VM Resource Allocation¶

Proper resource allocation for KVM/libvirt VMs ensures stable performance and prevents host resource exhaustion.

Memory Allocation¶

Static Allocation¶

Set fixed memory in VM XML:

<memory unit='GiB'>16</memory>
<currentMemory unit='GiB'>16</currentMemory>

Or via virsh:

# Set maximum memory (requires VM shutdown)
virsh setmaxmem myvm 16G --config

# Set current memory
virsh setmem myvm 16G --config

Memory Ballooning¶

Ballooning allows dynamic memory adjustment. The guest OS releases unused memory back to the host.

Enable in VM XML:

<memballoon model='virtio'>
  <stats period='10'/>
</memballoon>

Adjust at runtime:

# Reduce to 8GB (guest releases memory)
virsh setmem myvm 8G --live

# Increase to 12GB
virsh setmem myvm 12G --live

Note

Ballooning requires guest cooperation. Windows needs the virtio balloon driver; Linux includes it by default.

Memory Locking¶

For latency-sensitive VMs, lock memory to prevent swapping:

<memoryBacking>
  <locked/>
</memoryBacking>

Requires ulimit -l adjustment for libvirt.

Hugepages¶

Hugepages reduce TLB misses for memory-intensive VMs.

Configure Host Hugepages¶

# Check current hugepages
grep Huge /proc/meminfo

# Reserve 16GB of 2MB hugepages (8192 pages)
echo 8192 | sudo tee /proc/sys/vm/nr_hugepages

# Make persistent
echo "vm.nr_hugepages = 8192" | sudo tee /etc/sysctl.d/hugepages.conf

For 1GB hugepages (better for large VMs):

# Add to kernel cmdline in GRUB
GRUB_CMDLINE_LINUX="hugepagesz=1G hugepages=32"

VM Hugepages Configuration¶

<memoryBacking>
  <hugepages>
    <page size='2048' unit='KiB'/>
  </hugepages>
</memoryBacking>

Or for 1GB pages:

<memoryBacking>
  <hugepages>
    <page size='1048576' unit='KiB'/>
  </hugepages>
</memoryBacking>

vCPU Assignment¶

Basic vCPU Configuration¶

<vcpu placement='static'>8</vcpu>

Via virsh:

# Set vCPU count (requires shutdown for increase)
virsh setvcpus myvm 8 --config --maximum
virsh setvcpus myvm 8 --config

CPU Topology¶

Define sockets/cores/threads to match guest OS expectations:

<vcpu placement='static'>8</vcpu>
<cpu mode='host-passthrough'>
  <topology sockets='1' dies='1' cores='4' threads='2'/>
</cpu>

CPU Pinning¶

Pin vCPUs to specific host cores for consistent performance:

<vcpu placement='static'>4</vcpu>
<cputune>
  <vcpupin vcpu='0' cpuset='4'/>
  <vcpupin vcpu='1' cpuset='5'/>
  <vcpupin vcpu='2' cpuset='6'/>
  <vcpupin vcpu='3' cpuset='7'/>
</cputune>

View host CPU topology:

lscpu --extended
virsh capabilities | grep -A20 '<topology>'

Emulator Pinning¶

Pin QEMU emulator threads separately:

<cputune>
  <vcpupin vcpu='0' cpuset='4'/>
  <vcpupin vcpu='1' cpuset='5'/>
  <emulatorpin cpuset='0-1'/>
</cputune>

NUMA Configuration¶

For multi-socket systems or large VMs, NUMA awareness improves performance.

