data domain

[mporliod]

Good morning, I wanted to know if it is wrong to use the data domain as the primary repository.
Many have advised against it because they say that restores are very slow, right ?
Are there any recommended configurations when using the data domain as the primary repository ?

thank

[foggy]

Hi Massimiliano, that's right, using any dedupe device as a primary backup target is against Veeam reference architecture, mainly due to performance reasons.

[mporliod]

What types of restores might be slow?
What are the specific problems that I might encounter in using the data domain as the primary repository ?

[Mildur]

I can tell you about my old datadomain.
Entire vm restore had a throughput of 100GB per hour.
It could be, that your datadomain is faster because it‘s newer, but personally, I will never go back to a datadomain as primary storage.
If a vm is down, I need it back running. If restoring a entire 3TB Filer will take somewhere between 10-30 hours, that would be catastrophic for production.

Instant vm recovery could give you also a very slow feeling on the entry level datadomains. I don‘t have any values, but my old datadomain was unusable for instant recovery session. Especially for database systems. They need higher stable IOPS, that‘s not a datadomain feature.

They are really good for long term retention. As a target for a backup copy job. And with their capability to schedule snapshots, there is also some sort of ransomware protection possible for the ddboost unit.

You need to think about your required RTO. If your server needs to be back in production in 1-2 hours, then don‘t use a datadomain as a primary target. Buy a small all in one server with internal disks, do a backup job to that server with a retention of 7-14 days and run a backup copy job to the data domain for long term retention (GFS).
You will get much better RTO for your environment.

[mporliod]

did your data domain use ddboost?
what model was it?

[Mildur]

Yes, it was ddboost. I would never use a dedup appliance with fileshare protocols

We had an old DD2500. As I told, it could be that newer models have a better performance with faster cpus and ssd disks for metadata.
It was replaced this year with linux hardened repos (HPE Apollo) and an S3 Object Storage.

If I were giving dedup appliance for primary storage a second chance, it would be an Exagrid with the landing zone (which is not deduped).

[Mildur]

You can find a discussion about very slow restores speed here. They see the same behavior like myself.

https://www.reddit.com/r/Veeam/comments ... ame=iossmf