Date: Thu, 25 May 2006 23:21:52 +0100 From: "Joao Barros" <joao.barros@gmail.com> To: "Mike Jakubik" <mikej@rogers.com> Cc: freebsd-performance@freebsd.org, Nash Nipples <trashy_bumper@yahoo.com> Subject: Re: Poor Samba throughput on 6.0 Message-ID: <70e8236f0605251521l1435d7besef7ccff83d7434c5@mail.gmail.com> In-Reply-To: <4475F75C.4090908@rogers.com> References: <20060525174211.46064.qmail@web36315.mail.mud.yahoo.com> <4475F75C.4090908@rogers.com>
next in thread | previous in thread | raw e-mail | index | archive | help
On 5/25/06, Mike Jakubik <mikej@rogers.com> wrote: > Nash Nipples wrote: > > Hi Guys, > > > > has anyone actually managed to speed up the thing up to 10-12 MB/s > > > > i have a good 7-9 MB/s on large files and that should be enough, but st= ill, out of curiosity? > > > > > > No, not really. The performance of samba on freebsd still sucks. I have > a gigabit link between my samba server and windows xp workstation, both > use the intel em card. The fastest i can download from the freebsd > server is 15.946 MB/s, fastest upload to is 13.594 MB/s. Thats pretty > poor for a gigabit link. However, 9MB/s on your link is not bad at all, > one has to consider the protocol overhead. > > > Oh by the way: > > 100 MB ethernet > > ping -s 65507 -f windowshost ~ 10-12 MB/s > > ftp open freebsdhost put/get 500MB.file ~ 10-12 MB/s > > net.inet.tcp.inflight.enable=3D1 who said u have to put it down? > > > > This feature limited the speed of LANs, however recent code has aimed to > fix that, but i don't think its in 6.0. I used a recent -CURRENT for the > tests, you should consider updating to -STABLE. That's supposed to be in 6.1, but still the speed still is sub par : On 2/16/06, Andre Oppermann <andre@freebsd.org> wrote: > andre 2006-02-16 19:38:07 UTC > > FreeBSD src repository > > Modified files: > sys/netinet tcp_input.c tcp_subr.c tcp_timer.h > tcp_var.h > Log: > Have TCP Inflight disable itself if the RTT is below a certain > threshold. Inflight doesn't make sense on a LAN as it has > trouble figuring out the maximal bandwidth because of the coarse > tick granularity. > > The sysctl net.inet.tcp.inflight.rttthresh specifies the threshold > in milliseconds below which inflight will disengage. It defaults > to 10ms. > > Tested by: Joao Barros <joao.barros-at-gmail.com>, > Rich Murphey <rich-at-whiteoaklabs.com> > Sponsored by: TCP/IP Optimization Fundraise 2005 > > Revision Changes Path > 1.288 +10 -0 src/sys/netinet/tcp_input.c > 1.237 +7 -1 src/sys/netinet/tcp_subr.c > 1.28 +3 -0 src/sys/netinet/tcp_timer.h > 1.127 +1 -0 src/sys/netinet/tcp_var.h > _______________________________________________ --=20 Joao Barros
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?70e8236f0605251521l1435d7besef7ccff83d7434c5>