HPC cluster features

[tinx]

First, congratulations for the integration ha cluster.
in the future will be integrated on proxmox 2 the features for HPC clustering?

thanks
luca

 

[tom]

how does HPC clustering applies for a virtualization solutions?

 

[Florian Heiden]

In the case of having 4 identical nodes as a HPC cluster the virtual machine can use all 4 resources instead of only the physical limit of one node.
E.g. a node has a 8 core CPU and 16 GB RAM the VM could use up to 24 Cores and 64 GB RAM.

Nevertheless Proxmox VE is a very nice solution for system virtualization.

Regards
Florian

 

[KonstantiH]

In the case of having 4 identical nodes as a HPC cluster the virtual machine can use all 4 resources instead of only the physical limit of one node.
E.g. a node has a 8 core CPU and 16 GB RAM the VM could use up to 24 Cores and 64 GB RAM.

Hi all,
Where can I find these magical settings?

 

[badji]

Bonjour,
You can create HPCs with pools of VMs on a Proxmox cluster. With OpenHPC, Dcos / Mesos projects ... for big-data, Data-sciences, IA ...
Note: I have never used GPU with Proxmoxve, perhaps with PCI-PASSTROUGH. For HPCs NVIDIA TESLA GPUs bring a lot, in terms of calculation ...
https://openhpc.community/
https://mesos.apache.org/
https://dcos.io/
Merci.

 

[KonstantiH]

Bonjour,
You can create HPCs with pools of VMs on a Proxmox cluster. With OpenHPC, Dcos / Mesos projects ... for big-data, Data-sciences, IA ...
Note: I have never used GPU with Proxmoxve, perhaps with PCI-PASSTROUGH. For HPCs NVIDIA TESLA GPUs bring a lot, in terms of calculation ...
https://openhpc.community/
https://mesos.apache.org/
https://dcos.io/
Merci.

Hi badji,

I just started to learn the HPC environment, and there is a lot of information for me.

Many thanks!

 

[tommisan]

Bonjour,
You can create HPCs with pools of VMs on a Proxmox cluster. With OpenHPC, Dcos / Mesos projects ... for big-data, Data-sciences, IA ...
Note: I have never used GPU with Proxmoxve, perhaps with PCI-PASSTROUGH. For HPCs NVIDIA TESLA GPUs bring a lot, in terms of calculation ...
https://openhpc.community/
https://mesos.apache.org/
https://dcos.io/
Merci.

Hi badji,
could you please share some more details about your HPC implementations (i.e. distributed Storage, networking, etc)?
Thanks in advance

 

[badji]

Hi,
Sorry, but I no longer work with these technologies (Dcos-Mesos ...)
I work on new technologies provided by CANONICAL among others with the Projects: MAAS + JUJU + K8S + CHARM + SLURM ...
The deployment is done by bundle of charms with the juju project.
Check out these two links:

https://discourse.charmhub.io/t/fee...3-kubernetes-1-19-3-all-on-focal-20-04-1/3783

https://omnivector.solutions/

juju deploy cs:bundle/slurm-openfoam-0

Bye.