Working with Azure Virtual Machine (VM) - my experience

[michaelsh]

Hello friends,

Three months ago I started working with Azure VM for my Power BI work.

I'll share some of my own experience, and would like to hear from others, too.

My home PC has 16 GB and for several of my large models it's not enough.

I wanted something with lots of memory and good CPU.

I also don't need the VM 24/7, but only when I change something in one of my large models, so I wanted a flexible payment per hour.

I tried "Virtual Desktop" and it was very complicated to set up, so I gave up.

Then I tried a VM  - I think it was Data Science machine d8 v3 / d8s v3 and e4-2ds, but it was slow... and expensive i think (not sure).

Then I tried the one I am still using now (for 3 months now) - Standard B16ms (16 vcpus, 64 GiB memory).

It is fast, plenty of memory. I pay per hour appx. $0.8. I do my work, publish .pbix to the service and switch it off.

I connect via RDP remote desktop, very convenient, directly from Azure portal. There is an option to also connect via browser - even for iPads (though I haven't tried it)

I installed there VPN connectionss to connect to my clients data from the PBI Desktop - works fine.

I do not have Gateways installed there, the Gateways are installed on my clients' servers.

But you can install there anything you want including Gateways (but you'll need to keep it alive 24/7 then)

I use unmanaged disks (I don't remeber why, but I changed from managed to unmanaged - maybe I made a mistake).

This is my only fixed cost (in addition to VM cost per usage). I pay $0.72 each day for this storage. It's $22 a month.

(*** Update *** I just switched to managed Premium SSD disk and shrank it to 32GB (from 128), so it is $5.81 a month only now)

Maybe there is a way to pay the storage also only per hour that VM was on, but I don't have time to explore it further...

This disk is only needed for programs installed on the VM - Windows, Power BI, Excel and others.

All my work files and .pbix models I keep in my OneDrive that I just synced to VM and also to my home PC, so I can work on them from any station.

So this is my setup. I don't understand that much in hardware, I just played around untill I found something that works for me. 

If you want to play around - there is a $200 budget you receive to play with for a month. You can try everything with it. You can see daily cost summary so you know what costs what.
I would love to hear more experiences - what setup do you use for your work scenario?

Michael

[jonespossibly]

is there any info on running PBI desktop on a VM other than this?

I really would like to go back to using a macbook

[ThomasDay]

@jonespossibly   I'm not quite sure what you mean by "on a VM other than this".  I have PBI running on a VM in Azure:  we use a Standard E8as v4 (8 vcpus, 64 GiB memory).  We have attached a 128G OS disk, and 

Premium SSD LRS drive 256G Max,  IOPS:1100, Max throughput (MBps) 125.  So the machine is really big.  It mirrors a computer at our office, actually, that we do our dev and uploading etc. work on.  Super fast.  The machine costs about $1 per hour to run and the disks aren't crazy expensive at all for our use.  The disk...can't remember, but it's small money.

We turn it on and off readily when we need it or do not...and all the data that it absorbs is located on DropBox...which we syncronize automagically to the VM disks when it fires up.  This isn't a big drain on machine time since you can manage that pretty simply through Dropbox smart sync.  We run SQL on it and PBI and it's really reliable and easy.  On the VM, we keep our models "Online only" until we're using them.  Same for some data sets.  

AND you can have multiple people access the VM from where ever they are (one at a time).  We have a SQL guy in a different time zone using it for SQL work related to the models...and others use the VM for PBI.  Obviously we have to coordinate if we just want to share one VM, but you get the idea.  It's a terrific resource for us.

Hope this is something like what you are wondering about.
Tom

[jonespossibly]

Hi Tom - great info, exactly the kind of thing i was looking for.

 By  "on a VM other than this", i mean't either on a different size box or running in AWS or similar.  Most of our cloud is in AWS, so i was curious to see what kind of VM i'd be looking at over there to run PBI Desktop successfully

[ThomasDay]

@michaelsh Would you mind providing an update?  Still happy with AVM (Azure Virtual Machine) service?  I'm presuming that's what you're using though you didn't say that directly.  And you "upload" directly to your workspaces...do you do more than hit publish to web?

Anyway, I'd love to hear what you think today and thanks for posting this.

Cheers!

Tom

[mike_l]

This is great information - thanks for sharing. Would like to set something like this up as well.

[pedroccamaraDBI]

Can i ask you guys something? "My company" wants me to start working with a virtual machine or virtual desktop i don't know. So far, i connect to their SQL databases through a VPN connection.
1. Everybody who has access to that dataset can change it, right?
2. Also, they didn't told me anything about connecting through a VPN. Shouldn't they?
3. Why is not safe anymore keep working the way i use to?

[michaelsh]

@pedroccamaraDBI 
If the company allocates you a virtual machine, there is a big chance it is already connected to their network in a secured way, so you won'y need a VPN.

If you do - you can install in on the VM and use it as you would on your PC.
They want you to work with their VM becase they have a lot more security control over it.

[michaelsh · Accepted solution]

*** Update ***

I just switched to managed Premium SSD disk and shrank it to 32GB (from 128) thanks to this post:

https://jrudlin.github.io/2019-08-27-shrink-azure-vm-osdisk/

Now, I am supposed to pay $5.81 a month fixed price for disk (instead of $22).

This is pretty affordable

[amitchandak]

@michaelsh , Please write a community blog for on that. Contact Admin for access

[riguberto]

Thanks for sharing!