base-2 or base-10 system when reporting semantic model sizes on Power BI Service?

[Trish_Rye]

Hi,

Does Power BI Service use base-2 or base-10 system for reporting dataset (semantic model) sizes?

[lbendlin]

the compression algorithms are bit based.  but a rather good approximation is to assume that everything behind the scenes is turtles text all the way down.

[Trish_Rye]

Hi,

I am not sure I understand your answer... Maybe I should have been more precise.

I am extracting info on unused semantic models' size using this api:
https://learn.microsoft.com/en-us/rest/api/power-bi/admin/groups-get-unused-artifacts-as-admin

Since data is in MB, I would like to convert them into GB in my KPIs. I asumed that since this is a storage measurement, I should do it like this: 

size in GB = Size in MB / 1024

But you are saying it's bits (not bytes as above)  which would be strange I think, as it's mostly used for network speed...

[lbendlin]

Ah, you want to know if it is counting in GiB or in GB.  As I said - it doesn't matter. The compression has a much bigger impact.

[Trish_Rye]

OK, compression algorithms make semantic models smaller when they land in PBI Service. However, I need to show how much "space" they occupy in my organization across projects / workspaces in PBI Service, especially when they are not being unused. The stakeholder wants to see the KPI in GB so I divided the MBs by 1024. It's a large global company so the numbers are realy big and in this case 1000 vs 1024 makes a difference.

The size of the semantic models outside PBI Service doesn't interest us so we can disregard the influence of the compression. Maybe I misunderstood what you were trying to tell me.

[lbendlin · Accepted solution]

Tell your stakeholders that is a meaningless KPI. Instead, learn about Fabric pricing for storage.

[Trish_Rye]

Well, I don't wanto to argue anymore but here it's not just the case of storage pricing. Of course, they look at it as a priority and they still have enough space.

However, they need to show the KPI to workspace owners so that they realize how much space is occupied by UNUSED semantic models. 

There is no point in occupying the space, however cheap, if the datasets are unused and they clutter the environment unnecessarily. It's easier to declutter on a regular basis and teach the owners to discontinue unused content. Besides, they believe not only in cost effectiveness but also in environmental issues and it's their mission to avoid accumulating unnecessary data using up IT resources.