cancel
Showing results for 
Search instead for 
Did you mean: 
cancel
514
Views
0
Helpful
1
Replies

Load Balancing with ASR9000 vN and multiple ISPs

Jordi Benet
Level 1
Level 1

Hi,

we will deploy a new DC as Active/Active.

We will have ISPA and ISP B in each DC. Internet users are anybody in the internet coming to our e-commerce DC application.

How could we do load balancing between ISPs using the ASR9001 and nV feature ?

There is any IOS-XR feature that could help us about to do load balancing between ISPs?

Thanks a lot.

Regards,

J

1 Accepted Solution

Accepted Solutions

xthuijs
Cisco Employee
Cisco Employee

You have 2 options here Jordi, either you can use BGP loadbalancing, this requires multipath as BGP by default would only install one route from the BGP table to the RIB hence FIB.

But this may result in excessive IRL (inter rack link) usage in the cluster when traffic coming in on rack0 wants to take the bGP path out on rack1

You could also use ABF (access-list based forwarding) to forcelly push traffic received on rack0 out on the link on rack0 and use an ipsla tracker to fallback to rack1 in case the uplink is gone.

Alternatively to extend this by IGP signaling to redirect traffic preferably to rack1 to start with to minimize the IRL usage.

And then you also have the ability to use RPL in the uplink path to make one link more preferred on teh internet then the other in case you want to control a bit which link is preferably used on rack0 or rack1

regards

xander
 

View solution in original post

1 Reply 1

xthuijs
Cisco Employee
Cisco Employee

You have 2 options here Jordi, either you can use BGP loadbalancing, this requires multipath as BGP by default would only install one route from the BGP table to the RIB hence FIB.

But this may result in excessive IRL (inter rack link) usage in the cluster when traffic coming in on rack0 wants to take the bGP path out on rack1

You could also use ABF (access-list based forwarding) to forcelly push traffic received on rack0 out on the link on rack0 and use an ipsla tracker to fallback to rack1 in case the uplink is gone.

Alternatively to extend this by IGP signaling to redirect traffic preferably to rack1 to start with to minimize the IRL usage.

And then you also have the ability to use RPL in the uplink path to make one link more preferred on teh internet then the other in case you want to control a bit which link is preferably used on rack0 or rack1

regards

xander