Forum Discussion

RLeon's avatar
RLeon
Level 6
14 years ago
Solved

Netbackup 7.5 Accelerator and Incremental Backups

Hi all,

Just wondering if I'm getting this right.
For simplicity, let's say the client is running on the Windows platform:

Full backup = File level all files

Incremental backup (both types) = File level changed files (ones with archive bit set)

Accelerator Full backup = Block level changed blocks since the last all-blocks Accelerator backup

Accelerator Incremental backup = Same as the above except that less catalog space is used

So in effect both Accelerator backups are like Incremental backups, but ones that operate on the block level instead of the file level.
And just like the normal Full and normal Incrementals, an Accelerator Full backup uses more catalog space than an Accelerator Incremental backup because it records all files instead of just the changed files. (The catalog operates on the file level when it comes to remembering things in a volume for restores.)

Did I get that right? Is there more to it?

Thanks all,

RLeon

  • I think that :

    1. VSS is used (by default) on all windows backup, with accelerator or not. 
    2. Deduplication is "global" (database is on the media server) but you have a local cache on your client (with source deduplication). So if you have similar block to previous in a new file, accelerator will send unknow blocks only, and flashbackup will send all new blocks.
    3. If you rename a file :
    • (not sure) incremental flashbackup will send all blocks of the file
    • (sure) accelerator backup will process the file, but send 0 block because they will all be in the local deduplication database 

    Regards

5 Replies

  • Ii think that your explanationis pretty good but I would have said, about backup with accelerator, that we are on file level checking of change (with OS jounalized FS) but with deduplication, you will only send new blocks of data.

    Also, Netbackup will do virtual synthetic : create full image or cumulative incremental backup with these new blocks.

    You are right when you say that you will have less catalog entries with incremental backups because you have less files in the image. 

    Note that with Netbackup 7.5, image database is a relational database. So, i don't know if Symantec developped optimization to keep only once filename, and will keep only pointers on it in database entries of your backup image (i don't know is it is vrey clear ;-) )

  • Thanks for your input Gautier. I understand that the part about "file level checking of change" is just like Flashbackup.

    I was looking through the 7.5 documents about whether the Accelerator deduplications are on the global level (i.e., Global Deduplication), but couldn't find anything solid. I guess it is implied.
    It would be a waste of resources if it is not Global Deduplication.

    With Accelerator, only changed blocks are backed up.
    With Flashbackup Incrementals, only changed blocks are backed up.

    Am I right in saying that the only difference is that Flashbackup uses snapshots and Accelerator doesn't?
    Can Accelerator be used with snapshots? If so, wouldn't the effect be exactly the same as when you use Flashbackup Incrementals MINUS global deduplication?

    Thanks for your help,

    RLeon

  • I think that :

    1. VSS is used (by default) on all windows backup, with accelerator or not. 
    2. Deduplication is "global" (database is on the media server) but you have a local cache on your client (with source deduplication). So if you have similar block to previous in a new file, accelerator will send unknow blocks only, and flashbackup will send all new blocks.
    3. If you rename a file :
    • (not sure) incremental flashbackup will send all blocks of the file
    • (sure) accelerator backup will process the file, but send 0 block because they will all be in the local deduplication database 

    Regards

  • Thanks for the info Gautier,

    From what you said, it would seem that Accelerator would be a lot quicker than Flashbackup Incremental because of the following two points:

    1. Accelerator would send only the changed blocks of a file, when Flashbackup Inc would resend all the blocks of the file even if only a portion of its blocks have changed. That would mean, with big files, Flashbackup Inc would be alot less efficient compared to Accelerator.

    2. Accelerator further decreases the total number of blocks that needs to be sent by being able to use a global level deduplication pool. (As opposed to a single client/host level blocks/hash comparison)

    Regards,

    RLeon

  • I think that FlashBackup will send only changed block of the file (and all block if the file was renamed). But in some type of files (compressed, crypted or database files), a little change will change all blocks.

    Accelerator will only send changed blocks that are not in the (global) deduplication database. So it will more effecient that FlashBackup.