On Wednesday 10 September 2003 02:43, Dancer Vesperman wrote: > No worries. Why does it work? > Every packet leaving an interface is up to MTU bytes in length. Let's > say that a byte is ten bits (8+ 1start+1stop). So a 1500byte packet is > 15000 bits. On a 28.8K connection, that means a 1500 byte packet takes > 0.52 seconds to transmit. Assuming that the queueing discipline is FIFO > (first-in-first out) and the queue length on the interface is, say, 10 > packets, a packet can take 10*0.52 seconds to get to the head of the > queue. Plus transmission time, that's 5.2 (for queuing) +0.52 for > transmission. Quite a delay. > > You may be able to do better by shortening the packet queue on the > interface (ip link set ppp0 qlen 4) for example. Or attaching an SFQ > queueing discipline (tc qdisc add dev ppp0 root handle 1: sfq perturb > 10) - although for SFQ, I think you'd want a more substantive queue > length. You can use sfq with a shorter queue : http://www.docum.org/stef.coene/qos/faq/cache/21.html Stef -- stef.coene@xxxxxxxxx "Using Linux as bandwidth manager" http://www.docum.org/ #lartc @ irc.oftc.net _______________________________________________ LARTC mailing list / LARTC@xxxxxxxxxxxxxxx http://mailman.ds9a.nl/mailman/listinfo/lartc HOWTO: http://lartc.org/
