Cisco Support Community
cancel
Showing results for 
Search instead for 
Did you mean: 
Community Member

Multihomed Default Route Question

I was wondering if someone out there could clear up a question/concept for me. Here is the situation we are about to set up:

1. We are an endpoint; hosting applications, sites, servers. We are implementing BGP for fault tolerance and don't plan on performing a lot of traffic engineering or implementing a more advanced BGP config.

2. We will BGP between 2 ISPs

3. Not going to receive full routes only receiving each ISP's Local Routes + Default Transit/Backbone Routes

If I understand correctly, as long as I have a default static (floating?) route, outbound traffic will always be able to exit our network, even if there wasn't a matching route sent by either ISP. Is this correct?

If so, my question is what are the appropriate route statements? I was thinking about doing something like this:

Note in the example lines below:

We are 3.3.3.0/23

ISP #1 Router: 1.1.1.1

ISP #2 Router: 2.2.2.2

ip route 3.3.3.0 255.255.254.0 null0 248

ip route 0.0.0.0 0.0.0.0 1.1.1.1 249

ip route 0.0.0.0 0.0.0.0 2.2.2.2 250

My hope with the above is this:

The first line ensures our locally sourced route will always be announced. The second line will send any traffic through ISP #1 if a route has not been learned/found in BGP. The third line says if ISP #1 is down then send that traffic that doesn’t have a matching route from BGP out ISP #2.

Is this correct? If not, what am I missing or misunderstanding?

Thanks in advance,

Hutch

1 ACCEPTED SOLUTION

Accepted Solutions
Community Member

Re: Multihomed Default Route Question

Hi Hutch,

as to 1) I just meant to say that if you do not specify the AD for the second static route it will assume the default AD for static routes (1) and hence be preferred over the third static route.

as to 2) If you have any doubts about your ISP, keep the static routes. And yes, you will need to manipulate the AD if you want the route to ISP#1 to be preferred over ISP#2. Whether you use an AD of 248 or the default AD does not matter, the second static route will be preferred in either case.

There is one thing I want to mention. You are now basically load-balancing between your two ISP's for all the destinations in the BGP table. Your router will choose an outbound path based on BGP metrics (number of AS-hops). Depending on your traffic pattern (destinations of your users) and the routes advertized by your ISP's, you could end up with filling up the 20MB of ISP#2 while you still have plenty of bandwidth on the path to ISP#1. In that case your users have response time issues with all destinations routed over ISP#2, while destinations routed over ISP#1 have no response time problems.

It could be you will need to make your BGP configuration a little more complicated after all in order to make sure you make the best use of the available bandwidth. You could, for instance, use Local Preference and route maps to prefer ISP#2 for all destinations in ISP#2's AS (local routes), and ISP#1 for all other destinations.

If you feel this is a potential problem, let me know and I can send you a sample config for that.

HTH, Thomas

6 REPLIES
Community Member

Re: Multihomed Default Route Question

sir,

U are trying to make routing decissions at the edge..

Once the packet arrives at one of the router there will be only on route in the routing table of each router....

U should be making routing decissions behind these two routers, or roun iBGP between these two routers and anounce the route on one of the router with higher priority.

Community Member

Re: Multihomed Default Route Question

Hi Hutch,

your config will achieve what you want. I have just two remarks:

- you do not need to set the admin distance of the second line to 249, you can leave it on the default value

- you could entirely remove the default routes (line 2 and 3) and ask your both ISP's to advertize a default in BGP (I am not sure from your information if that is already the case or not). If you receive the default routes via BGP, there would be no need for any static routing, and you could simplify your configuration.

HTH, Thomas

Community Member

Re: Multihomed Default Route Question

Thomas-

Good news. Would you mind clarifying a couple concepts I'm trying to understand?

1. The main reason I was planning on setting specific distances for the static routes was that one of our ISPs is giving us more bandwidth (in my example ISP1 is giving us 100mbit and ISP2 is limiting us to 20) so my thoughts were I might as well send any traffic out ISP1 if no matching route was found in BGP and then fallover to ISP2 if ISP1 is dead for some reason.

2. Also, is there a harm (besides addition extraneous statements into the config) in keeping the routes there? The only reason I ask is I don't particularly trust one of the ISPs technical skillset! My concern is that they wouldn't send a default route and cause problems because we aren't receiving the whole routing table. Is this a valid concern of mine? Regardless, if I want to "prefer" (Note: not talking BGP local preference) ISP#1 over ISP#2 in the event a route isn't found in BGP, would I need the "distance" parameter to make ISP1 look more attractive to outbound traffic? I'm a fan of keeping it simple, so if I could/should remove those lines I certainly will!

Thank you very much for your advice,

Hutch

Community Member

Re: Multihomed Default Route Question

Hi Hutch,

as to 1) I just meant to say that if you do not specify the AD for the second static route it will assume the default AD for static routes (1) and hence be preferred over the third static route.

as to 2) If you have any doubts about your ISP, keep the static routes. And yes, you will need to manipulate the AD if you want the route to ISP#1 to be preferred over ISP#2. Whether you use an AD of 248 or the default AD does not matter, the second static route will be preferred in either case.

There is one thing I want to mention. You are now basically load-balancing between your two ISP's for all the destinations in the BGP table. Your router will choose an outbound path based on BGP metrics (number of AS-hops). Depending on your traffic pattern (destinations of your users) and the routes advertized by your ISP's, you could end up with filling up the 20MB of ISP#2 while you still have plenty of bandwidth on the path to ISP#1. In that case your users have response time issues with all destinations routed over ISP#2, while destinations routed over ISP#1 have no response time problems.

It could be you will need to make your BGP configuration a little more complicated after all in order to make sure you make the best use of the available bandwidth. You could, for instance, use Local Preference and route maps to prefer ISP#2 for all destinations in ISP#2's AS (local routes), and ISP#1 for all other destinations.

If you feel this is a potential problem, let me know and I can send you a sample config for that.

HTH, Thomas

Community Member

Re: Multihomed Default Route Question

That makes sense. Thanks for all your help on this issue.

In response to your comments about unequal bandwidth - yes, that is a concern. I was planning on using BGP Link Bandwidth to handle this. [I actually have another topic started with some specific questions I have relating to the config.]

If that doesn't do the trick I'll definitely need to resort to a more complicated traffic engineering config.

I appreciate the offer for the sample config. I should know in a couple/few weeks if the BGP Link Bandwidth works for me. If it doesn't, I'll probably contact you again if that's ok. I see your email in the Cisco profile *@equant.com - let me know if that's not correct.

Thanks again!

Hutch

Community Member

Re: Multihomed Default Route Question

Hutch,

ok, my email is thomas.cordier@equant.com. Good luck!

259
Views
0
Helpful
6
Replies
CreatePlease to create content