VBR13 MSSQL Plugin Backup in Azure

[PospiKM]

Hello,

I would like to discuss or get better overview of community experience with MSSQL Plugin Backups in Microsoft Azure that are maintained from the VBR (version 13) side, not standalone. As I'm new to such concept, I am very curious about some performance topics.

We use Azure Blob storage as a target for MSSQL backups and for now we have several policies (there will be much more in the future), backing up tens of servers (hundreds in the future).

What is the ideal or recommended configuration to get the best performance from the backups? Especially talking about the number of backup stripes/channels/parallel processing (data/logs) versus data reduction?
What is the maximum recommended MSSQL servers/instaces in one backup policy in terms of performance?
What is the recommended number of datamovers/gateways versus number of MSSQL servers/instaces to backed-up at the same time? Is anyone using the gateways or prefer direct backups?

Thank you.
Tom

[PetrM]

Hi Tom,

We have the following results based on our testing:
1. The optimal number of channels is 4 for full backups and 1 for log backups.
2. We tested 10 policies with 15 hosts/clusters per policy. We support 150 servers per VBR server (we plan to increase our scalability limits in future releases), 10 000 databases per VBR installation, and 2 000 databases per server.
3. The number of gateway servers depends on CPU and RAM, as these directly impact the number of parallel tasks that can be run. See more information on this page.
4. Please note that there is no physical direct mode for plug-ins. A gateway server always exists: either the gateway server specified in the repository settings or the Mount server if direct mode is selected in the repository settings.

Thanks!

[PospiKM]

Hi Petr,

1. Will test it and check the performance.
2. OK, our setup will have 30-40 policies, including +-200 SQL servers with xxxx databases so I believe that it should be fit the limits.
3. Is it better to have less GW servers but powerfull in terms of performace or more GW servers with lower performance tier? This is something interesting regarding cost scenarios in Azure. Is there any guideline/blueprints for optimal design based on the expected workload/resources (vs number of GW servers let's say)?
4. Noted.

Thank you.

[PetrM]

Hello,

3. I think it’s up to you and what’s easier to manage. I’d recommend sticking with a few powerful gateway servers. Cost scenarios in Azure do not depend on the number of gateway servers, the main factors affecting cost are the volume of data stored and the frequency of requests to Object Storage (put/get/list, etc.), which in turn depends on the full, differential, and transaction log backup schedules.

Could you please tell me more about the database backup schedule and the amount of data? I’ll try to provide rough statistics on API requests so you can make some rough cost estimations.

Thanks!