We’re excited to announce that we’ve reached the final milestone in our dataset scale-out public preview journey! We started the preview without auto-sync and with single read-only replica per dataset. A few months ago, we introduced auto-sync, and now Power BI can create as many read-only replicas as your Power BI capacity supports. Dataset scale-out is no longer limited to a single read-only replica per dataset.
Dataset scale-out is a feature that enables you to create read-only replicas of your datasets and distribute the query load among them automatically. A read-only replica is a dataset copy that can serve queries but cannot be updated by refresh operations. By having multiple read-only replicas, you can reduce the query latency and increase the throughput of your dataset. This is especially useful for scenarios where you have a high number of concurrent users or complex queries that consume a lot of resources.
To use dataset scale-out, you need to have a Fabric, Power BI Premium, or Power BI Embedded capacity that supports this feature. On capacities smaller than a P3, datasets can only have a single read-only replica because the compute power of these capacities typically doesn’t warrant higher replica counts. For P3 and above, datasets can scale to two or more read-only replicas depending on client load and processor utilization.
To enable scale-out for an individual dataset, you must set the maxReadOnlyReplicas parameter in the queryScaleOutSettings to a non-zero value, as demonstrated in our previous announcement of automatic replica synchronization for dataset scale-out. We recommend setting the maxReadOnlyReplicas parameter to a value of -1, so that Power BI can manage the read-only replica count automatically for you. You can verify the current number of read-only replicas at any time by using the /queryScaleOut/syncStatus API, as shown in the following screenshot. This API returns the sync status of each read-only replica:
We’re super excited that dataset scale-out is now feature complete because it can provide many benefits to your Power BI experience, such as:
- Improved performance - By having multiple dataset replicas, you can reduce the query latency and increase the throughput of your datasets. This means faster and smoother reports and dashboards for your users.
- Reduced refresh impact - By having read-only replicas separate from the read-write replica, you can reduce the impact of refresh operations on your reports and dashboards. When you refresh your dataset, Power BI only updates the read-write replica while the read-only replicas continue to serve queries. The subsequent sync operation is fast and transparent with less disruption and more consistency for your users.