DD Boost & DataDomain newbie questions....

[SteveLaw]

Hi,

I'm confused about how the DD Boost plug-in works. Can anyone help explain?

We are setting up a new backup infrastructure with Netbackup 7.5 on two sites with mixed Windows and Unix clients writing to Data Domain boxes which replicate via SLP across the wan.

We have a capacity license and large servers >50gb data are being installed as media servers with the DD Boost OST plugin. These servers will only back themselves up - other media servers will be promiscuous and pickup other clients at random. At this point each large server will probably need its own storage unit so it can only back itself up. 

1) My understanding is that the DD boost plugin will dedupe at source, so that after an initial full backup subsequent backups will be much smaller. How does it do that? Does it record all the blocks of data sent onto a list, then not send any matching blocks in future backups? Or does the plugin communicate directly with the DataDomain storage boxes?

2) If a server has Netbackup media server and the DD Boost plugin installed, but is backed up by a second media server (which may or may not have the DD Boost plugin installed), will that backup be deduped at source or not? In other words, will the first media server dedupe its data then send it on to the second media server to then be sent to the DD storage box?

3) Forgetting dedupe at source for the moment, as I understand things once an original full backup is written to the DataDomain box then future backups of the same client are radically 'squashed' after they arrive - only blocks of data that aren't already present in the original full backup are retained. (Kind of like full and differential or incremental backups work on a file level). But what happens if you eventually write your original full backup (written for example on 1st October) off to tape and its retention period expires? All the subsequent backups rely on the 1st October template for their integrity, but if the original is no longer present on the DD box then they can never be 'reinflated' as full backups in their own right.  :o

Is the data Domain box a lot cleverer than this - does it continually keep the blocks it needs and discard those it doesn't, so even if you wrote to tape and deleted the 1st October backup those blocks that were still needed for subsequent backups would be retained?

Thanks!

Confused of Hampshire

[SteveLaw]

An extra question!

What's the best way of confirming that DD Boost is actually doing something? Is there a log or alternative way of getting some stats and percentages out and wave in front of senior managers?

[Nicolai · Accepted solution]

1: Not quite right. Data Domain with boot need to transfer all the data to the media server that will then transform the data into hash vales and then only send new data block to the data domain itself. If you want to dedupe at the client you need either Netbackup MSDP or Symantec 5020 appliances.

BOOST communicate with the Data Domain about unique data blocks.

Maybe this video will help : http://www.youtube.com/watch?v=A0vKfAcIve4

2: I do not know - I expect data will not be resend

3:The data domain will only relase a data block if it's no longer referenced by other backups. This is what the weekly cleaning does.

Bonus qustion: Go to the data domain GUI under Data Management -> DD BOOST -> Staticis

Or ssh to the command line and run command 

ddboost show stats interval 2