On 1/13/2010 8:51 AM, Sorin Srbu wrote:

...

Between compression and pooling, I get about 10x the raw data being archived with backuppc - it beats juggling tapes and you can let the users access the backups of their own machine through a web interface. There are some down sides to plan around though: the compression takes some CPU and is slower than a stock rsync run, and the pooling is done with hardlinks which forces the archive to be on a single filesystem and makes it hard to duplicate for offsite copies. There's an RPM in epel that is easy to install on Centos.

Sound very interesting indeed!

I don't think the performance will be a problem, the server's a calculation machine that has now been scrapped running a dual-xeon@2,something GHz and some 4GB RAM IIRC. Do you think the software-raid5 array used, would be a problem in this case?

I've never had any problems with software raid5 in linux before, but you never know...

There's a big write performance hit from raid5 (software or not). It may not be enough to be a showstopper but I wouldn't recommend it. Can you reconfigure to a 0+1 or some other type that has better performance without losing too much space? The archive does have to be on a single filesystem, though, and if you use the epel RPM it makes things easier if you mount the volume at /var/lib/BackupPC before the install. See 'how to change archive directory' from: http://sourceforge.net/apps/mediawiki/backuppc/index.php?title=Tips_and_Tric... to do it after the install.