[Click] Traffic Shaping in Click

[Roman Chertov]

Let me see if I get this straight:

Input 150 Mbps -> Queue -> Output 40 Mbps

There is no way to avoid packet loss here as the queue will eventually 
overflow.  40/150 = 0.2666 meaning that .733333 of traffic will get 
lost.  That is similar to the 80% number that your observed.  Large 
queue can only help you to deal with a burst, but it will eventually 
will overflow during a long enough burst.

Roman

Latency Buster wrote:
> Does irqbalance interfere with Click's packet forwarding performance?
> Using PollDevice() I am still getting extremely high packet loss
> rates. 80% loss at an input rate of 150Mbps with 40Mbps link shaping.
> I am currently invertigating and will throw more info as I gather
> stats.
> 
> Thanks,
> LB
>