View Host NUMA¶

numactl --hardware
virsh capabilities | grep -A50 '<numa>'

VM NUMA Configuration¶

<numatune>
  <memory mode='strict' nodeset='0'/>
</numatune>
<cpu mode='host-passthrough'>
  <numa>
    <cell id='0' cpus='0-7' memory='16' unit='GiB'/>
  </numa>
</cpu>

I/O Resource Control¶

Disk I/O Limits¶

<disk type='file' device='disk'>
  <driver name='qemu' type='qcow2'/>
  <source file='/tank/vms/myvm/disk.qcow2'/>
  <target dev='vda' bus='virtio'/>
  <iotune>
    <read_bytes_sec>104857600</read_bytes_sec>   <!-- 100 MB/s -->
    <write_bytes_sec>52428800</write_bytes_sec>  <!-- 50 MB/s -->
    <read_iops_sec>1000</read_iops_sec>
    <write_iops_sec>500</write_iops_sec>
  </iotune>
</disk>

Set at runtime:

virsh blkdeviotune myvm vda --read-bytes-sec 104857600 --live

Network Bandwidth¶

<interface type='bridge'>
  <source bridge='br0'/>
  <bandwidth>
    <inbound average='125000' peak='250000' burst='256'/>   <!-- KB/s -->
    <outbound average='125000' peak='250000' burst='256'/>
  </bandwidth>
</interface>

Resource Monitoring¶

VM Statistics¶

# Overview of all VMs
virsh list --all

# CPU and memory for running VMs
virsh domstats

# Specific VM stats
virsh domstats myvm

# CPU usage
virsh cpu-stats myvm

# Memory stats (requires balloon)
virsh dommemstat myvm

Detailed Statistics¶

# Block device stats
virsh domblkstat myvm vda

# Network stats
virsh domifstat myvm vnet0

# Complete domain info
virsh dominfo myvm

Real-Time Monitoring¶

# Watch VM stats (updates every 2 seconds)
watch -n2 'virsh domstats --cpu-total --balloon --block --interface'

# virt-top for interactive view
sudo apt install virt-top
virt-top

Live Resource Adjustment¶

Memory¶

# Adjust current memory (within max)
virsh setmem myvm 8G --live

# Check current allocation
virsh dominfo myvm | grep memory

vCPUs¶

# Reduce vCPUs (if hotplug enabled)
virsh setvcpus myvm 4 --live

# Check current vCPUs
virsh vcpucount myvm

I/O Limits¶

# Adjust disk I/O
virsh blkdeviotune myvm vda \
  --read-bytes-sec 209715200 \
  --write-bytes-sec 104857600 \
  --live

# Adjust network bandwidth
virsh domiftune myvm vnet0 --inbound 250000,500000,512 --live

Best Practices¶

Windows VMs¶

Setting	Recommendation
Memory	Fixed allocation, no ballooning
vCPUs	Even number, matching host topology
Storage	virtio with latest drivers
Network	virtio with latest drivers

Windows-specific XML:

<features>
  <hyperv mode='custom'>
    <relaxed state='on'/>
    <vapic state='on'/>
    <spinlocks state='on' retries='8191'/>
    <vpindex state='on'/>
    <synic state='on'/>
    <stimer state='on'/>
  </hyperv>
</features>
<clock offset='localtime'>
  <timer name='hypervclock' present='yes'/>
</clock>

Linux VMs¶

Setting	Recommendation
Memory	Ballooning enabled for flexibility
vCPUs	Match workload, can overcommit
Storage	virtio-scsi for multiple disks
Network	virtio (default driver)

General Guidelines¶

Don't overcommit memory for production VMs
Reserve host resources: Leave 4-8GB RAM and 2-4 cores for host
Use CPU pinning for latency-sensitive workloads
Enable hugepages for VMs >8GB
Monitor regularly with virsh domstats and virt-top

Resource Budget Example¶

For a 128GB / 16-core host running mixed workloads:

Workload	Memory	vCPUs	Notes
Host reserved	8GB	4	OS, Docker, services
Windows VM	24GB	6	Gaming, GPU passthrough
Linux VM	16GB	4	Development
LLM inference	64GB	8	Ollama containers
Available	16GB	-	Headroom

See Capacity Planning for system-wide resource strategy.

Next Steps¶

KVM Setup for installation and configuration
GPU Passthrough for dedicated VM graphics
Windows 11 VM for gaming VM setup