Return to BSD News archive
Xref: sserve comp.protocols.tcp-ip.ibmpc:18673 comp.os.386bsd.questions:5025 Newsgroups: comp.protocols.tcp-ip.ibmpc,comp.os.386bsd.questions From: daveh@fusion.demon.co.uk (David Hodgkinson) Path: sserve!newshost.anu.edu.au!munnari.oz.au!news.Hawaii.Edu!ames!agate!doc.ic.ac.uk!warwick!qmw-dcs!qmw!demon!fusion.demon.co.uk!daveh Subject: Re: Data transfer rates - strange differences! Keywords: ftp, tcp, transfer rates Organization: Fusion Systems Ltd. References: <25na11$51s@olivaw.apanix.apana.org.au> <1993Aug30.174939.16586@adobe.com> <hastyCCL7K1.628@netcom.com> Date: Wed, 8 Sep 1993 11:35:46 +0000 Message-ID: <CD19Jn.AqE@fusion.demon.co.uk> Sender: usenet@demon.co.uk Lines: 30 In article <hastyCCL7K1.628@netcom.com> hasty@netcom.com (Amancio Hasty Jr) writes: >In article <1993Aug30.174939.16586@adobe.com> jsundar@adobe.com (Jagane Sundar) writes: >> >> I have noticed similar differences in transfer rates. >> In using the same PC with both 386bsd and DOS/Windows, I >> found that I could get from 400K to 600 K using 386bsd(600K >> with the new ed drivers). However the same machine gives as little >> as 100K with the "best" commercial windows TCP package. Its because >> of Windows build in stupidity. Sounds like a load of bollocks to me :-) I have routinely seen 400k-700k between PCs. Have you tried to investigate _why_ this might be the case? Often, the default settings for buffers are really puny - for example PC/TCP has about 4-6 small and large buffers by default. If you are pushing for streaming transfers, you might want to go to ten or twenty. Have you looked at the TCP stats? Dropped packets? Duplicated packets? Have you tried 'put' to nul rather than a file? That is a fairer measure of throughput. Just my .02c :-) Dave -- Dave Hodgkinson - Wizard for hire. +44 71 377 9009