Gateways & Scheduled Refresh
Understand when a data gateway is required, how to configure scheduled refresh, and keep your Power BI datasets up to date automatically.
Keeping data fresh
Imagine a delivery service. Your Power BI report lives in the cloud, but your data might live on a server in your office building. A gateway is the delivery truck that carries data from your on-premises server to the cloud.
Without a gateway, cloud-based Power BI can’t reach on-premises data. Scheduled refresh tells Power BI when to send the truck — daily, hourly, or at specific times.
When is a gateway required?
| Scenario | Gateway Needed? | Why |
|---|---|---|
| On-premises SQL Server | Yes ✓ | The server is behind your firewall — cloud can't reach it directly |
| On-premises file share (Excel, CSV) | Yes ✓ | Files on a local network aren't accessible from the cloud |
| Azure SQL Database | No ✗ | It's already in the cloud — Power BI can connect directly |
| SharePoint Online | No ✗ | Cloud service — directly accessible |
| Dataverse / Dynamics 365 | No ✗ | Cloud services — no gateway needed |
| On-premises Oracle or MySQL | Yes ✓ | On-premises databases need the gateway bridge |
Simple rule: If the data source is behind your organisation’s firewall (on-premises), you need a gateway. If it’s a cloud service, you don’t.
Gateway types
| Type | What It’s For | Managed By |
|---|---|---|
| On-premises data gateway (standard) | Shared gateway for multiple users and datasets | IT admin — installed on a dedicated server |
| On-premises data gateway (personal) | Single user, single machine — development only | Individual user |
Kenji at Apex Manufacturing (🏭) needs data from the factory’s on-premises SQL Server. The IT team installs the standard gateway on a dedicated server. Kenji configures his semantic model to use this gateway for refresh.
Exam tip: gateway installation
- The gateway must be installed on a machine that is always on and has network access to the data source.
- The gateway should be installed on a dedicated, always-on server with network access to the data source — avoid co-locating with the data source when possible for performance reasons.
- The personal gateway is for development/testing only — never for shared production use.
- Gateway supports DirectQuery AND Import mode for on-premises sources.
Scheduled refresh
Scheduled refresh automatically refreshes Import-mode semantic models at configured times.
How to configure:
- In the Power BI service → go to the semantic model settings
- Gateway and cloud connections → select the gateway (if on-premises)
- Data source credentials → enter/update credentials
- Scheduled refresh → set frequency and times
| Setting | Options |
|---|---|
| Frequency | Daily or Weekly |
| Time slots | Up to 8 per day (shared capacity), up to 48 per day (Premium, PPU, or Fabric capacity) |
| Time zone | Set to match your business hours |
| Send failure notifications | Email the dataset owner + additional contacts |
Dr. Ethan at Bayview Medical (🏥) schedules refresh for 6:00 AM daily — so by the time staff arrive, the patient dashboard shows yesterday’s complete data.
Riley at Coastal Fresh (🛒) needs more frequent updates — she schedules refresh every 3 hours using 8 daily refresh slots (Pro maximum).
DirectQuery and refresh
DirectQuery models don’t need scheduled refresh for data — they query the source live. However, some metadata (like security role changes) may still benefit from a refresh cycle. A gateway is still required if the source is on-premises.
Composite models (mix of Import and DirectQuery) need scheduled refresh for the Import tables only.
DirectLake models (Fabric) don’t need refresh either — they read current files from OneLake directly.
Knowledge check
Kenji's Power BI report uses Import mode connected to an on-premises SQL Server. The report data in the Power BI service is stale. What TWO things does he need?
Dr. Ethan's hospital has data in both Azure SQL Database (cloud) and an on-premises Oracle database. Which data source requires a gateway?
Next up: Row-Level Security and Access Control — control who sees what data.