Unsolved
This post is more than 5 years old
116 Posts
0
1058
weekly, monthly full (daily incr.) vs. restore speed (DD)
Dear All,
Are you using monthly 1 full daily incr filesystem backups with NW+DD?
Makes it any sense to do the weekly fulls (or at least to define some level N backups within the month in case of monthly 1 full) - or is it sufficient to have only 1 full in a month + incr. on all other days if the backup target is a DD?
According to the admin guide the restore from incr. or level n backups is slower because more volumes are needed. I belive this thing comes from the "tape times"..
What kind of schedule do you use in case of monthly retention policy?
TIA,
Istvan
ble1
2 Intern
2 Intern
•
14.3K Posts
1
November 17th, 2015 15:00
I use weekly full, but I also keep data for 14 days (I would keep it less if you asked me). If for some reason I had to keep it for a month, I would once again run full on weekly basis. No need for me to keep data for 2 months (eg. if you had full on 1st and last incr on 30th and you keep it for month - it means that full would be kept until 31st of next month due to backup cycle dependency). Even if your backup goes to disk, restore will take longer with monthly full schedule. However, I would do it on weekly for another reason - security. Mistakes happen and if for whatever reason one ssid is gone, with monthly approach by whole backup cycle is gone and that spreads across whole month. Quite a loss in case that I have to go back. In case of weekly, I still have loss, but 1/5 of it which is something that I can (hopefully) cope with.
bingo.1
2.4K Posts
1
November 17th, 2015 23:00
The question is really how fast you need to restore (what is your service level).
In general it makes sense to use a weekly schedule, if possible.
However, as some customers intend to use the backup as archives, where only a fraction of the total amount changes every full, it might make sense to switch to a monthly schedule. We have 1 server where a full backup of millions of small files would take about 24hrs. This one is backed up using a specific monthly full schedule for the group.
You may ask why we do not use synthetic fulls for that purpose? - because we use 2 DDs as backup devices and SFs only work if all save sets reside on the same DD. However, we configured the backup devices for redundancy (same pools on both machines) so NW will distribute the backup streams randomly. Consequently, SFs will not work any more.
oldhercules
116 Posts
0
November 18th, 2015 00:00
Thanks for the answers!
I also realized that the monthly 1 full may consume more space on the DD because of the increased backup cycle.
I think I'll implement monthly full only for a few huge file systems where the change ratio is not so high but to create the full (and clone it with CCR soon) needs a lots of resources (eg. a DMS system where the full is 5TB+ but the daily incrementals are only ~50GB). Maybe I should also use monthly full / weekly level 1 / daily incr. schedule for safety..
I've read too much DDOS fixes related to virtual synth. backups in DDOS release notes, that's why it is not implemented here..(EMC also offered a DDOS upgrade support from 5.4 to 5.5 because some special steps needed in the upgrade process if the user has synth. full backups).