On Tue, Jul 29, 2025 at 07:08:16AM +0000, Martin wrote: > Mike, does real interrupt controller implemented in vmd already? It should > significantly improve the performance of the vmd by allowing real hardclock > interrupts which significantly reduce CPU usage. >
vmd emulates a legacy i8259 and i8253 for irq routing and clock gen. you can use pvclock or vmm_clock if you want better clocking in guests until something like an hpet or acpitimer is implemented. > More vCPUs per VM moves vmd to the near one level of commercial hypervisors > but better in many cases and much simpler configured. > agree > On Monday, July 28th, 2025 at 7:30 PM, Mike Larkin <mlar...@nested.page> > wrote: > > > On Mon, Jul 28, 2025 at 03:27:15PM +0000, Martin wrote: > > > > > On Monday, July 28th, 2025 at 12:11 PM, Mike Larkin mlar...@nested.page > > > wrote: > > > > > > > On Mon, Jul 28, 2025 at 06:06:46AM +0000, Martin wrote: > > > > > > > > > Hi list! > > > > > > > > > > More than one vCPU per VM is implemented/planned? > > > > > > It is really good idea to have more than one vCPU per VM. vmd looks > > > stable enough from 7.7 as I can test in many production cases. Maybe try > > > to add some multi vCPU feature next? > > > > > > good idea > > > > > > not implemented as of now. > > > > > > > > > Can anybody share some way how to increase VM quantity per host (more > > > > > than four VMs running simultaneously) if host has CPU with 16 kernels > > > > > and 32 threads or more? > > > > > > Just add mote /dev/tapX interfaces to increase VM count running on my > > > hosts. Host RAM is enougth to run 16+ VMs simultaneously for now. > > > > > > makes sense > > > > > > I don't know what this means. You can have as many VMs running as you > > > > want. > > > > The constraint is host RAM vs the memory assigned to each VM. > > > > > > > > > Thanks. > > > > > > > > > > Martin >
