Recommended settings for offsite backups with WAN acceleration

[pumox]

Hi,

recently we upgraded from VBR 9.6 Standard to 10 Enterprise Plus. The background was, that we wanted to use WAN Acceleration for offsite backups. The sites are connected using VPN with max. 20 Mbit/s over cellular. We have two Backup Jobs, each with 14 restore points, approx. 2TB (VMware VMs) and 400 GB (Windows Agent Backup) that we want to copy offsite. On both sites we have CIFS Repos on Synology boxes.

As we are a small company with limited IT knowledge on this topic we need some advice, how to configure the backup (copy) jobs and repos.

** Main Site **

Current settings:

Backup Job (to Synology CIFS repo in main site over lan)
- Incremental with syntehtic full backups on saturday
- Compression level: Optimal
- Storage optimization: LAN target

Backup Copy Job (secondary target for the backup job to offsite with WAN Accelerator)
- Enable inline data deduplication: yes
- Compression level: auto

Repository at main site

- Align backup file data blocks? yes
- Decompress backup data blocks before storing? yes
- Use per-VM backup files? yes


** Offsite (Windows Server 2019 VM on Synology with WAN Accelerator, CIFS Repo on the same Synology) **

Current settings:

Repository at offsite

- Align backup file data blocks? yes
- Decompress backup data blocks before storing? yes
- Use per-VM backup files? yes

We'd be glad, if you could advice us, what the best settings would be for our setup.

As what we've read, one potential problem with "Use per-VM backup files" is, that if we create another VM in the main site, data deduplication won't work for the offsite backup and the whole new VM would be transfered? We are also not sure, if the repo on the main site can have different settings then the repo offsite, as we want to use data deduplication as much as possible, to lower the amount of data to be transfered offsite.

[foggy]

Hi Jakub, first of all, the source and target repositories can have different settings, this doesn't affect the amount of transferred data, especially when WAN acceleration is in use. Secondly, WAN acceleration ensures that only new blocks (not stored in cache) are transferred to the target, so per-VM backup chains setting also doesn't have much effect even if you add a new VM to the job. And finally, there's no need to enable decompression on repositories as compression is transparent to WAN acceleration and I bet you'd want to save some space on repository anyway.

[pumox]

Hi foggy,

thanks for your answer.

1. "the source and target repositories can have different settings, this doesn't affect the amount of transferred data, especially when WAN acceleration is in use". Ok, thanks. I thought that these settings for both repos have to be the same in order to match as much blocks as possible when copying offsite, i.e. that the option "Align backup file data blocks" on both repos is important.

2. So, should I set compression e.g. to optimal on the Backup Job in the main site, where the storage is on LAN and set compression to Dedupe-friendly only for the Backup Copy Job that sends offsite?

3. As of https://helpcenter.veeam.com/docs/backu ... ml?ver=100 "If you enable the Use per-VM backup files option, data deduplication between VMs will not work. For more information, see Data Compression and Deduplication.". I thought, that if I don't enable per VM files, that I can benefit in case of adding new machines to the backup job, because similar blocks could be taken from other VMs in the job.

4. As of https://helpcenter.veeam.com/docs/backu ... ml?ver=100 "Writing compressed data to a deduplicating storage appliance results in poor deduplication ratios as the number of matching blocks decreases. To overcome this situation, select the Decompress backup data blocks before storing check box.". I thought, that this setting is mandatory for the offsite repo, if the original Backup Job that has to be copied over has a compression level set. So you mean it is not relevant when using WAN acceleration?

Sorry for all these questions, but I'm a bit overwhelmed with all these config options and relations between them.

[foggy]

1. Alignment allows for better data deduplication on the storage itself and doesn't affect data transfer. WAN acceleration uses its own blocks of variable size for deduplication.
2. To optimize traffic between the sites, you need to set compression to Optimal on the backup copy jobs as well. Dedupe-friendly makes sense if you have a deduplication-enabled repository on the target storage and cannot perform decompression prior to writing data to it. If I get your setup right, this is not the case with your setup.
3. That is correct for cases without WAN acceleration. With WAN acceleration, data that exist in WAN cache on target will not be transferred regardless of whether per-VM chains are enabled.
4. Again, this is valid for cases where you have dedupliction appliance as the target repository.

[pumox]

Hi foggy,

thanks. So just to sum up the settings:

Backup Job
- Compression level: Optimal
- Storage optimization: LAN target

Backup Copy Job
- Enable inline data deduplication: yes
- Compression level: Optimal

Both repositories (main site and offsite):
- Align backup file data blocks? no
- Decompress backup data blocks before storing? no
- Use per-VM backup files? yes

Are these settings now optimal for my setup?

And one last question. Should I use forever incremental mode or incrementals with periodically synthetic fulls in the Backup Job or doesn't that make any difference in regards to offsite backups?

[foggy]

Looks good. Please make sure the remote CIFS repository has a gateway server assigned to some server located at the same site (could be the WAN accelerator itself). Here's also our best practices guide for further reading.

[PetrM]

Just a couple of ideas:
- You may try to make an NFS share instead of CIFS on the Synology box to accelerate local backup jobs
- Don't forget to uncheck the box "High-bandwidth mode" in WAN-accelerator settings as it's recommended only for links with bandwidth higher than 100 Mbps.
You should use "low-bandwidth" mode as you have 20 Mbps link between remote locations.

Thanks!

[pumox]

Thanks foggy and PetrM.

Could you please answer my last question, if I should use forever incremental mode or incrementals with periodically synthetic fulls in the Backup Job or doesn't that make any difference in regards to offsite backups?

[foggy]

The backup method used by the source backup job doesn't affect backup copy.

[pumox]

Thank you foggy!