[ Date Index ] [ Thread Index ] [ <= Previous by date / thread ] [ Next by date / thread => ]
On 07/09/12 13:44, Julian Hall wrote: > > Basically the main > measurement is by what Windows says, however I have done it previously > the hard way, with a stopwatch and timing how long it took to transfer > 60Gb - result that time was 5Mb/sec, so there is definitely a problem > somewhere. Indeed but if it is SMB some of the problem might be the protocol level. If you have to use SMB use large individual files. I tend to use "ftp" for bandwidth testing, but Kai was using some udp transfer utility which does achieve closer to bandwidth. I vaguely recall SMB over untuned gigabit ethernet gave me 25Mbps for copying large hierarchies of small files. The issue is the protocol is chatty and the ethernet adaptor default settings were fine for 10Mbps, jumbo frames and some other tweaking improved things closer to disk drive limits but a long way from optimal. Also you may want to use tests that fit in RAM so you can run them twice and throw away the first figure which will include disk subsystem performance. -- The Mailing List for the Devon & Cornwall LUG http://mailman.dclug.org.uk/listinfo/list FAQ: http://www.dcglug.org.uk/listfaq