Recently we faced an issue with AAS being down for around 4 hours which impacted the production server and many users were unable to do their work.

Our solution to mitigate was create a new AAS instance in another region and restore the backup.

Just like Azure SQL, it would be really useful to have failover solution to AAS, so in cases like that we could just automatically start the offline/secondary instance with lower down time.

Are additional configurations planned different ratios QPUS/RAM? We are often times hitting the memory limit but are no way near the QPUS limits. Also what are the plans if our model will not fit an S9 instances?
We would be looking for an S4 with 800GB ram or S5/6 if available with 1,5TB

If you decide to change your M query for a table, with the typical case being that you needed to add an additional column, you currently have to edit the model.bim file to add the new column into your actual table object.

This new feature would allow us to see what columns get surfaced by the M query and select which ones we want imported into the model (which would typically be all of them, because otherwise you would have excluded the column in your m query).

As a best practice, we'd typically use Windows auth for on-prem SSAS where the SSAS service account is granted access to the data source to process the model. Allow us to change the Azure AS service account to be used when processing from SQL Server.

Business users and report-writers that interface with the tabular model often have many questions about what the relationships between tables are, what the specific measure definitions are, or just want to know general description information about the model.

While I can create a generic "metadata" model myself that connects to a targeted AS database's DMVs, it'd be nice if Microsoft created a standard template as a model.bim file and published it so that we can very easily edit the datasource to select our targeted model.

We should be able to set the KPI status thresholds using table fields or measures. In my business KPI thresholds change with seasons and item categories. So by being able to set these thresholds using table fields or measures, the thresholds can be set without having to edit the datamodel.

It's not always easy to re-produce a crash in SSDT, and at times it can feel like the crash was due to typing out a measure too quickly, coming up with wacky m queries, or perhaps even because it is too windy outside. An improved error log would at least give me some solace that SSDT isn't prejudicing my model development due to my taste in music.

I would like to see a feature in Azure analysis service where when if a server or cube is getting processed the report connection can be redirected to another server. This functionality can be done in on premise cubes with help of load balancer and shell scripts. Anything of this sort on Azure can be great full.

Similar to overdraft protection on a bank account... would be nice to have a similar concept for Azure AS instances where the instance auto-scales up if memory consumption exceeds the current limit (instead of an error - for example during processing)

Currently, we are using the on-premise data gateway to connect Azure Analysis services to Snowflake.
Instead, it would be great to have a direct query connection established between Azure Analysis services and Snowflake.

Update Azure Analysis services so it auto-pauses after 1 hour of inactivity and then auto-restarts when the first person tries to access it similar to the functionality for Azure SQL Server.

It's available for PBI, would be nice to have it on Azure Analysis Services as well so we can build and manage our models in a single place. Doesn't make sense to pull data from one in-memory solution to another.

Analysis service can not pull data from azure storage account if the VNET Integration is enabled on the storage account in the same region.

Enable VNET Integration like App Service VNET Integration preview feature as one example.

Currently users have to have their IP address whitelisted in order to promote or manage the model from Visual Studio. This can be difficult with remote employees as IP addresses can change and it is difficult to manually input a large amount. Ideally it would be nice if a new way could be created that is easier to manage large number of remote users. Maybe a security token of some sort?

To me, the BIGGEST drawback of SSAS-TAB/AAS is the lack of security around columns and measures. In many cases, we need to give access to certain users to only part of the model. I.e: some users can see only sales, while others can see net margins (including costs).

Today, the only way to do this is to duplicate the model and drop columns or measure, which adds overhead in time and cost.

Maybe, if users could be locked to a perspective (and prevented from using freeform DAX), this could be accomplished ? Or something like that.