Power BI Global Scale out Deployment

If you are a global organization, you can take advantage of Azure to provide globally scaleable, high performance BI solution without purchasing tons of hardware which is going to sit ideal during off hours.

We have two major goals for the deployment.

1. Provide Best Performance to each user based on global location.

2. Provide always on availability.

You can deploy Power BI Report server to closest data center to each of your facility and then use ATM (Azure Traffic Manager) to route traffic based on IP Location, to use Azure Traffic Manager you will have to setup report manager site for external access. You can also have copy of report server VM as passive standby in another azure region in same country for business continuity scenario if there is data center wide outage. You will have to deploy same setup for each country you do have a location, you can have your source databases / Datawarehouse replicated using geo replication or Always on Cluster to same azure regions in each country.

Here is how it will work from user standpoint.

User will open report site URL configured in ATM, ATM will route that URL to report manager URL local to the user based on his IP Address. Report manager will run report against data source in the same data center, so performance will be best. Let’s say local Active azure data region has outage, ATM will detect it automatically and initiate failover to passive data region in same country, while passive VM is being spun and made ready users will be routed to Active data region in another country. There will be small performance impact but only for time it takes to activate passive and failover.

Usually you have very little or no traffic after business hours, so you can follow the sun model to auto shut off Active VM’s after business hours and save money, users will be routed automatically to report manager in different country.

CONSIDERATIONS:

1. Report Deployment: This architecture makes your users very happy, but your BI developers will have to do some extra work, they will have to deploy each report / Project to multiple report servers, they will have to use separate connection string for each data source, so report runs against replica local to Azure Data Region. This can be simplified using configurations in Visual Studio.

2. External Facing SSRS Site: ATM only works with external facing sites, so you will have to make report manager external facing or use third party load balancer to route traffic.

3. Site to Site VPN: You will have to setup site to site VPN from each of your facilities to azure data centers you going to use.

Same architecture can be followed for POWER BI online, you can setup tenant in each country or purchase Power BI Premium capacity in each Azure Region.

Note: if you are a very large organization you may consider doing scale out deployment for each report server and data source within same data center as well to further distribute traffic. Power BI Report server currently does not support scale out deployment, so you will have to use SSRS 2016 or wait for local scale out.