1 Controller 2 Nodes = One Large VM?

kobaltz

New Member
I'm relatively new to the whole private cloud business but was wondering if this is possible with Proxmox. I've looked around and couldn't find a clear answer. Let's suppose I have 3 machines. I have 1 primary controller and two nodes. Let's assume that all machines have the following specs.

Core i7 (4 Cores)
8GB RAM
500GB HDD

So in total, my private cloud has 12 Cores, 24GB RAM and 1.5TB HDD space. I am looking to create two VMs that each will use 5 Cores, 10GB Ram, and 100GB HDD.

So in this scenario you can kind of see my question. Can you stretch the number of resources allocated to a single VM across different nodes to create one powerful VM? I know that there may be some issues with IOPS, but my question really is in regards to sharing resources on Proxmox v1.8.
 
I'm relatively new to the whole private cloud business but was wondering if this is possible with Proxmox. I've looked around and couldn't find a clear answer. Let's suppose I have 3 machines. I have 1 primary controller and two nodes. Let's assume that all machines have the following specs.

Core i7 (4 Cores)
8GB RAM
500GB HDD

So in total, my private cloud has 12 Cores, 24GB RAM and 1.5TB HDD space. I am looking to create two VMs that each will use 5 Cores, 10GB Ram, and 100GB HDD.

So in this scenario you can kind of see my question. Can you stretch the number of resources allocated to a single VM across different nodes to create one powerful VM? I know that there may be some issues with IOPS, but my question really is in regards to sharing resources on Proxmox v1.8.
Hi,
your VMs can only use things from the node on which they run. In your case max. 4 cores, round 6-7GB Ram and the free space on the logical volume pve-data.

Udo
 
Thanks for your reply Udo. This could potentially be a deal breaker. Is what I'm asking for even possible with other virtualization softwares (is ESXi, UEC, etc)?
Hi,
i don't know UEC, but with ESX it's the same. I can't belive that your wish is practicable - if you use one core on another node, you must sync all the memory from the VM in both directions on both nodes (over the network). I guess you will never see such an slow VM before!
Memory is cheap and cores are also not so expensive (if you take the right cpus ;-) )

Udo
 
I don't think that a "classic" cloud infrastructure as of today can deliver what you want.
If you have distributed resources in your cloud, you can however make them work to deliver on the same goal, like combining their powers to find a solution to a computational challenge.
See here: http://en.wikipedia.org/wiki/Distributed_computing
A well known example of this is the SETI research programme, using the BOINC application control layer.