A premium capacity node can be thought of as a fully managed server in the Azure cloud which runs the Power BI service. The capacity node is dedicated and isolated to the organization that provisioned the capacity and the same user experience and functionality is delivered as the shared (free) capacity provided by the Power BI service. Each capacity node has a set of processing and memory resources (v-cores and RAM), bandwidth limits, and a cost that aligns with these resources. For example, a P1 capacity node includes 8 v-cores and 25 GB of RAM at a cost of $5,000 per month, while a P2 capacity includes 16 v-cores and 50 GB of RAM at a cost of $10,000 per month. When app workspaces containing Power BI content (datasets, reports, and dashboards) are assigned to premium capacity nodes, the resources of the given capacity node are used to execute Power BI activities associated with this content, such as query processing and data refresh operations.
Chapter 12, Administering Power BI for an Organization, referred to the v-cores (virtual processing cores) of Premium capacity nodes but didn't provide details on other resources (RAM and bandwidth) and their relationship to Power BI workloads. For example, if all Power BI reports will utilize a DirectQuery dataset or a Live connection to an Analysis Services model, then the amount of RAM provided per capacity will be much less important than the limits on the number of connections and the max page renders at peak times. In these deployments, the resources provisioned for the data source system (CPU cores, clock speed, and RAM), as well as the latency and bandwidth of the connection between the source system and the data center region of the Power BI tenant, would largely drive query performance.
The following table identifies the resources associated the six EM and P Premium capacity nodes currently available:
As shown in this table, as of February 2018 the largest premium capacity node includes 32 v-cores, 100 GB of RAM, and supports a max of 120 DirectQuery or Live connection queries per second. Larger capacity nodes, such as a P4 with 64 v-cores and 200 GB of RAM, will likely be released later in 2018 and will complement a scale out (multi-node) capacity as identified in the Power BI Premium whitepaper for October of 2017.
Microsoft Azure resources such as Azure Analysis Services or Azure SQL Database, which can be created within the same region as the Power BI service tenant, provide their own user interface and tools for scaling up and down as the needs of workloads dictate. Guidance on identifying the location of your Power BI tenant, and thus the preferred location for Power BI data sources, is included in the Top gateway planning tasks section of Chapter 9, Managing the On-Premises Data Gateway. The minimal distance between a Power BI tenant and an Azure data source in the same data center region provides a natural performance advantage over connections to on-premises sources via the On-premises data gateway.