It seems that when I see a description of a fully-meshed peering system, the exponential growth of peer connections is often referred to as a negative.

However, some mentions of this were specifically towards fully meshed peering in an AWS VPC specifically.

Surely, if connections between peers is done entirely automatically, and the amount of connections for any individual VPC increases linearly, then this shouldn't be an issue?

  • Do you have a specific problem? – Zac67 Sep 25 '19 at 18:23
  • @Zac67 I am considering using a fully meshed set of VPCs for a specific project I'm working on, and I'm not sure why having a large amount of connections is bad per se, when everything can be scaled automatically. (Sorry if I'm using incorrect terminology, I'm not a networking guy). – RecyclingBen Sep 25 '19 at 18:34

By exponentially increasing the the number of connections, you also exponentially increase the amount of control traffic, and you could end up with the control traffic using most or all of the available bandwidth of the system.

Some protocols will have mitigations for this behavior. For example, OSPF uses the DR/BDR to break the exponential number of connections, and iBGP can use route reflectors or confederations to do something similar.


In addition to Ron’s point about the amount of control traffic: The additional cpu load resulting from processing route updates on every router often isn’t worth it either when your network grows. You don’t want every router in your network to do that, that’s where for example route reflectors come in. Having many routers processing route updates can lead to nasty routing inconsistencies if not all routers process updates equally fast.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.