On Sat, 2005-27-08 at 22:38 +0200, Eric Dumazet wrote: > Hi David > > I played and I have very good results with the following patches. > > # tc -s -d qdisc show dev eth0 ; sleep 10 ; tc -s -d qdisc show dev eth0 > qdisc pfifo_fast 0: bands 3 priomap 1 2 2 2 1 2 0 0 1 1 1 1 1 1 1 1 > Sent 440173135511 bytes 3211378293 pkt (dropped 240817, overlimits 0 > requeues 27028035) > backlog 0b 0p requeues 27028035 > qdisc pfifo_fast 0: bands 3 priomap 1 2 2 2 1 2 0 0 1 1 1 1 1 1 1 1 > Sent 440216655667 bytes 3211730812 pkt (dropped 240904, overlimits 0 > requeues 27031668) > backlog 0b 0p requeues 27031668 > > (So about 360 requeues per second, much better than before (12000 / second)) >
I suspect what you are doing is shoving a lot more packets than the wire can handle. Thats why you are getting the backpressure. I read back and it seems you have a 4 way opteron? What kind of bus? Are you filling up that wire? What exactly is this traffic - is it something you are sending from user space? > oprofile results give > 0.6257 % qdisc_restart (instead of 2.6452 %) > > This patch has 3 parts It would be interesting to see the contribution to performance for each of these three changes. For opterons/NUMA the last change seems interesting and could be the most beneficial; although the most important change could be the one you did on tg3 actually. cheers, jamal - To unsubscribe from this list: send the line "unsubscribe netdev" in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html
