Over the past several months, we have received positive feedback about the Azure Analysis Services preview. Many customers have moved their models to the cloud and have enjoyed the improved manageability of the platform-as-a-service offering. We have expanded regions where Azure Analysis Services is available, and we are working on several of the improvements asked for on the Azure Analysis Services feedback site.
One scenario customers have asked for is a way to support smaller workloads in the cloud. While you can run multiple databases on a single Standard S1 instance, you may want to start with something smaller based on your model size and query volume. We are introducing a new smaller pricing level, Standard S0, which has 40 QPUs and 10 GB of RAM for models.
This new size offers the same features and capabilities as the rest of the Standard tier. You can continue to scale up and down based on your expected load to achieve the best experience for your users. For example, if you need more RAM when processing data, you can scale up during processing and scale down after processing. You can also scale up during business hours for better query performance, scale back down in off-hours, or even pause when needed for further cost savings. You can track QPU utilization in the Azure portal and through the Azure Monitoring APIs. Note, your model must fit in the available RAM. It is a good idea to check your memory consumption the Azure Portal to ensure Standard S0 is appropriate for your workload.
Please try out the Standard S0 size and let us know how it works for you. You can share your experiences on the Azure Analysis Services MSDN forum. The forum is also a great place to get help from the community.