Dear esteemed experts,
We are running DSA backups for approximately 6 months. Now we want to improve our backup strategy for PRD, than what we currently have, to fully benefit from the DSA features and capabilities.
For PRD our shop has a 2690 appliances (4nodes) and about 25TB of compressed data. (all BLC hardware compressed) We currently run on TD15.10 and have DSA 16.20
Our current plan was based upon the old strategy at the times we used ARCMAIN.
- Every Sunday, except the last month of each month, a full backup of the entire PRD machine. retention 5weeks.
- Every last Sunday of each month, a full backup of the entire PRD machine, retention 9months.
- Every day, a functional backup of a selection of tables and some full databases, work area's, sandboxes. retention 2weeks.
Of course we did not just copied / migrate the arcmain scripts into DSA, but made new DSA jobs to benefit from the DSA capabilities.
What I had in mind was the following strategy:
Monday Tuesday Wednesday Thursday Friday Saturday Sunday
D D C D D D F
(D=delta, C=cumulative, F=full)
I heard some comments, that for a small system like ours, Delta backups will not be so beneficial but cumulative ones can be considered. (of course it will require testing). There would be a lot of overhead etc. etc. I'm still trying to figure out what and why, because i have not fully understood it I think.
How do others handle this, also having a small system like ours?
What is it they mean with overhead of delta's vs cumulative?
What are good common backup strategies using DSA?
I'm eager to hearing all your thoughts, comments, suggestions ...
Both Delta and Cumulative backups are changed block backup, not changed row backup. So benefit is primarily for large tables where only a fraction of the blocks have any changes. Typically to see any benefit you need to design the tables with partitioning that will limit the number of changed blocks.
Cumulative backups would include all changed blocks since the last Full backup; Delta backups only include changed blocks since the last backup of any kind. So Delta could potentially be faster. The added overhead is mainly on restore, since you would need the most recent Full, most recent Cumulative (if any), and any more recent Deltas.
Thank you for answer.
I have a another question in my mind, regarding the strategy i mentioned. Suppose you take the backups with a retention of 5weeks. First the full backup will be kept 5 weeks, next I suppose, logically the other types taken (delta or cumulative) for that cycle, will have the same retention, hence also 5weeks, correct?
But a certain point, the full backup will reach 5w and expires, rendering the rest of the week's cycle unusable, because of the delta/cumulative types no longer having their base backup.
That would mean with a backup cycle having a retention of 5w, you could go only go back 4w??? or am i seeing this wrong?
How do others generally manage that?
Correct - you can't go back further than the oldest full backup.The "orphan" incremental backup files (also referred to as dependent save sets) are invalidated and DSA won't let you use them.
I suspect most sites just allow for this when sizing backup capacity and setting retention.