Your experience running Power BI Desktop (and other BI tools) in a virtual machine?

[otravers]

I've seen a few mentions of this in passing, but I'd love to get more details on how it's been working out for people running PBI Desktop and related BI development tools (Excel, Visual Studio, SSMS, R Studio etc.) in a VM (aka desktop as a service, DaaS, VDI, etc.). Ideally in Azure but I'd be open to Amazon Workspaces, Google, or other providers based on price and functionality. On Azure alone, there's a dizzying array of options (the E series are focused on RAM so look to be a good option for PBI), I know Cirix and VMWare also have DaaS Azure offerings, frankly I'm a bit confused by all the options and not sure how well this works in practice.

What I'm looking to accomplish:

- Use a 64GB RAM Windows 10 PC with an SSD and a decent CPU on demand, and pay only for time used (i.e. PAYG, ideally per second). Low latency is a must-have so that working remotely (presumably via RDP?) doesn't feel like a slog. RAM is the main bottleneck when I work with PBI/PP on my 14GB desktop, I often need to shut down other apps and idly waiting for an ETL process or data refresh to complete is not my idea of productive work.

- Set up everything to my liking in terms of OS preferences, installed apps, and loaded data, so that I can quickly do BI development as well as run demos from that machine, on demand.

- Spend maybe $25/mo? I wouldn't need to do all my work on this machine, only stuff that's heavier or that I'd want to easily demo without requiring my own PC.

Meaning I'd need to be able to image/snapshot, say, a 50GB OS+apps+data image that I could start up/shut down quickly without having to reinstall anything.

Thanks for any real world feedback from people who are doing something like this.

[v-frfei-msft]

Hi @otravers,

To work smoothly on any machine, At first, we need to optimize data model to improve the performance and save memory.  Such as removing useless columns, apply Mark as Date Table to tables with dates. For more details, you can refer to the both online documents. What's more, In my opinion, it would be better to refresh the report when the server (which has On-Premises data gateway and data source) is not heavy.

https://www.sqlbi.com/articles/data-import-best-practices-in-power-bi/

http://blog.pragmaticworks.com/power-bi-performance-tips-and-techniques

Regards,

Frank

[otravers]

Frank, thanks but I'm already aware and apply these best practices. Nonetheless PBI (and Excel) can require a lot of RAM to run.

[mcnater]

Did you ever get a good answer on this? We are trying to set up an environment the same way so we don’t have to have desktop on every single users computer and constantly keep them updated.

[otravers]

@mcnater sorry somehow I missed your question. I ended up using the Azure DSVM with Power BI Desktop, SSMS, and other tools:

https://azure.microsoft.com/en-us/services/virtual-machines/data-science-virtual-machines/

I took a 4-vcore, 28GB RAM VM with a 120GB SSD and have been using it for a year without issues. It costs me maybe $40/mo on average between the hours I use it ($1/h), storage, and a fixed IP. There's a bit of UI lag as I connect to Brazil from Chile with about 120ms latency, but it's bearable.

The only drawback is that Azure doesn't support hibernating VMs, so you have to do a cold boot, i.e. you can't leave your apps and files open and get that state back after deallocating/restarting the VM:

https://feedback.azure.com/forums/216843-virtual-machines/suggestions/4844519-a-sleep-state-of-vm-wh...

[michaelsh]

Here is my experience:

https://community.powerbi.com/t5/Desktop/Working-with-Azure-Virtual-Machine-VM-my-experience/m-p/142...

Somehow DSVM was slow for me.... CPU was 100%, so I use B16ms today.

I don't know the differences in specifications but it is working good for me.