Re: [rdiff-backup-users] Network performance

[Robert Nichols]

On 05/21/2015 09:50 AM, Laurent De Buyst wrote:
> Hello,
>
>
> I was wondering, if I run rdiff-backup and point it to a remote server,
> how much work gets done by that remote server?
>
> Or to be a bit more specific: would there be a serious difference in
> performance between:
>
> A) rdiff-backup to a remote server
>
> B) use sshfs to mount the remote server's disk locally and use
> rdiff-backup locally
>
>
> I would like to do B because I have a rather piecemeal network and this
> would allow me to handle a lot more clients in a consistent manner, but
> if it's going to have a huge performance hit, then I'll pass.

A lot of CPU-intensive stuff gets done on the server.  That's where the
diffs are generated and compressed, compressed snapshots are made for
files being deleted from the mirror, and, finally, the metadata files
are compressed.  Whether (B) would be a performance hit depends on the
relative CPU power of the two machines.  The client machine is mostly
waiting while that activity is happening on the server.  There would
of course be a lot more network traffic for case (B).

You can really see the server CPU effort when an old kernel version
has been removed, and snapshots have to be made for all of the files
under its /lib/modules directory.

--
Bob Nichols     "NOSPAM" is really part of my email address.
                Do NOT delete it.