On Fri, Oct 17, 2008 at 04:39:05PM +0200, Paul Bijnens wrote:
> On 2008-10-17 15:23, Greg Troxel wrote:
>> Gene Heskett <gene.heskett AT verizon DOT net> writes:
>>


>> There is certainly merit in the hard drive approach, but you can lose
>> two critical properties if you aren't careful:
>>
>>   full backups offline (not writable by computer) even while making next
>>   full backup.
>>
>>   full backups taken to a remote location
>
> I solve these issues by doing backups to an external harddrive.
>
> USB-2 works reasonably fast, if the amount of data is not too large.

I concur. I back up my SOHO network to an external USB HD. I then
rotate two more external USB HDs for off-site backup. Backing to the
off-site machines is scripted and uses rsync. The two off-site HDs are
encrypted with ecryptfs, and the passwords exist in one place.

The backup server is a FIT-PC (http://www.fit-pc.com/new/), so my
power costs are zilch, five to maybe 15 watts for the FIT-PC and two
external HDs (when they're powered up). Try that with your SCSI tape
drives. :-)

As an extra added benefit, I can do an integrity check on all my
"tapes" with "diff -r --brief". It takes a while, but with screen, who
cares?

Another benefit is that I can also copy my bare metal restore data
onto the same off-site HDs, again using rsync and ecryptfs.


>
> I will be experimenting with eSATA "real soon now".
> One of the ideas I have is to make a mirror with LVM of my vtapes
> to an external disk for offsite storage.

I went with rsync because I know it only copies changes. I don't know
whether LVM copies will do that. Also, I can encrypt on the fly while
copying to the off-site HDs. I haven't tried that with LVM.

> That disk gets exchanged on friday each week, and stored offsite.
> Using the USB subsystem makes the server (also small) rather unresponsive.
> Only workable in the weekends and nights.  I hope eSATA will make
> better use of system resources.

Interesting. I'm running a diff right now, and top indicates that diff
itself is far and away the biggest resource hog. The two USB processes
are there but less than diff by two orders of magnitude. Maybe your
server has a more CPU intensive USB controller than mine?

-- 

Charles Curley                  /"\    ASCII Ribbon Campaign
Looking for fine software       \ /    Respect for open standards
and/or writing?                  X     No HTML/RTF in email
http://www.charlescurley.com    / \    No M$ Word docs in email

Key fingerprint = CE5C 6645 A45A 64E4 94C0  809C FFF6 4C48 4ECD DFDB

signature.asc
Description: Digital signature