Hi there,
Consider a 3 node setup with each node having 2x 10GB & 6x 1GB Nic.
Is it more interesting to speedup replication between the nodes using 10GB interconnects and have the VM's communicate on 1GB (scenario 1) or is it more interesting to speed up VM <—> desktop access (scenario 2)
My incline (but I can be wrong of course ;-) ) is that replication on 1GB (scenario 2), should be sufficient as it is always connected & buffered (? not sure ?) whereas access to VM will be more event-driven (spiky bandwidth usage) & give lower access-time on a 10GB link.
In scenario 2, the 10GB link would be shared by a reasonable big ERP system (DB is loaded quasi 100% in RAM VM, so disk access is not a restraint there) & by a file-server.
I would like to use a mesh-setup for the data replication, thus 2 Nics are required for the interconnect (https://pve.proxmox.com/wiki/Full_Mesh_Network_for_Ceph_Server).
Switches for connection to network would be redundant too (stacked).
Thanks for your input on this guys & gals!
Jord
Consider a 3 node setup with each node having 2x 10GB & 6x 1GB Nic.
Is it more interesting to speedup replication between the nodes using 10GB interconnects and have the VM's communicate on 1GB (scenario 1) or is it more interesting to speed up VM <—> desktop access (scenario 2)
My incline (but I can be wrong of course ;-) ) is that replication on 1GB (scenario 2), should be sufficient as it is always connected & buffered (? not sure ?) whereas access to VM will be more event-driven (spiky bandwidth usage) & give lower access-time on a 10GB link.
In scenario 2, the 10GB link would be shared by a reasonable big ERP system (DB is loaded quasi 100% in RAM VM, so disk access is not a restraint there) & by a file-server.
I would like to use a mesh-setup for the data replication, thus 2 Nics are required for the interconnect (https://pve.proxmox.com/wiki/Full_Mesh_Network_for_Ceph_Server).
Switches for connection to network would be redundant too (stacked).
Thanks for your input on this guys & gals!
Jord