Planning for the On-premises data gateway involves identifying which data sources require a gateway and understanding the role of the gateway in each deployment scenario. For example, if an import mode Power BI dataset or an import mode Azure Analysis Services model simply needs to be refreshed with on-premises data every night, then gateway resources (hardware) should be provisioned to support this specific nightly workload. This deployment scenario, with the refreshed and in-memory data model hosted in the cloud, is preferable from a user experience or query performance standpoint, as the queries generated in the Power BI service do not have to access the on-premises source via the On-premises data gateway.
Alternatively, when the data model or data source accessed directly by Power BI reports is located in an on-premises environment, the On-premises data gateway is used to facilitate data transfer between the data source and the queries from the Power BI service. For example, a DirectQuery Power BI dataset built against an on-premises Teradata database results in report queries being sent from the Power BI service to the Teradata database via the On-premises data gateway and the results of those queries being returned to the Power BI service via the On-premises data gateway. This deployment scenario can naturally require alternative gateway resources, such as additional CPU cores, given the potentially high volume of queries being generated dynamically based on user activity.
The following two sets of questions address essential, high-level planning topics including the administration of the installed gateway. The following section, Top gateway planning tasks, as well as the Gateway architectures section later in this chapter, contain higher detail to support gateway deployment:
- Where is the data being used by the Power BI dataset?
- Confirm that a gateway is needed to access the data source from Power BI
- This access includes both scheduled data refresh and any DirectQuery or Live connections to the data source
- Additional details on sources requiring a gateway are provided in the next section
- If a gateway is needed, is the data source supported with a generally available (GA) data connector?
- If a source-specific connector is not available, the gateway supports Open Database Connectivity (ODBC) and OLE DB connections as well
- The current list of supported data sources is available at http://bit.ly/2EN1BCg
- Data connectors labeled as (Beta) in the Get Data window of Power BI Desktop should only be used for testing
- Is the on-premises data or the IaaS data being imported to the Power BI dataset(s) or an Azure Analysis Services model?
- If yes, the gateway will support the scheduled refresh/processing activities for these datasets
- If no, the gateway will support user queries of the data source via DirectQuery or Live connections
- Will a standard On-premises data gateway be used or will a personal gateway (personal mode) be used?
- In all corporate BI deployments, the default and recommended on-premises gateway will be installed by the IT organization on IT-owned and maintained servers.
- However, in certain business-led self-service projects or in scenarios in which an IT-owned gateway server is not available, the personal gateway could be installed on a business user's machine, allowing that user to configure scheduled refreshes of import mode datasets.
A single gateway can be used to support multiple datasets, both import and DirectQuery. However, it can be advantageous to isolate the alternative Power BI workloads across distinct gateway clusters, such as with an import gateway cluster and a DirectQuery or Live connection gateway cluster. Without this isolation, the scheduled refresh activities of import mode datasets (Power BI or Azure Analysis Services) could potentially impact the performance of user queries submitted via DirectQuery and Live connection datasets. Additionally, as mentioned earlier, scheduled refresh activities can require far different gateway resources (for example, memory) than the queries generated via DirectQuery datasets or Live connections to on-premises SQL Server Analysis Services (SSAS) databases.
In addition to provisioning hardware and installing the gateway(s) for each scenario, BI teams must also plan for the administration and management of the gateway. Answering the following five questions contributes to planning the implementation:
- Which users will administer the gateway in Power BI?
- This should be more than one user; preferably, a security group of multiple gateway admins can be configured
- These users do not need Power BI Pro licenses if they're only administering gateway clusters
This isolation of hardware resources between corporate and self-service BI (that is, business user/team owned) can also be implemented with Power BI Premium capacities, as described in Chapter 13, Scaling with Premium and Analysis Services. The essential goal of this isolation is to provide the self-service projects with resources aligned to these needs while ensuring that high priority and widely utilized corporate BI assets are not impacted by self-service content or activities.
- Which authentication credentials or method will be used to configure the gateway data sources?
- For SSAS and Azure Analysis Services, this should be a server administrator of the Analysis Services instance
- For certain DirectQuery data sources, a single sign-on (SSO) option is supported in which the Power BI user's identity is passed to the source system, thus leveraging the source system's security.
- The DirectQuery datasets section later in this chapter contains details of this configuration
- Which users will be authorized to use the gateway?
- Users or security groups of users must be mapped to the data source of a gateway
- These are usually report authors with Power BI Pro licenses
- Where will the gateway recovery key be stored?
- This will be necessary for migrating, restoring, or taking over an existing gateway
- Who will be responsible for updating the On-premises data gateway as new versions are released?
- Just like Power BI Desktop, a new version of the On-premises data gateway is made available each month and includes new features and improvements, such as the support for datasets with both cloud and on-premises data sources
- The Power BI team recommends staying up to date with new releases and will not support old gateway versions
- For example, as of March 15, 2018, gateway versions older than the August 2017 release will not be supported
In the following image, two users are added to a security group in Azure Active Directory (AD), dedicated to the administration of the On-premises data gateway:
In this example, mapping the On-Premises Gateway Admins security group to a gateway in Power BI would allow Anna and Brett to configure data sources for the gateway and to authorize users or security groups of users to utilize the gateway. The Managing gateway clusters section later in this chapter includes details on using the gateway portal in the Power BI service.