[ale] Today's lesson: rdiff-backup restores

[Lightner, Jeff]

To be clear:  I wasn’t saying you’d do 10 GB for the replication.  I was saying you’d do 10 GB for the initial backups.

The replication is envisioned as a WAN connection for DR purposes but no doubt would work on LAN between buildings.    The replication as I noted previously is a background process to sync the units – not a live copy that runs at same time as the original backup.   A replication that could give you another copy that took hours is still much better than your monthly or our weekly off-siting of tapes.


From: ale-bounces at ale.org [mailto:ale-bounces at ale.org] On Behalf Of Jim Kinney
Sent: Tuesday, February 23, 2016 9:59 AM
To: Atlanta Linux Enthusiasts - Yes! We run Linux!
Subject: Re: [ale] Today's lesson: rdiff-backup restores


On Feb 23, 2016 9:48 AM, "Lightner, Jeff" <JLightner at dsservices.com<mailto:JLightner at dsservices.com>> wrote:
>
> Big boys?   Do you know of a FOSS deduplication (block level) solution?

Other than the ZFS implementation, I'm not certain of it's FLOSS credentials, no.
>
>
>
> Backup to dedupe appliance can be done in various ways.   The two big ones are “inline” and “post-ingest”.   The former has higher resource requirements on the servers being backed up than the latter.   The latter doesn’t take any more resources than backup to tape generally.    You can also do 10 GB fibre on NFS/CIFS mounts or you can do OST over fibre SAN.
>

I can't get 10G between the 2 locations available to me and one location now has a moratorium on new hardware.
>
>
> One of my great annoyances with Quantum DXi was that they made a “firmware” change a few years back that actually turned out to be full upgrade of the embedded Linux they were using (CentOS based I think) wherein they switched from post-ingest to inline.   It killed our backup window to the point I refused to even try using that new “firmware”.  They finally put us back on the old one.   They continually say they’ve made improvements to that newer “firmware” and have better hardware that makes inline work better.    Given this was a few years ago we had the issue it is possible they’re correct now but we don’t use them anymore.
>
>
>
> I liked Data Domain but after EMC bought them we were no longer keen on them for non-technical reasons.    Since Dell bought EMC it will be interesting to see if the culture of Dell overrides EMC or vice-versa.
>
>
>
> At the end of last year we got ExaGrid units and so far we’re quite happy with those.   We’re getting faster backups  over NFS (on 10 GB) than we do to our SAN attached tape library.
>
>
>
> The “Vaulting” I spoke of is a setup that allows us to do copies of all backed up images from one storage unit to another based on the vaulting policy setup rather than having to do manual individual duplications of images.    The catalog that lets us search for backups shows both copies (until the primary expires from dedupe at which point we see only the duplicated image.   In a litigious society having access to old backups is important for most corporations.
>
>
>

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.ale.org/pipermail/ale/attachments/20160223/67ab59a2/attachment.html>