On Tue, Jan 11, 2000 at 11:26:05AM +1030, John Pearson wrote: > On Mon, Jan 10, 2000 at 11:20:06PM +0100, Carel Fellinger wrote
> > using ncftp I get the expected 1.0+MBs transfer copying a large file into > > /dev/null. Quite reasonable on a 10Mbs ethernet considering ftp and tcp each > > adding their things. But if I do dd -if=/home/biggy of=/dev/null it takes > > 10 times more then doing it locally and only fills a little more than > > half the ethernet bandwith! how come? would it help to switch to 100Mbs > > Ethernet? or has nfs so much overhead (but ftp does okee)? or is the 486 > > the culprit? > > > > NFS is stateless and thus has significantly higher overhead than > (say) ftp (although, being stateless has its advantages - you > can reboot the server and suffer nothing worse than a long pause). Okee, so there is more overhead. So more bytes have to be transfered. But almost doubling it seems a bit overdone, doesn't it. So I'm still wondering... Is the overhead mainly in the extra bytes to be sent, then a 100Mbs Ethernet card would improve things. Or is the overhead also in the pre/post processing, so a faster computer is wat's needed? -- groetjes, carel
