SSAS Multidimensional & Tabular on Same Server

[blynchdata]

Hello,
We are planning a migration of our SSRS server to an on-premise Power BI Report Server with the longer term plan to migrate everything to Azure & Power BI Service. 

There is guidance out there to not install SSAS Multidimensional & Tabular on the same server for performance reasons as they allocate and use the hardware differently, etc. 

My question is: What about for lighter workloads?

We have a low user count. Over the past 2 years our SSRS server has served:

40 unique users per month at peak. 

On average 7 users per day with a peak of 21 on one of those days. 

Our data volume is quite low:

Multidimensional cubes = 2GB. 
Tabular Model = 500 MB. 
Most of the rest of the reporting is via T-SQL procedures.

If this is seen as a lighter workload scenario, could it be appropriate to install two instances of SSAS, one in each mode to buy time to then plan to move to the cloud?

Or do you think it would make all performance horrendously slow, including any SSIS jobs?

Thank you for any insights you may have.

Brad

[josef78 · Accepted solution]

I'm not sure in azure, but in on-premise deployment is not problem with two different instances (OLAP and Tabular) on same server. We are using onpremise SSAS server with three instances (OLAP, Tabular and second tabular for workspace) on same server, and all three instances are used from PBIRS farm without problems.

From performance perspective, yes OLAP and tabular use HW resources very differently (OLAP is I/O intensive, Tabular is CPU and Memory intesive), but there is not big difference if are CPU and Memory resources shared in two instances on single server, or on two virtual servers on single HW. In heavy load can be trouble in I/O workloads, because are used very differently and I/O request can be queued, but can be sloved by separated LUNs. In lighter load, I think there no performance problem.

[blynchdata]

@josef78, 

Thank you so much for this. Having more than one SSAS instance on the same server would be such a critical stop-gap for us to give us some time to plan our move from on-premise to most likely Azure; while getting our teams used to Power BI. 

Thank you!