]

I don't know if I came to the wrong place. But every other forum out there just kinda sucks. 

 

So right now, I'm running a WISP network with multiple P2P Bridges feeding towers for redundancy. I am using STP on tough switches (going to be switching them to Netonix) for the redundancy portion of this. 

To reduce overhead, Im thinking about putting my bridged links in router mode, and routing the traffic through, forcing packets to use a link based on the route in the routers, but then that would require me to reroute data if a link went down. 

 

Could I use BGP to handle the weight of an IP and where it goes, then if it fails have it pick the other link? 

 

Or should I just stick with STP and bridge mode and call it a day? 

 

Side note, I would just use Vlans on the tough switches along with cost and priority to achieve less overhead, but Tough switches kind of suck when it comes to CPU and memory. When I tried that, my throughput was cut in 3/4ths ... like 30 Mb on a 100 Mb link. 

[Lurick]

Yah, you can set weight or a ton of other attributes to locally influence which routes are most preferred and even suggest to the other router which path you would like them to take to you.

[Kookieman]

4 hours ago, Lurick said:

Yah, you can set weight or a ton of other attributes to locally influence which routes are most preferred and even suggest to the other router which path you would like them to take to you.

Okay, the only reason I ask is because I've never dealt with BGP before. Just know of it. But I hear its way faster than STP and even RSTP. Do I need to have a device on the far end that can do BGP to achieve this? Or can I have a Cisco on the close end, and just use the UBNT radios routing capabilities after that? 

I would assume if I routed the IPs on all my links to the far end, my Cisco would send packets where they need to go, but coming back concerns me. I dont think UBNT has an option for BGP (at least not that Ive found).

I might just be screwed coming back because the UBNT radio would just send out its GW.

(Welcome to confused)

[Lurick]

2 hours ago, Kookieman said:

Okay, the only reason I ask is because I've never dealt with BGP before. Just know of it. But I hear its way faster than STP and even RSTP. Do I need to have a device on the far end that can do BGP to achieve this? Or can I have a Cisco on the close end, and just use the UBNT radios routing capabilities after that? 

I would assume if I routed the IPs on all my links to the far end, my Cisco would send packets where they need to go, but coming back concerns me. I dont think UBNT has an option for BGP (at least not that Ive found).

I might just be screwed coming back because the UBNT radio would just send out its GW.

(Welcome to confused)

Unless both sides support BGP, or any other routing protocol for that matter, then you can't do it. BGP negotiates with the other side and once everything is configured properly then it comes up. Now you could setup static routing and just give the least preferred path a higher Administrator Distance so it's less preferred.

[WINcon_Gaming]

What does your network diagram look like? 

[Teshy09]

depends on the size on the network, by the sounds of it this could be achieved with other routing protocols, 

But a network diagram will give us somewhere to start from

[leadeater]

@Wombo Advice?

 

You don't have to use BGP there are other routing protocols that you could use, OSPF/IS-IS etc, but picking the correct one leads to many detailed network design questions.

 

If your going for a routed network design then yes each end of each link will need a routing capable device. Also be careful since we are talking about wireless links which from a networking standpoint are unreliable and you could end up with many link down/route convergence events that are unnecessary which is a bad thing.

 

I would actually suggest talking this to the Ubnt forums, keep us in the loop as I'd like to hear how this goes.

[Wombo]

On 2016-09-10 at 8:11 AM, Kookieman said:

I don't know if I came to the wrong place. But every other forum out there just kinda sucks. 

 

So right now, I'm running a WISP network with multiple P2P Bridges feeding towers for redundancy. I am using STP on tough switches (going to be switching them to Netonix) for the redundancy portion of this. 

To reduce overhead, Im thinking about putting my bridged links in router mode, and routing the traffic through, forcing packets to use a link based on the route in the routers, but then that would require me to reroute data if a link went down. 

 

Could I use BGP to handle the weight of an IP and where it goes, then if it fails have it pick the other link? 

 

Or should I just stick with STP and bridge mode and call it a day? 

 

Side note, I would just use Vlans on the tough switches along with cost and priority to achieve less overhead, but Tough switches kind of suck when it comes to CPU and memory. When I tried that, my throughput was cut in 3/4ths ... like 30 Mb on a 100 Mb link. 

 

14 hours ago, leadeater said:

@Wombo Advice?

 

You don't have to use BGP there are other routing protocols that you could use, OSPF/IS-IS etc, but picking the correct one leads to many detailed network design questions.

 

If your going for a routed network design then yes each end of each link will need a routing capable device. Also be careful since we are talking about wireless links which from a networking standpoint are unreliable and you could end up with many link down/route convergence events that are unnecessary which is a bad thing.

 

I would actually suggest talking this to the Ubnt forums, keep us in the loop as I'd like to hear how this goes.

I would not recommend BGP for this. BGP is not meant to be an IGP and really shouldn't be looked at as a protocol that supplies any kind of reliable failover. BGP is for route replication across the entire internet, by design it is VERY slow because of this.

 

If your network is segmented into spans no greater than 8 devices in a chain you should be good to stay with layer 2 protocols such as STP. If you branch out to anything larger I would recommend going to a routed layer 3 design with a good IGP such as OSPF, IS-IS or even EIGRP if you find enough platforms that support it.

 

If complexity is low you should stay with Layer2. As complexity increases, move to a layer 3 design that bridges the layer 2 segments together and handles the routing/redundancy. As spans increase, implement optics (layer 1). As you start to get into a larger network designs all 3 elements will have to be considered. The most typical modern service provider designs will involve layer 2 for transport and layer 3 for routing/redundancy with optics spanning the long haul.

 

As an example, modern service provider MPLS networks involve layer 2 transport, typically Ethernet, with routing choices for these layer 2 frames performed via higher levels protocols such as OSPF at layer 3 for finding the best path between nodes. One way I like to think about it is every time you step closer in the network view you go deeper by 1 layer. At a high-level you have the layer 3 routed design showing the paths between nodes and peers (Ex. router-router). As you go in deeper you see the layer 2 paths between the network elements, typically the MPLS paths for forwarding traffic between layer 3 devices. Going deeper yet you find the layer 1 paths between your devices, these would be the physical optical links, or in this case, wireless shots.

 

BGP, as its name implies, should really only be used at the network border for route advertisement and replication to the greater internet. BGP is not an IGP.

[Kookieman]

On 9/10/2016 at 10:53 PM, Teshy09 said:

depends on the size on the network, by the sounds of it this could be achieved with other routing protocols, 

But a network diagram will give us somewhere to start from

I will post a diagram when I get the chance. Thank everyone so much for helping.

[InVis]

why not run it as a service in MPLS and use IS-IS?