Re: : 6.x, 4.x ipfw/dummynet pf/altq - network performance issues


Sack was never enabled, the packets in the flood had sack set. rtmaxcache was default, what made you think I had changed it? I was not running SMP, as I explained.

More over suggestions to do ether.ipfw result in terrible performance, etc. A 4.11 bridge and 4.11 router in series move all packets without issue. I'll play with rtexpire and rtmin


garcol@xxxxxxxxxx wrote:
Hi, if you disable sack, what's happend?
(sysctl net.inet.tcp.sack.enable=0) (Are Memory and cpu OK?)


For route problem you can set this to a low value, for example 10 sysctl net.inet.ip.rtexpire: 10
See http://www.freebsd.org/doc/en_US.ISO8859-1/books/handbook/securing-freebsd.html

Why do you modify this?, reset to original 4.x values.
net.inet.ip.rtmaxcache=1024


Try to set this to 0 disable the optimization of the bandwidth delay product for all TCP connections (but it's useful on GE interface).
sysctl net.inet.tcp.inflight.enable=0



If you use a SMP kernel you can try to set this or try to set anyway:

sysctl machdep.hyperthreading_allowed=1
sysctl machdep.cpu_idle_hlt=0

See kern/83406: [smp] em/bge drivers: severe performance loss under SMP


Let's me know
Regards Alessandro
_______________________________________________
freebsd-performance@xxxxxxxxxxx mailing list
http://lists.freebsd.org/mailman/listinfo/freebsd-performance
To unsubscribe, send any mail to "freebsd-performance-unsubscribe@xxxxxxxxxxx"



--
Justin



_______________________________________________
freebsd-performance@xxxxxxxxxxx mailing list
http://lists.freebsd.org/mailman/listinfo/freebsd-performance
To unsubscribe, send any mail to "freebsd-performance-unsubscribe@xxxxxxxxxxx"

Relevant Pages

  • 6.x, 4.x ipfw/dummynet pf/altq - network performance issues
    ... I think you can try to check/tuning this sysctl variables and the isr ... such a DoS/DDoS UDP floods the machine ... Little packets with ip lengths of 28-29 bytes seem to do the ... saturation or PPS limits of the hardware, which in this case was around ...
    (freebsd-performance)
  • Re: 6.x, 4.x ipfw/dummynet pf/altq - network performance issues
    ... Another scary discovery - if you've got 6.2 setup to route, even with static routes, 1Mbps of TCP SYN traffic will cause it to start dropping packets in every direction. ... Could I get an explanation as to what the queue and isr sysctl values are actually doing? ... UDP floods are much better handled - an ipfw block rule for the packet type and the machine responds as if there were no flood at all. ...
    (freebsd-performance)
  • Re: Who wants SACK? (Re: was My planned work on networking stack)
    ... MS>> Selective ACKnowledgment (SACK) allows acknowledgment of received ... MS>> packets in a TCP window so that only the missing/damaged packet needs to ... The difference is that SACK allows the transmitter to transmit the ...
    (freebsd-net)
  • Re: Who wants SACK? (Re: was My planned work on networking stack)
    ... MS>> Selective ACKnowledgment (SACK) allows acknowledgment of received ... MS>> packets in a TCP window so that only the missing/damaged packet needs to ... The difference is that SACK allows the transmitter to transmit the ...
    (freebsd-current)
  • Re: read() returns ETIMEDOUT on steady TCP connection
    ... 2331504 data packets ... 2056 connections established ... 116632 segment rexmits in SACK recovery episodes ... input SACK chunks ...
    (freebsd-net)