Review Request 112982: copyToFile support for kio_smb

[Mark Gaiser]

> On Oct. 5, 2013, 4:01 p.m., Mark Gaiser wrote:
> > Tested it.
> > PRE patch: ~17MB/s
> > POST patch: ~27MB/s
> > 
> > So in functionality terms this patch makes a file copy from a windows share (note: a linux machine sharing through samba, not an actual windows machine) much faster. However, mounting the same share through CIFS (mount.cifs) gives mu much greater speed with ~86MB/s. Is there anything you can do to make this about equal in speed compared to cifs?
> 
> Dawit Alemayehu wrote:
>     Can you provide the mount options you used? We use libsmbclient for so there should not be such a large difference. I will have to check if there are known performance issues with libsmbclient itself.
> 
> Mark Gaiser wrote:
>     Hi Dawit,
>     
>     I did't provide any mount options besides my username. The actual line:
>     mount.cifs //<ip> data -o username=mark
>     
>     Yes, that's all :)
> 
> Dawit Alemayehu wrote:
>     I found this gnome vfs change: https://mail.gnome.org/archives/commits-list/2009-August/msg06304.html. It also checked to see if that code remains the same today and it seems like it does. See http://code.ohloh.net/file?fid=ZGtFwvN9JyhO9xIWTTQS-H6jau8&cid=Wya2WEWTTh4&s=gvfsbackendsmb.c&pp=0&ff=1&filterChecked=true&fp=309479&mp=1&ml=1&me=1&md=1&projSelected=true#L0
>     
>     Do you get comparable speed if you change MAX_XFER_BUF_SIZE definition in kio_smb.h to 65534?
> 
> Mark Gaiser wrote:
>     Well, your patch changed it to "65536". I doubt decreasing it to "65534" will magically make it equally fast as mount.cifs. My "POST patch" result is with MAX_XFER_BUF_SIZE set to 65536. One thing i do notice in searching for CIFS is that it's advertised as the SMB successor. Perhaps it's just better and faster?
> 
> Dawit Alemayehu wrote:
>     No. It is the same thing. CIFS is just a new name for the same old SMB protocol in its modern incarnation. If you are testing it against a Windows share and not a Samba share, then that might make a difference. You can also try to double the buffer to 128 KB and see if that makes any difference.

I will try that tomorrow.

Also, i don't think it should make a difference if i copy from windows or from a samba share. I just did test a copy using Windows 7 as the destination and the same samba share as source. Even that gives me slower speed then mount.cifs although just slightly. Using Windows 7 gives me ~80MB/s. Before i add confusion, this has nothing to do with your patch, just comparing copy speeds on my pc but with different distributions.


- Mark


-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
http://git.reviewboard.kde.org/r/112982/#review41277
-----------------------------------------------------------


On Oct. 5, 2013, 3:07 p.m., Dawit Alemayehu wrote:
> 
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> http://git.reviewboard.kde.org/r/112982/
> -----------------------------------------------------------
> 
> (Updated Oct. 5, 2013, 3:07 p.m.)
> 
> 
> Review request for KDE Runtime.
> 
> 
> Bugs: 176271 and 291835
>     http://bugs.kde.org/show_bug.cgi?id=176271
>     http://bugs.kde.org/show_bug.cgi?id=291835
> 
> 
> Repository: kde-runtime
> 
> 
> Description
> -------
> 
> The attach patch adds support for the following to kio_smb:
> 
> - copyToFile optimization so downloading files from window shares is faster.
> - partial download resumption as part of the copyToFile implementation.
> - preservation of modified file timstamp. Again as part of the copyToFile implementation.
> 
> Note that in this patch the latter two features only apply to "smb" -> "file" downloads. The second part of this patch will that will follow soon will add support for the other half, the "copyFromFile" optimization.
> 
> 
> Diffs
> -----
> 
>   kioslave/smb/kio_smb.h 55efb44 
>   kioslave/smb/kio_smb_dir.cpp 5573266 
>   kioslave/smb/smb.protocol 654bcfb 
> 
> Diff: http://git.reviewboard.kde.org/r/112982/diff/
> 
> 
> Testing
> -------
> 
> 
> Thanks,
> 
> Dawit Alemayehu
> 
>

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.kde.org/pipermail/kde-core-devel/attachments/20131005/5b4c73c8/attachment.htm>