BGP Full Tables Issue

[oliver998]

I have a Cisco 6504 taking 3 full BGP tables. One of our BGP feeds becomes unstable and the BGP session drops. This causes a massive knock on effect for all traffic routing though the Cisco 6504.

How can I keep 3 full BGP feeds and if 1 drops not take 2 - 5 minutes to become back stable?

[Mark Malone]

bgp convergence times are slow by design as its based for internet traffic but there are a few things you can do to speed it up see this link , you need to be careful what you implement though

http://www.networkers-online.com/blog/2008/12/bgp-performance-tunning-convergence-stability-scalability-and-nsf-part-1/

3 full BGP tables on one device , you should try split the load across another device or minimize what you learn in from the ISP using prefix-list and route-maps to ease the load , the memory and cpu on that device are probably struggling with the normal load

[oliver998]

Thanks for the fast reply.

How do big tier-1 ISP run BGP between each other?

It seems that I am very unlucky with the fact my BGP sessions are not stable with an average up-time of no more than 2 months.

[Mark Malone]

What do you mean how do they run it , I suppose that depends on what's required but a standard setup most ISPs would use MPLS as there underlying technology and BGP as the transport

Then end customer would be CE to a PE router and that's how the customers would connect through by being put into individual vrfs to isolate their traffic from other customers and pass through mpls at layer 2 to their far end remote sites connections

Are you sure your routers not maxing out and dropping the connection , 3 BGP tables with full load is a lot to take , how much ram do you have in memory