Comprehensive data protection for all workloads
Post Reply
Keyser
Enthusiast
Posts: 50
Liked: 9 times
Joined: Feb 13, 2014 10:11 am
Contact:

Forward Incremental Feature Request

Post by Keyser »

Hi Veeam

I setup a lot of Veeam Solutions with my customers, and recently I have been involved in using Deduped storage as a backend.

When creating backup jobs I have stumbled upon a need for an additional feature in some setups.

I have several customers with huge amounts of data compared to their I/O needs, and subsequently the storage backend for Veeam contains lots and lots of large but slow harddrives.
The problem is that the amount of data makes it impossible to create synthetic fulls or do Active Fulls regularly due to the time involved in handling the several hundreds of Tbytes.
On the other hand, doing reverse incremental backup's is just to I/O consuming on the backend and thus forces the jobs to run for far to long.

I think you need a third scheduling option to the FORWARD INCREMENTAL feature.
Right now I can select:

1: Synthetic Full which essentially asks my backend to copy the entire full backup to a new full (Way to time consuming when you have huge amounts of data)
2: Active Full which copies all the data from my production to the backend (LOTS of time and IO involved)

I think you need a third FORWARD INCREMENTAL option allowing me to essentially schedule transformation IO on the backend to run in production hours (where i'm not doing backup).
You could call the option "ROLLING CONSOLIDATION". Basically it should allow me to select normal forward incremetal backup, and then allow me to select a schedule for when to apply the oldest VIB file to the VBK file and thus deleting the oldest restore point. The schedule option could be something like: Right after the backup job, or Clock scheduled.
The point of this model is that I have a backup model that resembles Reverse Incremental in space usage and numbers of FULL backups on my backup. But i Have the backup window performance of Forward Incremental, and I never have the need to handle the full amount of data in the job when doing consolidation/backup to allow the oldest restore points to be deleted.

I'm aware this will make restores slower due to always having my FULL backup file in the end of a long chain of incrementals. But that is a decision for me to make on a per job basis which makes it possible to prioritize.

I hope my request is making sense.
veremin
Product Manager
Posts: 20270
Liked: 2252 times
Joined: Oct 26, 2012 3:28 pm
Full Name: Vladimir Eremin
Contact:

Re: Forward Incremental Feature Request

Post by veremin »

Are you essentially after the same mechanism that we already have for backup copy job where the following increments are inserted inside the oldest restore point (.vbk)? Thanks.
Keyser
Enthusiast
Posts: 50
Liked: 9 times
Joined: Feb 13, 2014 10:11 am
Contact:

Re: Forward Incremental Feature Request

Post by Keyser »

Hi Eremin

Yes, that is exactly the behaivour I'm looking for - with the added option of allowing me the schedule when the block injection and deletion of the oldest VIB file is done.

From a performance standpoint, this option would make a LOT of sense for some backup's that contains very large amounts of data.
veremin
Product Manager
Posts: 20270
Liked: 2252 times
Joined: Oct 26, 2012 3:28 pm
Full Name: Vladimir Eremin
Contact:

Re: Forward Incremental Feature Request

Post by veremin »

Got it. The functionality you're talking about (backup copy retention mechanism for ordinary backup job) makes sense, indeed. Thank you for the feedback; much appreciated!
Gostev
Chief Product Officer
Posts: 31460
Liked: 6648 times
Joined: Jan 01, 2006 1:01 am
Location: Baar, Switzerland
Contact:

Re: Forward Incremental Feature Request

Post by Gostev »

Hi, Keyser. This backup mode is available for backup jobs in v8, and is actually the default backup mode for the newly created jobs starting v8.
Keyser wrote:I'm aware this will make restores slower due to always having my FULL backup file in the end of a long chain of incrementals.
Not necessarily. In the long run, I don't expect significant differences comparing to the existing backup modes that keep a single full backup file on disk. Remember that transforms noticeably fragment full backup file over time. The fact that your recent restore point data is logically located in a single file does not really matter when it comes to getting the required block from the disk. It is physical block placement on disk that determines the amount of IO operations required to read the entire VM image from the backup file.

Thanks!
Post Reply

Who is online

Users browsing this forum: No registered users and 207 guests