Availability of H100 and 2 vCPU / Private image for OS

Hi, I have 2 questions:

  1. We want to use H100, where generally they are more available, at fin-01 or at fin-03? I also spotted that at fin-02 there is a new instance type with additional 2 vCPU, is it more available that previous one at other regions?
  2. We use cloning of disks to create disk for vm ready to use, vm later joins k8s, so disk already has all dependencies and, more specifically, installed docker images for our inference. so we use disks as base images of os. but it takes a while to clone. I wonder if we can somehow make our own private image for os? It will increase speed of provisioning and allow more flexibility for us.

Hi,

  1. Currently H100’s availability is best on FIN-02. You could also consider H200 which has even more VRAM compared to H100 that is also available there.
  2. We will get back separately about the question relating to the disk creation setup.

I also am interested in using private images for the os. We are also doing disk cloning to create the base image for the VM for exactly the same reason (environment preinstalled and ready to use) and it is the most time consuming part of spinning up a VM.

1 Like

I want to raise question about gpu capacity once again.

Today we’ve been experiencing unavailability of H100 GPU instances at fin-03. Are there more instances available at fin-02 than fin-03? It won’t be a problem for us at all to use any region, but for now we have to clone volumes with preinstalled stuff for our work.

This also raises my old question/request about possibility to create our own os image template. Are there any news about this inquiry?

@NovaPulse Hello, in the future, there’ll be a feature called “Clone” in the “Storage ” section in which you can use to copy your Disks to other Data Centers. 1 Full H100 Node is available right now in FIN-02.

@proteindesigner @NovaPulse Customer Images are in our future roadmap so it will take roughly the same time before we eventually get this feature. In the meantime you have to change the existing images to your needs. Hope this helps!

If both of you have any further questions, feel free to reach out!

Nova this is exactly what we do… 1 template image (HDD is OK) at each location. Then you clone → NVMe. It works, but takes an extra couple minutes per instance.

1 Like

@proteindesigner Wow awesome, thank you for the tip :sweat_smile: