Availability statistics

[lesyndrome]

Hello
I have queued a GPU job for over an hour now and as a brand new user I don't know if that's normal or not
I'm sure that you have the analytics for your hardware's availability and I think it would be cool if you could share some of the patterns you see with your users(like at which part of the day the severs are very busy/available). This way we can make better decisions when to use your service for lower queue times and you can better optimize the hardware that you have available.
I found this thread from '17 that proposes a feature that shows some queue time estimation, but I don't see such a feature either.
Love your service, but I don't know if I'm ready to pay for it, if I don't know if I have to wait hours to test a tweak in my NN

Hope I didn't break any forum conventions.. first question here (:

[redeipirati]

Hi @lesyndrome,

I have queued a GPU job for over an hour now and as a brand new user I don't know if that's normal or not

I'm sorry for the bad experience. Usually, the queue should be automatically resolved in 10 minutes, the time requested to auto-scale process to spin up new machines. Longer queue time is definitively a symptom that something is not working as expected. In this case, please reach us out at support@floydhub.com.

I've shared this issue with the Dev Team to investigate, We will update you as soon as possible.

FYI you can also inspect the status page to check the main functionalities of the service.

---

UPDATE

We have just released a fix for this. You shouldn't wait longer than 5-10 minutes when your GPU job/workspace will hit the queue, if not the case please reach us out.

Hope that helps.