Veeam + DataDomain old issue resolved in 5 ? Clarification

[stormlight]

Hello,

1.What is recommended for Veeam 5?

In the Veeam FAQ is states to enable built-in de-duplication
http://www.veeam.com/forums/viewtopic.p ... 745#p20745

In the DataDomain Veeam integration guide it says the opposite. (However this paper was for 4.0)

DataDomain Veeam guide:
Ensure that inline de-duplication is disabled and that compression is set to None for each configured job.

2. In this post there was an issue with DataDomain and Veeam Synthetic backups based on how DataDomain suffers because Veeam wants the backup to re-assemble. Is this still an issue with DataDomain and Ver 5? If so what are the latest recommendations on how to backup to DataDomain in regards to synthetic fulls forward incrementals ect.
http://www.veeam.com/forums/viewtopic.php?f=2&t=2131

Also note in the same DD/Veeam paper referenced in question 1 DataDomain said this.
Note: The Veeam synthetic full backup method may result in a smaller increase in the Data Domain system’s compression factor over time compared to other VMware backup applications.

Which conflicts or what is isnt recommended by Gostev or other posters due to performance issues.

3. In a number of post I have seen 3 ways to backup to a data-domain. I only understand the cifs can I get a little help with the other 2.

1. DataDomain as a Cifs share (tell Veeam this computer and point it to a my DD cifs share \\datadomain\cifsshare\)
2. A linux box with the DataDomain mounted to it. (Add that linux server to Veeam destinations servers and then backup to the linux box which in turn redirects it to the DD Kind of like a proxy)
3. Mount the NFS DataDomain share to the ESX host. ??

I really don't get 3. In the GUI you have to select a specific host and browse to /vmfs to find your mounted NFS DataDomain target thats connected to ESX. What happens to a VM when its not on that specific host that you choose? In addition does anything change with Esxi since you cant upload anything into the service console? Vranger called this direct to target and didn't make you select a particular host. They allowed you to choose the same data path to the DataDomain NFS target that all host would have access to.

thanks

[Gostev]

From Veeam perspective, recommendations are all the same as you have already found on this forum. There were no changes in v5 as it regards DataDomain interop. We have not tested many DataDomain models with v5, so I can only judge based on customer reports, which I've seen both negative and positive. I understand that on lower-end DD storage, backup speed suffers because the storage does not perform well. However, with higher-end backup DD storage, customers were reporting great results. I would recommend that you test with your specific device.

3 is basically a legacy method, just a variation of 2, as "fat" ESX is not really different from Linux server. You should not consider deploying this method at this time, because we will be in ESXi-only world in just a few month. For performance reasons, I would strongly recommend option 2 (NFS via Linux box). 1 is not ideal both because of DD specifics.

Thanks.

[stormlight]

Thanks as always Gostev,

However I still have a few unanswered questions from my post.

enable or disable veeam de-duplication when backing up to datadomain. Listen to your faq or datadomain paper?

It wasn't clear on what was the actual recommendation. There was talk about bringing back traditional full and incremental in 4.5 or 5 using that, as well as talk of doing fulls each night as the data-domain will discard what it doesn't need inline.

Can you state what is recommended?

[Gostev]

Actually, I have answered these questions already, but I guess I need to elaborate to make them more clear.

It is totally up to you who to listen to. Our product configuration recommendations are from perspective of actual product supplier. These recommendations will allow you to achieve the best results using our product (for example, shortest backup window). I cannot comment on DD guide, because I never actually saw it (was never run by me, but instead produced by DD on their own). I have no idea what their recommendations are based upon.

I am also unable to answer what backup mode would be better for your DD device and your specific environment, because I did not test v5 with different DD models. This is why I suggested that you simply test what works best for you, depending on your requirements. You can refer to backup modes comparison in sticky FAQ topic to help you choose between different backup modes depending on your needs. This is storage-agnostic comparison though, so results may vary depending on specific backup storage make and model (which is why again, you should simply test what works best for your environment and suits your requirements best).

Thanks.

[jgremillion]

I just implemented a DD670 with Veeam v5 and left the inline deduplication turned on, and turned compression off. This seems seems to be working well with our DD.

The compression ratio isn't wickedly high, like DD likes to tout, but I am get an overall 6.1X and climbing (after 5 weeks of Data) with a 24 hour compression anywhere between 7 - 12x. Also, our backup speeds have increased by about 2%.

I backup 30 TB of data from VMware and store 30 days (weekly fulls, incrementals in-between) of restore points on 30 TB of DD space. Plus and am only using 35% of the space on the DD. ....
So far all, the hndful of test restores I have done work good.

I don't really care if Veeam, or DD is doing the dedupe/compression, I just want to be store a good amount of data in a smaller footprint.

[stormlight]

You have synthetic fulls on when you do your weekend fulls? If so do you see that your weekend full takes longer then your first full backup or that your daily incrementals take a long time?

You writing to the data domain via cifs from your veeam box?

I am leaving veeam duplication on as I don't see the dif in turning it off. However I need to read more about what it is de-dupping before I can say. (and how the de-dup changes with the block size/ local target, wan target, LAN target)

[Gostev]

Veeam dedupe is well covered in the sticky FAQ topic, in case you have missed. Thanks.

[jgremillion]

No, I don't do any synthetic fulls. We Just run the traditional incremental.

Like The Maestro Gostev said, there is a lot of info in the sticky FAQ at the top of this forum.