cancel
Showing results for 
Search instead for 
Did you mean: 
cancel
775
Views
5
Helpful
3
Replies

Outbound load balancing

Paul Harris
Level 1
Level 1

We have two locations with connections to two different ISPs which provides redundancy/failover should either connection go down. This works fine but I would also like to implement some kind of load balancing to shift some of the load from ISP A to ISP B as the split is currently something like ISP A-95% and ISP B-5% and we quite frequently go above the commited rate on our ISP A connection.

The most simple method of doing this that I have found by googling is to use the BGP local preference attribute to force the routers to prefer ISP A for half of possible IP ranges (1.0.0.0 - 128.0.0.0) and to prefer ISP B for the rest.

I was wondering if anyone had tried this and how well it worked?


My concern is that it is an arbitrary way of determining path selection and as a result some customers may see performance degradation by having their traffic routed over a sub-optimal path.

Paul

1 Accepted Solution

Accepted Solutions

Varun Uniyal
Level 1
Level 1

How are the devices on the inside learning the BGP routes? iBGP or some

IGP such as OSPF/EIGRP in which scenario you must be redistributing BGP into IGP?

You can always set local preference conditionally as follows :

neighbor x.x.x.x route-map TEST in

route-map TEST permit 10

match ip address

set local-preference

route-map TEST permit 20

match ip address

set local-preference

If you can run OSPF on the inside, then you can consider redistributing BGP into OSPF with E1 route-type , though redistributing BGP into IGP is not really a good idea.

If you have no IGP on the inside try running GLBP between the two edge routers...i am not sure of your topology so cannot say how feasible this will be.

View solution in original post

3 Replies 3

paolo bevilacqua
Hall of Fame
Hall of Fame

My concern is that it is an arbitrary way of determining path selection and as a result some customers may see performance degradation by having their traffic routed over a sub-optimal path.

With the fast Internet of today, that is no more a concern anyway. Unless your access circuits are undersized, or your ISP backbone is not up to standards, you have nothing to worry about, no matter what is routed where.

Varun Uniyal
Level 1
Level 1

How are the devices on the inside learning the BGP routes? iBGP or some

IGP such as OSPF/EIGRP in which scenario you must be redistributing BGP into IGP?

You can always set local preference conditionally as follows :

neighbor x.x.x.x route-map TEST in

route-map TEST permit 10

match ip address

set local-preference

route-map TEST permit 20

match ip address

set local-preference

If you can run OSPF on the inside, then you can consider redistributing BGP into OSPF with E1 route-type , though redistributing BGP into IGP is not really a good idea.

If you have no IGP on the inside try running GLBP between the two edge routers...i am not sure of your topology so cannot say how feasible this will be.

Yes the internal devices are using iBGP.

I will have a go at setting the local route preference and see if it makes a difference.

Cheers

Paul

EDIT:

Actually if someone could just check the config I am intending to use to make sure it makes sense (I'm fairly new to BGP)...

Router A going to ISP A

router bgp 12345

neighbor x.x.x.x remote-as 56789

route-map test in

route-map test permit 10

match ip address 10

set local-preference 500

acess-list 10 permit 128.0.0.0 127.255.255.255

Router B going to ISP B

router bgp 12345

neighbor  x.x.x.x remote-as 54321

route-map test in

route-map test permit 10

math ip adress 10

set local-preference 500

access-list 10 permit 0.0.0.0 127.255.255.255

Then presumably I'd need to restart the BGP session on both routers?

Review Cisco Networking for a $25 gift card