Performance to expect

[hansmbakker]

When using the SQL Server docker image it is useful to be able to persist the data, but it seems that performance when persisting to File Storage is an issue then.

It would be good for this documentation page to mention some usecases or performance expectations on this page.

A workaround could be to put the /var/opt/mssql folder in the emptyDir storage and make e.g. an hourly backup of it to file storage. Another workaround could be to recommend AKS for more demanding workloads.

---

Document Details

⚠ Do not edit this section. It is required for docs.microsoft.com ➟ GitHub issue linking.

• ID: b730fc95-951e-1acd-770d-88b3e6f2f6ef
• Version Independent ID: b6445fd2-46ea-aedd-2f2c-294faa660660
• Content: Mount Azure Files volume to container group - Azure Container Instances
• Content Source: articles/container-instances/container-instances-volume-azure-files.md
• Service: container-instances
• GitHub Login: @dlepow
• Microsoft Alias: danlep

[Karishma-Tiwari-MSFT]

Thanks for the feedback! I have assigned the issue to the content author to investigate further and update the document as appropriate.

[dlepow]

@hansmbakker - We don't currently document performance expectations for Azure file shares with container instances specifically. Azure docs include a number of related topics about file share performance, which should help for deployments with many services. Examples:

• Azure Files performance and scalability targets
• Planning for an Azure Files deployment
• Troubleshoot Azure Files performance issues - such as latency due to file share in a different region.

Do these help for your scenario? As you note, there are other volume mounting options with container instances, useful for different scenarios. If you use an emptyDir volume, the data does not persist there after the container instances stop.

[hansmbakker]

My scenario was that we tried running sql server in a docker container in ACI. We tried various locations for the /var/opt/mssql/ mount and got different results, but bottom line none of the options with ACI gave normal container startup time (and the query performance was not good either). The sql server docker container has a normal startup time on developer laptops.

In both cases we used an unmodified sql server docker from Microsoft.

The ACI instance cpu / ram were not the bottleneck (not all resources were used). Therefore I assumed that disk IO was the bottleneck. After searching online I saw people request ACI support for managed disks for performance reasons.

The reason we tried ACI were these requirements:

• we needed a full SQL server for compatibility reasons (SIMPLE RECOVERY support for bulk import performance - neither Azure SQL databases nor SQL Managed Instances support that
• more lightweight and easier to integrate with a devops setup than SQL VM)
• less complexity & learning curve for a single container than kubernetes
• compatibility with SQL (Synapse has lots of SQL schema limitations)

If it would have been documented that ACI (disk) performance cannot be considered comparable to even a developer laptop, it would have saved us time trying to tune ACI to run sql server.

[dlepow]

@hansmbakker - Thanks for providing the details. I am sharing with engineering partners.

[hansmbakker]

Thank you!

[dlepow]

assign @dkkapur

[dlepow]

Reassigning doc ownership

reassign: @macolso

[cynthn]

Thanks for your dedication to our documentation. Unfortunately, at this time we have been unable to review your issue in a timely manner and we sincerely apologize for the delayed response. The requested updates have not been made since the creation of this issue, and the timeline for resolution may vary based on resourcing, so we've created an internal work item to incorporate your suggestions. We are closing this issue for now, but feel free to comment here as necessary. #please-close