Understanding Fabric Capacity Metrics

[JeremyYoung]

Hi All

I am trying to work out what my Spark job converting 50k deeply nested jsons to lakehouse tables will cost me in terms of CUs. I did the work in January. I have a Trial Fabric licence. I have today installed the Metrics app and am seeing this:

1. It appears that it only analyzes the last 14 days. Is that correct? Can I backdate it to look at January?

2. It is showing 18727 seconds (=5hrs) of Synapse Notebook usage on 24th Feb (my birthday) when nothing was running AFAIK. Any ideas what this would be?

3. What is the relationship between what I see here and the number of CUs I need? The trial is F64 and I will want to buy F2. How do I understand what this workload (whatever it is!) will look like on an F2?

The storage tab looks like this:

That 3GB of storage was taken up in January. What is the Cumulative billable storage widget showing me? The description on the website is unhelpful. "Cumulative billable storage is calculated as a sum of billable storage from the start of the selected date time period." The date-time period is the two weeks the report is allowing me to look at. How can the cost be zero at the start of each rolling two-week period?

Any pointers gratefuylly received!

Thanks.

[R1k91 · Accepted solution]

my 2 cents

2. It is showing 18727 seconds (=5hrs) of Synapse Notebook usage on 24th Feb (my birthday) when nothing was running AFAIK. Any ideas what this would be?

not sure but:

- notebooks session stay alive even if you closed the notebook itself. 

Microsoft Fabric – Notebook session usage explained (And how to save CU’s or billed time) - FourMoo ...

- bursting is something that happens behind the scenes also borrowing capacity up to 24 hours for background operations.

3. What is the relationship between what I see here and the number of CUs I need? The trial is F64 and I will want to buy F2. How do I understand what this workload (whatever it is!) will look like on an F2?

the only secure way is to try on a F2 but you can image that it could be 32 times higher on F2. again you should also consider smoothing and bursting and maybe you can fit in an F2 because they smooth your workload avoiding spikes over 100%.

[Anonymous]

Hello @R1k91 ,

We haven’t heard from you on the last response and was just checking back to see if you have a resolution yet .
In case if you have any resolution please do share that same with the community as it can be helpful to others .
Otherwise, will respond back with the more details and we will try to help .

[Anonymous]

Hi @R1k91 ,

We haven’t heard from you on the last response and was just checking back to see if you have a resolution yet .
In case if you have any resolution please do share that same with the community as it can be helpful to others .
Otherwise, will respond back with the more details and we will try to help .

[JeremyYoung]

I have been looking further at the metrics and decided it's not fit for purpose. It does, indeed, seem a no-brainer to just provision an F2 and try it out. I have requested permission from the powers-that-be in my organisation to do that. Thanks