Date: Thu, 25 Apr 2013 13:04:14 -0700 From: Adrian Chadd <adrian@freebsd.org> To: Erich Weiler <weiler@soe.ucsc.edu> Cc: freebsd-net@freebsd.org Subject: Re: pf performance? Message-ID: <CAJ-Vmo=vWB6Y3uCAkLD_6UfgCWYxKx1yA_y4JMDjiZXV8X=hYQ@mail.gmail.com> In-Reply-To: <51797E3F.1030400@soe.ucsc.edu> References: <5176E5C1.9090601@soe.ucsc.edu> <201304240134.22740.vegeta@tuxpowered.net> <517974DA.5090809@soe.ucsc.edu> <CAJ-Vmom9AcEGKYHNDBkJ_yUo4%2BbMrbxfbLBV6HrUD2UW_0_crw@mail.gmail.com> <51797E3F.1030400@soe.ucsc.edu>
next in thread | previous in thread | raw e-mail | index | archive | help
If it contends on the global pf lock, you're short of luck. There may be some hack to enable in sysctl that defers part of the packet processing into a taskqueue, but I dont' know if that's for general IP processing or just socket iO processing. One of the network stack peeps will know. ADrian On 25 April 2013 12:04, Erich Weiler <weiler@soe.ucsc.edu> wrote: >> ... please ask the pfsense guys to either migrate to -9, or backport >> the -head pf (with the locking fixes!) to -8 for that. >> >> Otherwise you're very likely going to be wasting time on something you >> can't really push that much harder. > > > I can ask for that (and will soon, likely), but to play with my current > setup in the meantime, can we logically say that if I have 4 cores, and one > interrupt queue is assigned to each core, and under I load I see each core > (via "top -P") at 100% in interrupt usage, would it be safe to say that more > cores (with additional interrupt queues accordingly) would mean more > interrupts overall being processed, which would mean more pps?
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?CAJ-Vmo=vWB6Y3uCAkLD_6UfgCWYxKx1yA_y4JMDjiZXV8X=hYQ>