ip sla with BGP

Unanswered Question

I'm running BGP with my own ASN, and peering to 2 providers, one 10Mbit, and the other 3Mbit. I prepend my own ASN as-path many times for the 3Mbit provider, so remote networks will almost always prefer the 10Mbit link. Ok, so now to the challenge.


This morning, the 10Mbit connection started acting really slow. Lots of latency, dropping 70% of the packets, etc. BGP didn't save my network from being severely impacted. After the provider of the 10Mbit link drug their feet on the issue, I shut down the interface, and let BGP flow over the the backup link.


Could I use ip sla, or something similar to automate this process in the future? How could I protect the network against something like this happening in the future? Any advice valued & rated! Thanks.

  • 1
  • 2
  • 3
  • 4
  • 5
Overall Rating: 5 (2 ratings)
Loading.
arun kumar Mon, 01/28/2008 - 21:43
User Badges:
  • Bronze, 100 points or more

Hi,


You can try Optimized Edge Routing now called Performance Routing which will automatically divert the traffic to the optimized path based on some pre-defined attributes.


Chk the URL's below:


http://www.cisco.com/en/US/products/ps6628/products_ios_protocol_option_home.html


http://www.cisco.com/en/US/prod/collateral/iosswrel/ps6537/ps6554/ps6599/ps8787/prod_presentation0900aecd806c5089.pdf



Hope this helps..



Rick Morris Fri, 02/01/2008 - 09:48
User Badges:
  • Silver, 250 points or more

I understand why you are prepending, but would it be plausible to use both links at the same time and let the BGP policy use the best path for route selection?


Do you get routes from your providers?


I would recommend getting full routes from your provider, removing the AS prepending and seeing if that speeds things up. Then any issue with latency will be handled becuase your traffic will not see a bunch of prepends and deny that traffic. Some ISP's will actually drop your announcement if you prepend more than 4 times. You can still prefer to use the 10M link for outbound if you like by using route-maps and if you are flapping you could try adjusting your keepalive timers to drop the session and force traffic out the other peer.

Actions

This Discussion