How does storage billing work for serverless endpoints?
I found this article here https://docs.runpod.io/references/faq#how-does-storage-billing-work
But it doesn't mention anything about how endpoints/workers/containers are billed for their storage. Of course network volumes are obvious, because they are separate storage instances.
Say I have an endpoint that has 20GB provisioned for the container disk. How much will I be billed for this? Is it a flat fee regardless of number of workers? Or is it per worker? Where is the pricing listed?
FAQ | RunPod Documentation
RunPod offers two cloud computing services: Secure Cloud and Community Cloud. Secure Cloud provides high-reliability, while Community Cloud offers peer-to-peer GPU computing. On-Demand Pods run continuously, while Spot Pods use spare compute capacity.
44 Replies
.02 per month per gb I think
FAQ | RunPod Documentation
RunPod offers two cloud computing services: Secure Cloud and Community Cloud. Secure Cloud provides high-reliability, while Community Cloud offers peer-to-peer GPU computing. On-Demand Pods run continuously, while Spot Pods use spare compute capacity.
Here
@nerdylive Yes that's the article I linked. I did see this already. My question is really: is this per worker? Per endpoint? During worker operation?
It's not clear
Hmm per gb
Per gbs allocated on your worker, pods
Worker = 1 pod, they don't share storage that's why every worker pulls your image
Imagine it that way
So say you got 2 active workers for a mo. With 5gbs in endpoint so 0.1 *2
* 5gb
That's Howmuch it would cost you
huh
Not sure I follow your post.... formatting issues perhaps?
But I guess I understand the gist, it's by worker at the price of Pod storage pricing, correct?
But complicated because workers are not always active.
So $0.10/GB-month while active and $0.20/GB-month while not active?
Yeah sorry I wrote it on 2 different messages hahah
Yes but I think serverless is never inactive
So only while active
But pods are different, they can be charge when they're inactive if you stop it ( for volume disk )
Okay so $0.10/GB-month?
So an example: our endpoint needs 20GB container disk. If we have 10 workers that would be:
20GB container disk * 10 workers * $0.10/GB-month = $20/month
?
If they're active for the whole months
yes
What do you mean active? Not idle?
you can try to count them to, per sec if im not wrong, try reaching out support on the website to make sure of this
active = running state
when its green
What about when they are idle?
No cost?
Yes i think it doesn't charge you for inactive disk cost
only for pods with no network storage do
oh srry Storage is charged per minute listed there
Should I just write support?
For example, if it's $0.10/GBmonth and applied to serverless active time (assuming per second)
that would be ~$0.000000038580247/s per GB
Sure yes
So, using a network volume might save money because all the serverless functions share the same storage. It’s charged based on the storage used, regardless of how many workers are running. Is that correct?
@Arjun how are you being billed in your idle workers? why am I being billed even no one is calling the serverless endpoint and its just idle.
@Arjun now I deleted the endpoint, my current billing is around 0.9, then after sometime even my serverless endpoint is deleted it is still incurring cost.
and still continue, LMAO, what is this a scam?
Yes
Are you using active workers?
Hmm that's weird, you got endpoints?
@Papa Madiator can you check this
also check if you delete network volume
Yea, but I think network volume is on the storage
?
need more context
Yea he said he got no running workers but still getting billed for serverless it wont happen right
I am now on mobile and it still continues, I already checked all, storage, serverless, template, etc. Check the deletion time and when the cist incur, I already deleted it and still ongoing. I might have different timezone. What is the explanation for this? Im so dissapointed.
What more context you need?
which category is charging you money? Endpoint, serverless or storage? Did you create any API keys?
@agentpabibi pod id or send me on dm your runpod email
That's what I'm trying to figure out. If it is billed only when active, then it could be more expensive to use a Network Volume because a Network Volume is always being charged even when it is not attached to a Pod or Serverless endpoint
both volume and network storage are billed
I understand.
My original question was about Container disks. When do they get billed?
- When they are active only? 24/7?
- Are they billed per worker? Or per endpoint?
- How much are they billed in which state?
Container storage when pod is active
Volume storage when pod is active and when it's stoped
NS when 24/7
Oh sorry, This conversation is in the context of serverless. so I'm talking about the cost of container disks on serverless endpoints.
Check my screenshott earlier, only serverless even after I deleted it. No network volume, storage, etc. I deleted it all already even api key. And why does api key cost something. I deleted all, as in all. I checked it all, all deleted like a newly vreated account.
Oh, if you deleted all then it shouldn’t charge you any money. API key is free, but I was suspect if someone using your api key and without you knowing.
But Arjun on serverless, the space taken from your docker image isnt the container storage on serverless.
I think it's not charged, so I guess using 5gbs on every endpoint is fine, unless your workers add some files on jobs, that's what counts on container disk in workers
So Ns and container disk, both are for different use in my opinion
Ns: more like for storing the files you use every time on your job, model files, codes you execute... Etc
@Arjun your charged only when workers are running, the disk charge is for container disk
for network volume your charged 24/7
serverless does not have pod volumes so you can ignore that
Oh and does the "image size" consumes disks space in serverless?
yes when running, we don't yet charge for image size when not running
Oh it's and is it different with the container disk?
when running container disk will get filled with image data
Oh
I have experienced when the image is bigger than container disk
Why does it still works
its how container image layers and container disk works, not everything has to be in container disk
Ahhh
Ic thanks for explaining