On 18/01/2021 11:28, Brock Palen wrote:
Disclaimer I have not used s3 with bareos but done many cloud
calculations.
Few things to think about using cloud.
Are you running your SD in the cloud?
Are your backup clients in the cloud?
If not what’s your bandwidth? It will impact your backup and restore
times significantly if you have modest WAN capacity for local clients
servers.
No, no. I was thinking about keeping an extra copy "off-site". I'm
mostly cloud-free at the moment and I do not wish to change it
significantly. I was thinking whether S3 could be an option for
extending my home backup setup.
Of course I understand the impact of bandwidth on the backup/restore
times. :-)
As for s3 pricing read this carefully
https://aws.amazon.com/s3/pricing/ <https://aws.amazon.com/s3/pricing/>
You have three components to pricing with s3 and I expect only two
move the needle on cost.
Data stored
Bandwidth and retrieval
Operations
Opts rations are so cheap and guessing how bareos uses virtual tape
volumes it’s prob not a big issue. Someone who has used it though can
speak.
That's a good observation. Thanks!
Data stored its straight $/gb/month. So you need to estimate your
total data stored for all your fulls and incrementals. Your right
these costs decline when you look at glacier but there is a trade off.
The cheaper to store the more expensive to access.
Retrieval fees come in two forms. The first is bandwidth. Which fit
most people is .09$/gb (unless your clients and servers are in the
same aws region) for my cloud activities this is 50% of my monthly
bill. It’s the thing that messes most cloud calculators for budget.
That said if your sever is on prem you likely will never pay this if
you don’t use always incremental or do any restores. So if your ok
paying for restores maybe it’s ok.
The cold tiers like glacier charge to access data. Again maybe fine if
you almost never read it. Glacier runs $10/tb or more for transfer vs
nothing for regular s3. With bandwidth your at ~$100/tb This is
something to avoid deep archive. Their sla is many hours to get data.
I don’t think deep archive is a backup replacement but a compliance
archive replacement
Well, that's what I'm counting on - it's better to have backup copy and
not need to use it than not having it ;-)
What I was also interested in was also how to approach the long SLA
regarding Bareos SD operation. Would I have to firstly request access to
the glacier data independently of the SD and after receiving
confirmation of data availability would have to run a restore job? Or
would I just run a restore job from storage using cold-tiered bucket and
the job would simply wait for data availability (similar to mounting tape)?
Also be aware glacier and deep archive have minimum retention times of
90 and 180 days. So you will always pay that at a minimum. Ok if your
keeping fulls for a long time. Look at the auto tier options to
manage aging volumes.
Yes, I noticed that
So YMMV. If you are 100% in the cloud or you don’t use always
incremental or have small data volumes or just a dr copy it works great.
Personally I run my servers in aws and my full bareos setup on prem
with a $400 tape library from eBay. This gives me diversity and most
of the data in the cloud is small (websites email text) while the on
prem is video photos and road warriors using always incremental.
So it all comes to "try the free tier and see for yourself" :-) I'll
have to do it anyway when I get some spare time just to see how it works
and get some understanding about achievable througputs, needed space and
so on.
Thanks for valuable insight!
--
You received this message because you are subscribed to the Google Groups
"bareos-users" group.
To unsubscribe from this group and stop receiving emails from it, send an email
to [email protected].
To view this discussion on the web visit
https://groups.google.com/d/msgid/bareos-users/f17e6555-1f48-3d1a-1ba6-f9681d1e8b73%40gmail.com.
