Report loads quickly on desktop but slow when published online

[Anonymous]

I had a very slow report. I deleted bunch of columns, rows, did other optimisations and on the PBI Desktop the visuals update much faster now. Like 5-8 times faster.

However, when I publish this report and try to look at it online it still loads at basically the same speed as the old, bulky report so I am thinking the issue must be elsewhere, not in poor optimisation?

All the data for report is saved on a Sharepoint and imported directly from there, no DirectQuery. I work for a corporation so all the gateways, etc. are not managed by me.

Also, we are required to use VPN as well, not sure if that could have an impact.

Would really appreciate any help on figuring out the bottleneck. Let me know if I should provide other info as well.

[vanessafvg]

the capacity of your laptop / computer is potentially different to what you have in the power bi service.

what capacity are you on?  Normal power bi pro?  or premium and if so what capacity?    Network latency can also be an issue.  Key thing is to understand that it is dependent on what accesss to compute you have in the cloud.

Also there might be more optmisations you need to do depending on that compute.

[Anonymous]

Thank you.

Just normal power bi pro, yes.

Also, the report is not always slow. If I try to load it during evenings or late in the afternoon, it will be near instant. But early in the morning, when I assume a lot of people are using the services, it can take as much as 20s or more to load anything.

So I wonder if there is anything that can be done to make report run smoother during those times or if it's out of my control.

The report itself is not massive too, 40 MB.

[vanessafvg · Accepted solution]

based on what you are saying i think it could be a capacity issue.  If you say the morning is busy and thats when its problematic then I would say that is very likely connected.   It might be worth optimising your model more if you aren't able to go to premium.  Also doing some performance analysis.  Some issues might happen due to code or modelling that uses a lot of compute, if you have multiple people running it at the same time, its a becomes apparent that is is an issue, however  if a person is running it on their own,  you might not notice because it does not exceeed the total processing available.

40mb in itself might not be the issue as there is storage but there is also the processing of compute to consider.

[Anonymous]

Thank you, I will look into the above options.