06-06-2019 10:47 AM
Hello
BGP synchronization provides black hole prevention in IBGP environment.
but
can anyone explain why " no synchronization" is default in Cisco routers ?.
Thanks in advance.
Solved! Go to Solution.
06-06-2019 11:03 AM - edited 06-06-2019 11:22 AM
Hello Sivam,
the default has been changed to no sync for the following reasons:
many years ago BGP was used on Border routers and it was redistributed into IGP.
The BGP sync feature was thought for this scenario: do not advertise a prefix to an eBGP peer if you don't see it in the IGP, to be sure that internal routers are not a black hole for traffic.
This was for the low number of prefixes involved and because BGP route reflectors and BGP confederations were to be introduced.
In modern service provider networks iBGP and MP iBGP are extensively used to advertise customer routes within services even for simple Internet connectivity (in several cases using a dedicated L3 VPN service).
The IGP is used to provide reachability between loopback interfaces that are used for iBGP sessions, specially in MPLS networks this is a key.
Actually if MPLS is deployed internal P nodes do not perform IP routing but simply MPLS label switching without the need to inspect the IP header (it is done but only for load balancing purposes).
In all these scenarios BGP is not redistributed into IGP anymore and so the BGP sync needs to be disabled.
In simple words the BGP synchronization can be considered an historical feature, that in most cases is not needed as we had to put the following two commands all the time we started a new router bgp process:
no sync
no auto-summary
Cisco received requests to change default behaviour for these two commands and so now they are disabled by default.
Now a full Internet table is around 660,000 or more prefixes and no IGP (even IS-IS) can scale to this so redistribution into IGP is not a viable option in all cases where a full internet table is received.
Hope to help
Giuseppe
06-06-2019 11:03 AM - edited 06-06-2019 11:22 AM
Hello Sivam,
the default has been changed to no sync for the following reasons:
many years ago BGP was used on Border routers and it was redistributed into IGP.
The BGP sync feature was thought for this scenario: do not advertise a prefix to an eBGP peer if you don't see it in the IGP, to be sure that internal routers are not a black hole for traffic.
This was for the low number of prefixes involved and because BGP route reflectors and BGP confederations were to be introduced.
In modern service provider networks iBGP and MP iBGP are extensively used to advertise customer routes within services even for simple Internet connectivity (in several cases using a dedicated L3 VPN service).
The IGP is used to provide reachability between loopback interfaces that are used for iBGP sessions, specially in MPLS networks this is a key.
Actually if MPLS is deployed internal P nodes do not perform IP routing but simply MPLS label switching without the need to inspect the IP header (it is done but only for load balancing purposes).
In all these scenarios BGP is not redistributed into IGP anymore and so the BGP sync needs to be disabled.
In simple words the BGP synchronization can be considered an historical feature, that in most cases is not needed as we had to put the following two commands all the time we started a new router bgp process:
no sync
no auto-summary
Cisco received requests to change default behaviour for these two commands and so now they are disabled by default.
Now a full Internet table is around 660,000 or more prefixes and no IGP (even IS-IS) can scale to this so redistribution into IGP is not a viable option in all cases where a full internet table is received.
Hope to help
Giuseppe
06-07-2019 08:02 AM
Makes sense.
Administrator can manipulate the protocol operations, But I think network devices default settings should be on the safer side, isn't it!
Thanks for the reply.
06-07-2019 08:26 AM
Hello Sivam,
it is also a question of implementation choices.
Just to make an example an ASR 9000 router running IOS XR by default does not advertise any prefix to an eBGP neighbor in address family ipv4 unicast you need to apply a routing policy that can simply states advertise everything or advertise some routes.
Hope to help
Giuseppe
06-07-2019 10:23 PM
Oh is it?
Discover and save your favorite ideas. Come back to expert answers, step-by-step guides, recent topics, and more.
New here? Get started with these tips. How to use Community New member guide