Understanding the Power BI Embedded Capacity Workspace
Power BI Embedded capacity and workspace management is the foundation of any successful embedded analytics deployment. Whether you are an ISV embedding Power BI reports in a SaaS application or an enterprise distributing analytics to external stakeholders, understanding how capacities, workspaces, and SKUs interact is critical for performance, cost management, and scalability. EPC Group has architected Power BI Embedded solutions for software companies and enterprises serving thousands of concurrent users across multi-tenant environments.
What Is Power BI Embedded Capacity?
A Power BI Embedded capacity is a dedicated pool of compute resources in Azure that hosts your Power BI content. Unlike shared (Pro) capacity where resources are shared across all tenants, Embedded capacity gives you guaranteed, isolated compute for rendering reports, running queries, and refreshing datasets.
- A-SKUs (Azure): Pay-per-second Azure resources (A1-A8) designed for ISVs and application embedding. Can be paused to save costs.
- EM-SKUs (Microsoft 365): Purchased through Microsoft 365 admin center (EM1-EM3). Cannot be paused but include Power BI Pro licensing benefits.
- P-SKUs (Premium): Full Power BI Premium capacities (P1-P5) that support both internal users and embedded scenarios. Include all Premium features.
- F-SKUs (Fabric): Microsoft Fabric capacity units that include Power BI capacity as part of the broader Fabric platform.
- Each SKU tier provides a specific number of v-cores, memory, and maximum model size. Higher tiers support more concurrent users and larger datasets.
How Workspaces Relate to Capacities
Workspaces in Power BI are containers for reports, dashboards, and datasets. Assigning a workspace to a capacity determines the compute resources available for that workspace's content.
- A workspace assigned to shared capacity uses pooled resources and requires Power BI Pro licenses for all viewers
- A workspace assigned to Premium or Embedded capacity uses dedicated resources and can serve free users (no Per-User license required for viewers)
- Multiple workspaces can share a single capacity, with resources dynamically allocated based on demand
- Capacity admins can set per-workspace resource limits to prevent a single workspace from consuming all capacity resources
- Workspaces can be moved between capacities with zero downtime, enabling capacity balancing
Capacity Sizing: Choosing the Right SKU
Selecting the right capacity SKU requires understanding your workload characteristics: number of concurrent users, dataset sizes, refresh frequency, and query complexity.
- A1 (1 v-core, 3 GB RAM): Development and testing only. Supports up to 1 GB datasets and approximately 10-20 concurrent users.
- A2 (2 v-cores, 5 GB RAM): Small production workloads. Up to 3 GB datasets and 25-50 concurrent users.
- A4 (8 v-cores, 25 GB RAM): Medium production workloads. Up to 12.5 GB datasets and 100-200 concurrent users.
- A6 (32 v-cores, 100 GB RAM): Large production workloads. Up to 50 GB datasets and 500-1000+ concurrent users.
- Always start with capacity that exceeds your estimated requirement by 20-30% for headroom during peak loads
- Use Azure autoscale rules to scale A-SKUs up during business hours and down during off-hours
Performance Optimization for Embedded Workloads
Capacity performance directly impacts end-user experience. Slow report rendering and timeout errors are often caused by capacity configuration issues, not report design issues.
- Monitor capacity utilization using the Power BI Premium Capacity Metrics app (available in AppSource)
- Track CPU overload events: sustained CPU above 80% indicates the capacity is undersized
- Schedule dataset refreshes during off-peak hours to avoid competing with interactive queries
- Implement query caching (enabled by default on Premium) to reduce redundant computations
- Use aggregation tables to pre-compute common query patterns and reduce query processing time
- Separate heavy-refresh workloads (large datasets) from interactive workloads (user-facing reports) on different capacities
Multi-Tenant Workspace Architecture
For ISVs and SaaS applications, designing the workspace architecture for multi-tenancy is a critical architectural decision that impacts security, performance, and cost.
- Workspace-per-tenant: Each customer gets their own workspace. Maximum isolation, easiest RBAC, but highest management overhead.
- Shared workspace with RLS: All tenants share a workspace and dataset. Row-level security filters data per tenant. Lower management overhead but requires careful RLS design.
- Hybrid approach: Large customers get dedicated workspaces; smaller customers share workspaces with RLS. Balances isolation with cost efficiency.
- For the shared model, use service principal profiles to manage embed tokens per tenant
- Implement tenant onboarding automation using the Power BI REST APIs to create workspaces, deploy reports, and configure RLS dynamically
Why EPC Group for Power BI Embedded Architecture
EPC Group has designed and deployed Power BI Embedded solutions for ISVs and enterprises serving thousands of external users. Our architects specialize in capacity sizing, multi-tenant security, and performance optimization for production-grade embedded analytics.
- End-to-end Embedded architecture design from POC through production
- Capacity sizing and cost optimization with autoscale configuration
- Multi-tenant workspace design with row-level security implementation
- Service principal and embed token management automation
- Performance monitoring and capacity health management
Need Help with Power BI Embedded Capacity Planning?
Contact EPC Group for a capacity assessment and architecture review for your Embedded deployment.
Frequently Asked Questions
What is the difference between A-SKU and P-SKU capacity?
A-SKUs are Azure resources purchased through the Azure portal with pay-per-second billing and the ability to pause (stop billing). They are designed for ISV embedding scenarios. P-SKUs are Power BI Premium capacities purchased through Microsoft 365 with monthly billing. P-SKUs include additional Premium features like deployment pipelines, XMLA endpoints, and Paginated reports. For pure embedding, A-SKUs are more cost-effective. For combined internal and external use, P-SKUs or F-SKUs (Fabric) provide better value.
Can I autoscale Power BI Embedded capacity?
A-SKUs support Azure autoscale rules. You can configure scale-up triggers based on CPU utilization (e.g., scale from A2 to A4 when CPU exceeds 70% for 10 minutes) and scale-down during off-peak hours. P-SKUs and F-SKUs support autoscale through the Power BI Admin portal with configurable maximum v-core limits. EPC Group configures autoscale with cost guardrails to prevent unexpected billing spikes.
Do embedded report viewers need Power BI licenses?
No. When using the "App Owns Data" (service principal) embedding pattern, end users do not need any Power BI license. Authentication and authorization are handled by your application. This is the standard pattern for ISV and customer-facing scenarios. The "User Owns Data" pattern requires each viewer to have a Power BI Pro or PPU license, which is typically used for internal embedding.
How do I monitor capacity health in production?
Install the Power BI Premium Capacity Metrics app from AppSource. It provides detailed dashboards showing CPU utilization, memory consumption, query durations, refresh statistics, and overload events. Set up Power Automate alerts for sustained high CPU, failed refreshes, and eviction events. For A-SKUs, also monitor Azure resource health and set up Azure Monitor alerts.
What happens when capacity is overloaded?
When a capacity is overloaded, Power BI begins throttling operations. Interactive queries slow down, and scheduled refreshes may be delayed or fail. In severe overload, the system may reject requests entirely (HTTP 429 errors). Dataset evictions occur when memory is exhausted, forcing datasets to be reloaded on next access. EPC Group designs capacity architectures with headroom and autoscale rules to prevent overload scenarios in production.