Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

credit-account-charges page lists overlapping charges #314

Open
matyasselmeci opened this issue Sep 20, 2022 · 4 comments
Open

credit-account-charges page lists overlapping charges #314

matyasselmeci opened this issue Sep 20, 2022 · 4 comments

Comments

@matyasselmeci
Copy link
Contributor

A lot of the tables in https://path-cc.io/credit-account-charges/ have overlapping numbers. For example, in the CPU Credits table, 2-8 cores per job are charged at 1.2 per core hour, and 8-32 cores per job are charged at 1.5 per core hour.

Also the Memory Per Core table shows that "2-8 GB greater than nominal" is charged at 0.125. Since "nominal" is up to 2 GB, that language makes it seem like the range is 4-10 GB.

Talk to @jasoncpatton to find out how the overlaps should be resolved.

@jasoncpatton
Copy link
Contributor

Currently, I've resolved the overlaps by being inclusive for the lower range and exclusive on the higher ranges, e.g. I charge the "2-8 core" rate for jobs requesting 8 cores, and the "8-32" rate for jobs requesting 9 (to 32) cores. This can be changed if needed.

@CannonLock
Copy link
Collaborator

CannonLock commented Sep 21, 2022

@jasoncpatton

Can you also clarify the discrepancies in "Additional GPU Credit Charge". The table head reads "per job" and the body reads "per GPU".

Same with "Additional CPU Credit Charge".

@jasoncpatton
Copy link
Contributor

@jasoncpatton

Can you also clarify the discrepancies in "Additional GPU Credit Charge". The table head reads "per job" and the body reads "per GPU".

Same with "Additional CPU Credit Charge".

Despite the wording being a little confusing, when parsing it out, I'm not sure I see where the problem is. There is a "nominal" amount of memory allowed without extra charge per CPU (or per GPU). Beyond that, you're charged for the extra memory usage per job per hour, not per core per hour, which is why the table has a different header.

@CannonLock
Copy link
Collaborator

The way it is currently displayed I would expect the appropriate charge range to be found using.

total_memory / num_gpus

rather then

total_memory - ( num_gpus * 128 ) 

The memory per gpu table should really be more like:

Memory (GB) over nominal * Number of GPUS

0

256

384

And the same to the other memory table.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

3 participants