Queue full on Intel 350 and difference between obytes and ibytes

[Alberto Vargas]

Guys,

I have read in another messages that queue full happens when the hardware limit is hit.

I'm generating stateless IMIX traffic with multiplier flag with 100% value. In this scenario I'm getting a bunch of queue full. The CPU utilisation is 69%. When I change multiplier flag to 99.9% there is no queue full and CPU utilisation is only 13.7%.

Is this .1% enough to avoid hitting hardware limitation? What limitation is it? This is conceptual doubt.

Why Am I getting in both scenario differences beetwen opackets, ipackets, and so? Should the number be the same with no loss?

TRex version is 2.88 and I'm running it with python CLI client.

The following is a screenshot that shows different values for RX and TX:

[hanoh haim]

Hi, 

Yes, 0.1% can be the difference between queue full and not queue full . 

Thanks

Hanoh