09-19-2016 05:43 AM
Rui Romano has a great tool over at https://ruiromanoblog.wordpress.com/2016/01/08/power-bi-desktop-trace-logs-analyser/ to anlyzing your Power BI Audit logs to try and figure out why your refreshes are slow. There is only one problem. I can't seem to find any documentation for translating the actions into english, so I can understand what the problem is and how I can work to resolve the problem.
Does anyone know where I could find translations for RemoteDocumentEvaluator, SimpleDocumentEvaluator, and RemotePageReader along with their child actions? I've got a small excel file (<500 rows) being used to filter a small (<2M row) Tabluar Model for a proof of concept. When I change to a SQL table, I get similar results as Excel. It takes about 3.5 minutes to read the data.
Only 15 - 20 seconds of that time is reading the Tabular model. When reading a SQL table, that SQL read takes 0.0 seconds. When reading the excel file, that too takes 0.0 seconds.
The major waits are:
- RemoteDocumentEvaluator/Service/OnBeginGetResult, GetResult
09-27-2016 05:03 AM
If the senior engineers have any documents on the audit logs, that would be a great help here. I'm really looking for a deeper understanding of Power BI Internals and the mashup workers. If I could understand these action descriptions, I think that would answer why the performance is so bad in this one case.
I have noticed that it appears that for each data pull the mashup appears to process that "file" of data three to five times. That seems odd to me. It could be various steps in compressing that data to package it up into the resulting report file.
Thanks for anything you can share!
09-28-2016 06:55 PM
@shannonlowder, I'm still in the progress to involve other engineers. You can also submit one document request on the Ideas forum here: https://ideas.powerbi.com/forums/265200-power-bi-ideas
09-29-2016 07:09 PM
@shannonlowder, I got updates from engineer from PG team:
"Those traces files are collected on the customer machine and should be provided to the engineering team to help troubleshoot issues. The engineering team does not expect customers to look at this data and do not plan to educate customer on how they can do this".