DD Boost and Backup Copy

[bleyers]

Hi,

Our current setup is to backup to an EMC nas.
Afterwards we do a backup copy to and EMC DD2500.

Is there a way to use DD boost during the backup copy?
This would really speed the backup copy.

thanks

bart

[veremin]

If the target repository is added properly (as the one supporting DDboost technology) to a backup console, DDboost should work regardless of job type: be it backup or backup copy job. Or you meant something different, Data Domain Replication, for instance?

[bleyers]

i have checked the logs from the DD and indeed I indeed see that DD boost in being used.

The speed of my backup job is between 30 - 60MB/s.
The bottleneck is the target (DD2500)

Are these speed normal ?

[veremin]

Hard to say without more information regarding your source and target infrastructures, such as backup copy job source size, connection between to two sides, full bottleneck statistics (the one with percent distribution), etc. Thanks.

[bleyers]

The destination for backups is an EMC VNXe3200. Has 2x10gb interfaces.

The source for the Backup Copy is the same EMC VNXe.
The destination for the Copy is a DD2500. Also 2x10Gb interfaces.

All in the same vlan, no firewall.
Both on different switches that are interconnected with 2x40Gb

Stats
------

If i look at the global i have 6 -2 - 1 - 36
If i look at individual vms in the job i moslty get for target in the 90 -99 range

The job in currently running so i will update the size when it's done...

[bleyers]

Jobs did not finish because took too long but here are the stats
7% - 2% -2%- 40%

Duration: 25h
Speed: 35MB/s

Processed 2.5TB
Read 2.5TB
Write 1.5TB

[foggy]

I'd say 35MB/s is quite decent speed for DD Boost target. You can ask support to review your job configuration and entire setup with the purpose of possible optimizations.

[Gostev]

@foggy single stream should go well over 100 MB/s normally (for full backup at least).

[bleyers]

I think the problem is the rehydrating on the DD when doing a merge..

I have read the v9 new features and i hope i can resolve this with the feature that allows backup copy to dedupe device
to write the whole backup file instead merging is on the dedupe device

[Gostev]

No, merge cannot be a problem with DDBoost, as the later enables in-storage merge when data is not physically moved (and so no rehydration is necessary). Only pointers to the existing data blocks are updated. Thus, merge becomes extremely fast comparing to one without DDBoost.

[bleyers]

I have a support case open because i have a merge that is running for 83 hours now and is still at 63%.
They say the cause is the size of backup chain AND the fact that it is on a dedupe device...

So no DD boost is used ?

[veremin]

Kindly, ask your support engineer the very same question, as information regarding used mechanism should be present in logs for sure. By the way, what's your support ticket number?

[bleyers]

The case number is 01183408