[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: A question about priorities

Hi Jamal,

On Tue, Aug 08, 2000 at 11:10:03PM -0400, jamal wrote:
> > It seems that traffic control comes into play only when traffic is to be
> > dequeued to be sent to the output interface. 
> true for the egress queueing. There is also ingress policing which happens
> on the input.

But ingress queueing shapes only the traffic coming from other interfaces,
in my face the traffic is coming from the UDP -> IP -> ...
I want to prioritize the order in which the ip_xmit's are called.

> > Does it also influence how IP layer hands down the  data to the lower layers?
> > 
> I think this would be the case when the NET_XMIT_* codes start to be used
> correctly by the upper layers .... For now the answer is no.

Is there a plan to do so in near future ?

> > For example, if I have two UDP sockets one set at a higher priority using
> > setsockopt than the other. Then though the packets for the higher priority
> > connection are drained first at the interface, if the IP layer hands
> > equal amount of data from the socket buffers to the layers below (which it will
> > if Traffic control does not affect how sockets are drained), the observed
> > throughput for the both the connectons in the long run will be the same,
> > (since IP hands equal amount of data from both the connections.)
> > 
> If you are doing this correctly, then the effect you are describing is
> what you should see. 

This is what amazed me.
 I thought that I should have seen 50-50 traffic for both the connections, since
I am not influencing the incoming rate, but I was completely surprised to
find that I was able to get almost 100% of the bandwidth for the higher priority
connection, when the connections were UDP.
I feel that something really strange is going on.
If you want I can send over the code for you to see.

When I had TCP socket connections, I was seeing about 50-50 bandwidth.