On 7/3/25 00:02, Anthony Fecarotta wrote:
Thank you. From what I can tell, after being on the mailing list for six
months, it seems most users are running Kubernetes.
The last Ceph survey we did (2022) showed a different picture:
https://ceph.io/en/news/blog/2022/ceph-user-survey-results-2022/
https://ceph.io/en/news/blog/2022/ceph-user-survey-results-2022/images/ceph-user-survey-results-2022.pdf
(see page 74).
It's mostly used in VM environments. I think this still holds true in
2025, although the balance might have shifted a bit towards Kubernetes
in the mean time.
I have been able to get by without implementing Kubernetes. I was looking into
OpenStack or OpenNebula, although many have deterred me from migrating to
OpenStack, citing unneeded complexity. With all that said, GPU time-slicing
(Nvidia MIG) is on our roadmap, and I have read the easiest way to achieve that
is through the Kubernetes GPU Operator. It just seems unnecessary to implement
currently.
We are using OpenNebula. They have support for various GPU
configurations [1]. We have Ceph clusters running as packages on bare
metal, and also (newer clusters) with containers (Cephadm). You
definitely don't need Kubernetes to achieve a high level of automation
with Ceph, but I guess it makes sense when you use Kubernetes
extensively anyway. Just pick what you feel comfortable with.
Gr. Stefan
[1]: https://opennebula.io/blog/product/gpu-and-vgpu-in-opennebula/
_______________________________________________
ceph-users mailing list -- ceph-users@ceph.io
To unsubscribe send an email to ceph-users-le...@ceph.io
