Hello list, I have observed quite low bandwidth usage when copying files over SMB. This happens for at least three cases I have tested: * Windows 98 SE reading from a Samba 3.0.22 smbd * Windows XP reading from a Samba 3.0.23d * smbclient reading from 127.0.0.1 (3.0.23d) According to iptraf, the average transfer rate was oscillating between 8604.60 kbps and 8654.20 kbit/s, copying a 172MiB sample file took about 5 minutes. On the other hand, copying the file using a more simple protocol resulted in: $ rsync -PHSav localhost:/tmp/172m-testfile /dev/shm/ 40361 kbit/s, 66 seconds total transfer time In the win98--3.0.22 network I have observed that smbd takes a lot of %system CPU time when transferring files. Since I observe this slowness on three different setups, the question arises: is SMB/CIFS badly designed? -`J' --
On Sat, Jan 27, 2007 at 10:10:36PM +0100, Jan Engelhardt wrote:> In the win98--3.0.22 network I have observed that smbd takes a lot of > %system CPU time when transferring files. Since I observe this slowness > on three different setups, the question arises: is SMB/CIFS badly > designed?Well, is it "designed" at all in any proper definition of "design"? Seriously, smbd should not eat significant amounts of CPU on a reasonably modern machine. Do you have your debug level set too high? Volker
On Jan 27 2007 22:46, Volker Lendecke wrote:>On Sat, Jan 27, 2007 at 10:10:36PM +0100, Jan Engelhardt wrote: >> In the win98--3.0.22 network I have observed that smbd takes a lot of >> %system CPU time when transferring files. Since I observe this slowness >> on three different setups, the question arises: is SMB/CIFS badly >> designed? > >Well, is it "designed" at all in any proper definition of >"design"? > >Seriously, smbd should not eat significant amounts of CPU on >a reasonably modern machine. Do you have your debug level >set too high?Ah I found it. This is smb.conf: http://pastebin.ca/330452 Removing SO_SNDBUF=8192 gives $ smbget smb://localhost/rt/blob.iso [blob.iso] 41.08Mb of 171.06Mb (24.01%) at 41.08Mb/s ETA: 00:00:03 perfect performance again. Wonder how that got in there *grumble* Jan --
On Sat, Jan 27, 2007 at 11:35:32PM +0100, Jan Engelhardt wrote:> > Ah I found it. This is smb.conf: http://pastebin.ca/330452 > Removing SO_SNDBUF=8192 gives > > $ smbget smb://localhost/rt/blob.iso > [blob.iso] 41.08Mb of 171.06Mb (24.01%) at 41.08Mb/s ETA: 00:00:03 > > perfect performance again. Wonder how that got in there *grumble*Yep, on modern kernels I don't think setting SNDBUF or RCVBUF is a good idea.... Jeremy. -- Jeremy - why is this? Ryan -------------- next part -------------- ------------------------------------------------- This email transmission and any documents, files or previous email messages attached to it may contain information that is confidential or legally privileged. If you are not the intended recipient, you are hereby notified that any disclosure, copying, printing, distributing or use of this transmission is strictly prohibited. If you have received this transmission in error, please immediately notify the sender by telephone or return email and delete the original transmission and its attachments without reading or saving in any manner. The Evangelical Lutheran Good Samaritan Society. ---------------------------------------------------------