On Mon, Jan 23, 2006 at 03:12:22PM +0200, David Randelman wrote:

> From my experience the main problem is believe it or not your download 
> speed as well, the ISP creates huge buffers of data being sent to you. 
> If you want low latency you will have to disable the ISP downlink buffer 
> or at least reduce it, normally from my experience a 1.5Mbit line needs 
> to be reduced by half at least. Once you have done this you will have 
> much lower latency.

How do you lower the download bandwidth?

I have used wondershaper for two years now, but it's mechanism for
lowering download is to "drop the packets on the floor" AFTER they
have been received so I turned that part off.

This just causes double transmission of the packets. The effective bandwidth
of the line drops, but the actual queues still stay large. The only way I can
see to accomplish real queue reduction would be to selectively increase
the latency of the line for low priority traffic. This could be 
accomplished by holding back TCP ack packets so that the originating host
really does not send faster than we want. 

Dropping the packets on the floor does lower the effective bandwidth,
but it really does not lower the line utilization or queue length. The
end to end throughput decreases but the bottleneck is after the line
not before it.

This really is very impolite, it causes the sender to resend packets they
had no reason to resend. It causes a lot of uncessary retransmission just
to keep you stats looking good, without actually doing any good.

Geoff.

-- 
Geoffrey S. Mendelson, Jerusalem, Israel [EMAIL PROTECTED]  N3OWJ/4X1GM
IL Voice: (07)-7424-1667  IL Fax: 972-2-648-1443 U.S. Voice: 1-215-821-1838 
The trouble with being a futurist is that when people get around to believing
you, it's too late. We lost. Google 2,000,000:Hams 0. 

=================================================================
To unsubscribe, send mail to [EMAIL PROTECTED] with
the word "unsubscribe" in the message body, e.g., run the command
echo unsubscribe | mail [EMAIL PROTECTED]

Reply via email to