Low POS throughput on ML100T cards

[cerp]

We have a new OC48 Ring utilizing 3 15454 Nodes and have ML100T cards in each node connected in a "ring" via STS3 circuits to the POS interfaces. On each of the ML100T cards I have several subnets on the fastethernet ports and am routing via eigrp. Latency is very good but my throughput seems very low. Doing file transfers my max troughput is around 90KB/s, which seems low for a 155mb link. Any ideas...would RPR be a better choice instead of using the POS links as point-to-point links? Thanks

[bkurtzer]

Are you seeing CRC and/or input errors on any of the ML cards? There is a bug that may address your issue: CSCsb43596 - CRC errors for 5% traffic out of ML100T ports:

Symptom: ML100T cards send out approximately 5% of frames (frame size 1518) with a bad CRC.

Conditions: This can occur in the following scenario.

1. Autonegotiation is disabled. Speed set to 100/ Duplex to Full.

2. ML startup configuration is saved on TCC and the ports are enabled.

3. Card is rebooted.

Certain ports ( at random) will display CRC errors coming out of them. The failing ports is not always consistent: sometimes after a reboot, different ports will show the CRC errors.

Workaround: Disconnect and reconnect the cable, and the errors go away. Errors also go away if the port is shut/no shutdown.

***************************

You can also try resetting the ML cards in each of the nodes to see if the throughput improves.

[Marvin Rhoads]

We use these cards as well. Your throughput should be near line rate. Something is likely amiss on your interface(s) (or the bug mentioned above).