Re: diff between incremental and archive backups

  • From: Tim Gorman <tim@xxxxxxxxx>
  • To: oracle-l@xxxxxxxxxxxxx
  • Date: Sun, 15 Nov 2015 08:28:34 -0700

Final comments (hopefully) on this thread...

Now you've stepped squarely into *my* area of interest as an employee of a vendor <http://www.delphix.com>.

Employing backups to copy production databases and application software to non-production environments like development and testing is a holdover from the 20th century, and is fully obsolete. Those continuing to provision non-production from restored backups or from SAN copies are placing themselves at a disadvantage to those who have adopted data virtualization <http://www.delphix.com/products/how-delphix-data-virtualization-works/>, plain and simple.
Tim, I don't think that this thread is the proper venue to discuss this, but "data virtualization" is a nice marketing buzzword which has a ZFS snapshot as its foundation. Commvault is using many of the same technologies as Delphix, but we are not Oracle specialists. We also backup MS Exhange, SharePoint, DB2, SAP Hana, MySQL, Postgres and many other things. Essentially, our aim is to cover the entire enterprise, not just Oracle. I don't think that this is an appropriate venue for vendor comparison, especially since the two participants in the discussion are both employees of their respective backup vendors. I have enormous amount of respect for both you and Kyle Hailey. Such debates do not make much sense without the impartial 3rd party, like Gartner Group. And by the reaction to my statements about the incremental backups, you can tell that Commvault technology far surpasses the traditional backup/restore technology, just as is the case with Delphix. Our marketing vernacular is slightly different, though. I will refrain from using it, since this is not the right venue for such things.

Discussion veered organically into this area, and so it is absolutely relevant. Just because vendor products other than Oracle are front and center does not automatically end discussion. And you, you sly devil, were comfortable touting Simpana at the outset of this thread as an employee of Commvault when you felt it became relevant to the discussion. You were even comfortable just now to continue to tout Commvault on other databases, etc, in the midst of *scolding* me? That's called chutzpah, my friend... :)

My point is that provisioning non-production environments from production backups is obsolete and inefficient. Moving on...


The original discussion was about incremental backups, plain and simple. As far as the incremental backups are concerned, if the underlying technology is rman, those do not make much sense. The only argument you listed is that they are cheaper than full backups to execute, which doesn't really matter, if the backups are taken on the proxy. Proxy box like a standby DB is dedicated recovery and it doesn't really matter whether the backup will burn 30 minutes of CPU time or 3 hours of CPU time. That time would have not been used otherwise.

Backups taken on a proxy absolutely do matter. Nothing is free. Everything adds up, and has a cost associated. There is no part of IT where you can wave your hand and say "Oh, that doesn't matter, it is free". In current IT organizations, unused CPU and RAM cannot be said to be "unused otherwise" due to virtualization. In fact, that's the main point of almost all forms of virtualization.

It is not compute, but the storage framework (and possibly network as well) that is being hammered by full backups. Storage is almost certainly a shared resource on a SAN; direct-attached storage for databases disappeared in the early 1990s. Nothing done on a SAN is done in isolation, and the sudden appearance of a frequent massive workload caused by daily full backups of multi-TB databases is not going to go unnoticed. Nor unpunished.

-------

Look, I think we got the message: incremental backup strategies are a tradeoff. You feel that not enough attention is paid to the costs on the restore side of that tradeoff, and that you know a solution, and you've brought attention to that. I feel that RMAN incremental backup strategies are a valid choice in the tradeoff. I agree that one must understand the implications of terms like RTO (recovery time objective), and how different mechanisms can be used to achieve that objective.

Let's move on?


Other related posts:

  1. Home
  2. oracle-l
  3. Archive
  4. 11-2015