Jerry Anderson (jerry@gi.net) writes:
3. Better TCP windowing and better performance when all
packets follow the same path.
When I wrote that I was thinking of multiple paths, rather than
asymmetric paths. Compare the TCP performance of packet-based
load balancing to session-based load-balancing.
Vadim Antonov (avg@ncube.com) writes:
2) Don't Do Any Dynamic Routing Where Only One Path Exists.
Alan Hannan (alan@gi.net) writes:
Certainly I would not agree with this rule.
Nor would I. It might make sense in a flat topology, but ours is
more hierarchical. I find network management easier when we push
the details of tail-circuit connections out to the access
router. The less complications in hub and core routers, the
better.
Paul Ferguson (pferguso@cisco.com) writes:
'Tis better to black-hole packets than to flap.
In core (EGP) routing, absolutely. However, I haven't noticed
any ill effects of flapping in our internal network.