Date: Wed, 15 Dec 1999 22:21:06 -0500 From: "Mikhail Evstiounin" <evstiounin@adelphia.net> To: <juha@saarinen.org>, "Mike Tancsa" <mike@sentex.net> Cc: <questions@freebsd.org> Subject: Re: Not such good networking performance with FreeBSD Message-ID: <000c01bf4774$979a1820$d3353018@evstiouninadelphia.net.pit.adelphia.net>
next in thread | raw e-mail | index | archive | help
-----Original Message----- From: Juha Saarinen <juha@saarinen.org> To: Mike Tancsa <mike@sentex.net> Cc: questions@freebsd.org <questions@freebsd.org> Date: Tuesday, December 14, 1999 10:30 PM Subject: RE: Not such good networking performance with FreeBSD >Hi Mike, > >> Are you sure its not a duplex issue with your network card ? > >Don't think so -- upon boot-up, the card (a DEC 21140 10/100 clone) is being >put into 100Mbps full duplex. At least that's what the system tells me. >Ifconfig -a seems to imply it's only running in 100BaseTX mode: > >de0: flags=8843<UP,BROADCAST,RUNNING,SIMPLEX,MULTICAST> mtu 1500 > inet 192.168.1.1 netmask 0xffffff00 broadcast 192.168.1.255 > ether 00:00:e8:4a:bf:96 > media: autoselect (100baseTX) status: active > supported media: autoselect 100baseTX <full-duplex> 100baseTX >10baseT/UTP <full-duplex> 10baseT/UTP > > Are your card connected to hub? Hub cannot operate as full duplex device - only half-duplex. I had awful performance on my system (10Mbps interface) both in FreeBSD and NT environment when I set full duplex on my cards - something like 30 times slower. Nevertheless, I saw a switch 10/100 in Best Buy for $120 ( 4 ports, either D-Link or Linksys ). Didn't test it yet. > >> If thats not >> the case, it might be that if you are getting a lot of errors on your xDSL >> connection. FreeBSD does not seem to fair well where there are a lot of >> errors. Have a look through the archives, there was a discussion about >> this a few months ago. > >Thanks -- the RADSL line here is reasonably free from errors, thank >goodness. I've got an external router connected to a switch to which the >rest of my small LAN is hooked up as well. > >I thought it might be the low values for the TCP receive and send buffers -- >16KB by default. I upped them to 65K with: > >sysctl -w net.inet.tcp.recvspace=65536 >sysctl -w net.inet.tcp.sendspace=65536 > >but it didn't really make any difference. > >Going through the sysctl options, I noticed several which may or may not >affect performance: > >Don't know what these do, but I presume they're for LANS: > > net.local.stream.sendspace: 8192 > net.local.stream.recvspace: 8192 > net.local.dgram.maxdgram: 2048 > net.local.dgram.recvspace: 4096 > >Could this one have anything to do with IP routing: > > net.inet.ip.fastforwarding: 0 > >RFC 1323 extensions are useful for me: > > net.inet.tcp.rfc1323: 1 > >Is this the default maximum segment size? > > net.inet.tcp.mssdflt: 512 > >The TCP send and receive buffers: > > net.inet.tcp.sendspace: 65536 > net.inet.tcp.recvspace: 65536 > >Hmmm... Delayed ACKs? > > net.inet.tcp.delayed_ack: 1 > >UDP datagram sizes? > > net.inet.udp.maxdgram: 9216 > net.inet.udp.recvspace: 41600 > >Now what's this then? I increased the raw.recvspace, but I'm not sure what >good it would do: > > net.inet.raw.maxdgram: 8192 > net.inet.raw.recvspace: 65536 > >Where are all these options described? The man page isn't much use. > > >Cheers, > >-- Juha > > > >To Unsubscribe: send mail to majordomo@FreeBSD.org >with "unsubscribe freebsd-questions" in the body of the message To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-questions" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?000c01bf4774$979a1820$d3353018>