Guy had shown some traffic graphs of the ENSS illustrated where the BARRNET transition occurred (in late January). The CIX interconnection showed peek utilization from October until the California IETF.
BARRNET is a net provider of information and so their traffic is asymmetric than before. Since the MCI transition, the CIX link utilization has gotten much more reasonable. The current situation had alot of delay but not congestion. Hopefully, once more interconnections are available on the West Coast the CIX problem should go away.
BARRNET intends to keep the CIX T1 after the MCI connection is up. Long term, it will depend on what happens to the CIX organizationally and the possibility that BARRNET may get a high speed connection to MAE-WEST itself.
Vince illustrated that under constant congestion, a specific TCP will try alot of retransmissions. Once BARRNET had statically routed that service via MCI and the transmissions were significantly reduced.
Vince also showed a problem that the current 10.2 test code from Cisco has inaccurately counting bps on the DS3s. The problem only seems to occur on serial (point-to-point) interfaces.
