Last month at the Microsoft Business Applications Summit (MBAS), Power BI program managers David Magar and Swati Gupta showed off a new load assessment tool for Power BI Premium capacity workloads.
This new tool was included as part of the BRK2046 session on Power BI Premium at MBAS. The whole session is valuable, but the tool itself comes in around the the 32 minute mark. There’s a demo at the 37 minute mark. The tool is available today on github.
This tool will help Power BI Premium customers better plan for how their specific workloads (reports, dashboards, datasets, dataflows, and patterns of access) will perform on a given Premium capacity.
The tool is built on top of a PBIE load generation tool my teammate Sergei Gundorov has built to help ISVs better handle load on their PBIE capacities. The tool grabs a user’s token and uses it to render reports again and again, cycling through preset filter values and incrementing a “render counter”. The tool stops rendering when the authentication token expires, so the result is an empirical benchmark: “report X can run against capacity Y, Z times in 1 hour”.
The tool that’s available publicly used Sergei’s work as the starting point and uses PowerShell to turn it into a simple “menu-based” UX that anybody can run. The tool enables users to:
- choose multiple reports to run at once
- choose the credentials used for each report
- define filter values to cycle through between renders for each report
- define how many users (browser windows) should request the report at once
Once all definitions are set the tool launches multiple browser windows, each targeting different reports and the users can see the load test happening on screen.
The tool was an effective way for David and Swati to generate “interesting” load scenarios for their MBAS workshop. They used it demonstrate how phenomena related to overloaded capacities (such as query wait time build up and frequent evictions) are visible using the Power BI Premium metrics app. If you haven’t already watched the session recording, be sure to check it out.
The folks at Artis Consulting have already taken the tool that Sergei developed and which was shown at MBAS and have released a “Realistic” load test tool, also on GitHub. This tool build on the original one, and makes it easier to simulate users interacting with reports in a more realistic manner, such as selecting filters and bookmarks.
If you’re interested in understanding how your Power BI application will scale under load on your dedicated capacity, check out these tools and consider how to incorporate them into your development and deployment processes.
 Yes, MBAS took place in June, and this is getting posted in October. I originally wrote this post in early July, and I put it on hold to wait for the official blog post to be done so I could include a link. It’s been languishing in my drafts ever since. Life comes at you fast…
 It’s worth emphasizing that this tool and this post apply only to dedicated Power BI capacity. If you are using shared capacity, you should not use this tool.