what's performance impact when we start offload the data to object storage

[apolloxm]

Hi,

Currently, We only have one veeam server. all the roles are installed in the same server. The backup server was a VM machine, all ISCSI disks come from QNAP storage works as SOBR repositories. We only created backup jobs.We are planning to start offload data to object storage with immutable.for the offload jobs policy we may choose copy mode, May I know where the offload data come from? Is it read the data from SOBR repository? What is the performance may impact for us?

[Mildur]

Hi Apollo

Offload traffic flow in your environment:
- QNAP > Backup Server > Object Storage

What is the performance may impact for us?

This depends on your infrastructure and what object storage you use. If it's a public object storage (like AWS, Azure or Wasabi), then the internet upload is also important. Without using throttling rules, the offload could block the internet connection for your end users, if the max upload is to low of your internet connection.

Thanks
Fabian

[apolloxm]

Hi Fabian,
Yes, it's a publick object storage.Except for the Internet,is there any other will be impact? Like read I/O from QNAP Strorage disk?

[Mildur]

Can I ask, what your internet upload speed is?
Because I believe, most likely your internet upload will be the bottleneck and not the I/O from your QNAP. If you can restore backups already in an acceptable speed, then the offload to a object storage will not be an issue for the QNAP.

[apolloxm]

it's 10Gb connection.

[jorgedlcruz]

Hello,
To send data to Capacity Tier, Veeam needs to read all the backup files, and slice them into blocks (usually up to 1MB but depending on the compression options). 10GbE to the Internet is a great speed, is it your Veeam Infrastructure on a data center, etc?

With 10GbE you can get around 950MB/s up to 1GB/s meaning transferred data. As Veeam will do multiple uploads/put from a single backup file, you should see the great speed in your scenario.

To your question, I think the real bottleneck here will be the QNAP Disks, as usually the read on QNAP, depending on RAID and disks, are quite slow.

See how it goes, and how the Backup Repository with those iSCSI disks performs when doing the offloaded task. Perhaps adjust the "concurrent tasks" under the object repo as well, not because the Object will have limits, but because your on-prem repository, another way of limiting the concurrent offload tasks will be from the normal repo.

Keep us posted