Date: Thu, 22 Nov 2012 09:09:06 -0800 From: Adrian Chadd <adrian@freebsd.org> To: Marc Peters <marc@mpeters.org> Cc: freebsd-net@freebsd.org Subject: Re: Low Bandwidth on intercontinental connections Message-ID: <CAJ-VmokpTkXg5_O-Ut6k6EUqEPD26vspYjJtP8=P9-EKD6TiBw@mail.gmail.com> In-Reply-To: <50AE47D4.7080608@mpeters.org> References: <50ACF62C.8000408@mpeters.org> <CAOgwaMuUuJ2%2BmKqsFVp=DyVFkfm8Et%2Brnt2iEGDO8i1Kt_kDVA@mail.gmail.com> <50ad087d.1892cc0a.2cce.3bf2@mx.google.com> <50AD1012.7020209@mpeters.org> <50AD14F8.8050001@xip.at> <50ADE5E4.9090708@mpeters.org> <50AE0B12.8000309@xip.at> <50AE1CCC.7080706@mpeters.org> <50AE2686.8070007@xip.at> <50AE47D4.7080608@mpeters.org>
next in thread | previous in thread | raw e-mail | index | archive | help
Hi Marc, You definitely have enough information now for a PR. Would you please file all of this into a PR so one of the IP stack people can take a look? Thanks, Adrian On 22 November 2012 07:42, Marc Peters <marc@mpeters.org> wrote: > On 11/22/2012 02:20 PM, Ingo Flaschberger wrote: >> Am 22.11.2012 13:38, schrieb Marc Peters: >>> interesting, the MTU is way lower, than i expected. Through the VPN >>> tunnel, only 1322 bytes are possible without fragmentation. ScreenOS >>> adds 42 additional bytes per paket and the FreeBSD box is receiving >>> 1364 bytes, according to tcpdump. From the outside (only one Netscreen >>> on the way), 1472 is the maximum possible size to send pakets without >>> fragmentation (-D). Which MTU would you suggest to use? Shouldn't the >>> MTU discovery of FreeBSD handle this correct? >> >> do you see fragmented tcp packets on the receiving site in tcpdump? > > nearly every packet is fragmented, if i read th [TCP segment of a > reassembled PDU] correct. Those have a length of 1364. Thera are also > lots of [TCP Window Update] (every two to three acks from the receiving > host, where the tcpdump took place). After some time, there were lots of > [TCP Dup ACK] from the receiving host and the throughput went to a crawl > and had lots of retransmissions. After 30 sec. everything went back to > "normal" as of transmitting pakets without dubs and retransmissions. > > The same tcpdump collect on the Linux hosts, the packets had a length of > 1514 and no dubs or retransmissions. > > I didn't get the option JUMBO_MTU removed, but removing TSO4 didn't > changed anything. > > Adding packetfiltering and the pipe didn't change anything, too. > >> >> When you load the tcpdump data (tcpdump -s 1500 -w filename ...) into >> wireshark, you can graph the speed (bit/sec, packets/sec) and do some >> more tcp analysis. >> >> Kind regards, >> Ingo Flaschberger >> _______________________________________________ >> freebsd-net@freebsd.org mailing list >> http://lists.freebsd.org/mailman/listinfo/freebsd-net >> To unsubscribe, send any mail to "freebsd-net-unsubscribe@freebsd.org" > > _______________________________________________ > freebsd-net@freebsd.org mailing list > http://lists.freebsd.org/mailman/listinfo/freebsd-net > To unsubscribe, send any mail to "freebsd-net-unsubscribe@freebsd.org"
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?CAJ-VmokpTkXg5_O-Ut6k6EUqEPD26vspYjJtP8=P9-EKD6TiBw>