Unable to review CPU as it occurs from background workloads

[dbeavon3]

At some point Microsoft started graphing "CU" in the capacity dashboard instead of CPU.  Their goal is to promote this concept of "carry forward" and "smoothing".  Customers are regularly borrowing from future capacity, and the Power BI service is slowly becoming a payday lender, in addition to a reporting service.    The claim is that it is easier to manage, but that is not my experience so far.

The CU changes have made this environment very troublesome to manage.  If we want to actually find *actionable* ways to fix specific problems, we need to be able to see more detail.  Eg. we need the CPU impact when it is *incurred*, rather then the smoothed version.  Instead, of managing CPU as it is incurred, Microsoft appears to be reporting a version of the CPU wherein smoothing-accruals are used and the resulting CPU penalties are spread over a long period of time.

Consider the CPU of background operations, based on the following chart:

This chart is almost useless where background CPU is concerned.  As you can see, our background consumption is growing out of control (around 30% of total available CPU for the day).  But the impact can't be easily managed.  We can't isolate whether the background is happening off-hours or not.  We can't see where the CPU is incurred.  All we have to look at is a big blue bar at the bottom of our utilization graph, and it rises on a daily basis.

We can drill through to timepoint detail, but that loses our ability to visualize usage-over-time.  The background operations in that screen are reported in a disorganized way.  It doesn't clearly indicate what periods are incurring the most CPU for background operations, so that it can be minimized.

My question is how do we visualize the background CPU in a way that tells us when the CPU cost are *incurred*.  That would be far more actionalbe than visualizing the "smoothed" version of this data. 

https://learn.microsoft.com/en-us/power-bi/enterprise/service-premium-smoothing

I find it really bothersome that Microsoft pushed these carry-forward and smoothing concepts as far as they did.  It is not likely that this was done based on customer requirements.  It seems these changes make it harder than ever for customers to try and manage their CPU usage and, by extension, costs.

[johnbasha33]

@dbeavon3 

It sounds like you're facing challenges in effectively managing and visualizing background CPU usage in the Power BI service, especially with the introduction of concepts like carry-forward and smoothing. Let's explore some potential approaches to visualize background CPU usage in a more actionable way:

1. **Detailed Time Series Analysis**: Instead of relying solely on aggregated daily or hourly data, consider analyzing CPU usage at a more granular level, such as per-minute or per-second intervals. This can provide insights into when CPU spikes occur and help identify patterns in background CPU usage.

2. **Custom Visualizations**: Explore the possibility of creating custom visualizations or dashboards tailored specifically to monitor background CPU usage. You can leverage Power BI's extensibility features to create visualizations that suit your specific requirements and provide more detailed insights into CPU usage patterns.

3. **Drill-Down Capabilities**: Enhance drill-down capabilities in your reports to enable users to investigate CPU usage at different levels of detail. For example, users should be able to drill down from daily or hourly summaries to per-minute or per-second breakdowns to identify specific periods of high CPU usage.

4. **Comparative Analysis**: Compare background CPU usage across different time periods, such as weekdays vs. weekends or business hours vs. non-business hours, to identify trends and anomalies. This can help in understanding when CPU costs are being incurred and optimizing resource allocation accordingly.

5. **Alerting Mechanisms**: Implement alerting mechanisms to notify administrators or stakeholders when background CPU usage exceeds predefined thresholds or exhibits unusual behavior. This proactive approach can help in identifying and addressing issues in a timely manner.

6. **Collaborative Analysis**: Foster collaboration among stakeholders, including IT administrators, data analysts, and business users, to collectively analyze and interpret background CPU usage data. This multidisciplinary approach can lead to deeper insights and more effective decision-making.

7. **Feedback and Advocacy**: Provide feedback to Microsoft about your experiences and challenges with the new concepts introduced in the Power BI service. Participate in relevant forums, user groups, or feedback channels to advocate for improvements that better align with customer requirements and use cases.

By adopting these strategies, you can enhance your ability to visualize and manage background CPU usage in the Power BI service, ultimately leading to more informed decision-making and optimized resource utilization.


Did I answer your question? Mark my post as a solution! Appreciate your Kudos !!

[dbeavon3]

Please don't use AI bots to reply to the community posts. 

I'm looking for a specific solution that would allow me to review the background CPU as it is incurred.  IE. the way to achieve the goal you listed as #1 : Detailed Time Series Analysis. 

You restated my goal , and didn't provide the solution.  Your AI bot is good at restating the question, and meanwhile introducing a great deal of "ambient noise" as well.  This was not what I was looking for, or I would not have posted to the community in the first place.