[c-nsp] Throughput Issues with Trunk Between Switches

[Erik Fritzler]

We have tried this across Cisco 3550's, 3560's, 3750's and 6509's.
The controllers are not showing any kind of errors during the testing. That is what is confusing me.
When testing with small packet sizes and approaching 98 -99% capacity on a 1 Gig link, that is when the errors appear. 
We are talking about 64, 128, and 256 bit packet sizes. There are errors even at the larger packet sizes, but not as many due to the number of packets that can passed over the 1G link.

>    Smart Bits----->Access port--->Cisco Switch-->Trunk-->Cisco Switch-->Access Port--->Smart Bits
>    If we leave the connection between the 2 switches as an access port link then we can test across at full 1.0Gbps.
>    However, it the link between the switches is configured as a trunk we experience packet loss.
>    The smaller the packets, the higher the throughput, the higher the errors.
>    We have tried both ISL and dot1q trunks, as well as setting the nonegotiate setting on all the ports.
>    Has anyone else seen anything like this?

> I haven't seen this - but I haven't tested for this sort of stress either. Which switches are you running?

> Where are you dropping the packets? you may want to check show controllers/ show interface controllers.

> You have changed the buffer allocation by changing between access and trunk ports.


> Cheers

> Andrew