Skip to main content
Microsoft Idea

Power BI

Needs Votes

Power BI - PaaS for scalability

Vote (7) Share
Shrikant's profile image

Shrikant on 28 Jan 2016 07:22:24

Move Power BI from SaaS to Platform as Service to scale users & performance.

Comments (1)
Shrikant's profile image Profile Picture

Lauren Dittmann on 03 May 2024 05:44:52

RE: Power BI - PaaS for scalability

Agreed. Unless MS can figure the scalability thing out first for us :) Although current roadmap developments (query scale-out and auto-scale) do help with scaling up for higher concurrent user rates, scaling up for faster interactive query performance remains out of reach - leaving smaller teams unable to 'buy some time' to optimize reports while still being able to deliver reports in a timely and non-aggravating manner to end users. Although there is the inherent limitation from the Vertipaq engine presented by the FE engine (which can only use 1 stream and 1 core at a time and therefore cannot be sped up), it is disheartening that even a high-SE engine-based query is not 'scalable' when additional capacity is purchased (meaning it will not speed up its performance by maximizing parallelization or automated temp SE aggregations or whatever other magic could be worked). If PaaS would offer us the ability to develop this functionality in-house, then at least our extensive optimization efforts will be seen to save time AND money :)Also - I am open to any other options, tools, work arounds or viewpoints that might help us 'scale up' slow queries, understanding that optimization is already being worked on heavily and continuously, and the hiring pool for folks who can reliably optimize DAX (and return valid results) is not very large. The goal here really is to enable a system where, when a query isn't fast, we can just throw some money at it until next month.If the VertiPAQ engine truly cannot be speed up by increasing resources, then a report which highlights which portions of current capacity usage are not scalable down to the per-query level and relevant statistics on how often the worst-performing queries are run and their total capacity impact would be much appreciated - the current Capacity usage metrics report does not get us there (or even in the neighborhood). Thanks for reviewing!