Hi Peter, If we look holistically at the problem statement, is there a common agreement that local distributed reaction on outbound link congestion of any node in the IGP is a good thing ?
Honestly I don't think so. It looks to me more like uncontrolled traffic shift ping pong instead of proper network operation practice. Moreover, let's observe that core links usually have much higher bandwidth as compared with capacity of external links. So to properly address the congestion I would rather search for higher level protocol adjustments (for example tune BGP best path selection to spread the traffic to different exit points) instead of flip flopping the link occupancy in the core back and forth. And if some core links are becoming congested with elephant flows the proper measure to me is end to end (ingress to egress) TE (does not need to be RSVP-TE to make it crystal clear) instead of local destination based redirects. Cheers, Robert
_______________________________________________ Lsr mailing list -- [email protected] To unsubscribe send an email to [email protected]
