[Virtualization] Virtual machine vcpu tied to core physical machine

Reference article
1: Summary of KVM virtualization CPU technology
Article 2: Configuration of virtual machine CPU and mem (cputune and numatune)
Article 3: Configuration of cpu and numa in libvirt
Article 4: How to improve virtual machine performance? Take advantage of this feature to make it happen!

1. NUMA

The NUMA architecture is a technical solution for multiple CPUs to work together .
There are three main architectures for multi-CPU working together:

1. SMP
2. MPP
3. NUMA

---

SMP and MPP will not be introduced here.
Here we only briefly introduce NUMA.
NUMA: Each processor has its own memory, and each processor can also access the memory of other processors.

---

2. Virtual machine xml configuration (key points)

Virtual machine configuration requirements

1. 8c16g
2. 100g-lvm
3. vcpu绑单NUMA核

---

Regarding the creation of lvm volumes, please refer to my article: [Operating System] Advanced Disk Management
Here is a brief description, taking as /dev/nvme0n1p1an example

# 1.物理卷
pvcreate /dev/nvme0n1p1
# 2.卷组
vgcreate vgdata /dev/nvme0n1p1
# 3.逻辑卷
lvcreate -L 10G -n testblk vgdata

Virtual machine xml configuration about disk

<disk type='block' device='disk'>
	<driver name='qemu' type='raw'>
	<source dev='/dev/vgdata1/testblk'/>
	<target dev='sda' bus='virtio'/>
	<address type='pci' domain='0x0000' bus='0x08' slot='0x01' function='0x0'/>
</disk>

---

Let’s start preparing for core binding! ! !

1.给虚拟机cpu绑定物理机核心
2.给虚拟机cpu分组（非必须）
3.给虚拟机cpu分配内存（非必须）

---

Physical machine numa allocation
lscpu | grep -i numa
Check the virtual machine cpu usage (18289 is the virtual machine process number)
cat /proc/18289/status | grep Cpus_allowed_list

The above are all checking steps and do not involve virtual machine numa binding. The following involves binding.
This article: Bind the vcpu of the virtual machine to the same numa node. The one selected this time is: NUMA node 0

Virtual machine vcpu is bound to the physical machine core

<vcpu cpuset='0-7'>8</vcpu>
<cputune>
	<vcpupin vcpu='0' cpuset='0'/>
 	<vcpupin vcpu='1' cpuset='1'/>
 	<vcpupin vcpu='2' cpuset='2'/>
 	<vcpupin vcpu='3' cpuset='3'/>
 	<vcpupin vcpu='4' cpuset='4'/>
 	<vcpupin vcpu='5' cpuset='5'/>
 	<vcpupin vcpu='6' cpuset='6'/>
 	<vcpupin vcpu='7' cpuset='7'/>
</cputune>

Attributes:
1.vcpu: corresponds to the virtual machine CPU number;
2.cpuset: corresponds to the physical machine CPU number;
(The above two lines of explanation may not be very professional, but they are easier to understand.)

Check the virtual machine core binding status on the physical machine
virsh vcpupin <vm-name>
or: Check the virtual machine CPU usage (18289 is the virtual machine process number)
cat /proc/18289/status | grep Cpus_allowed_list

---

Group virtual machine CPUs (can be omitted)

Divide the vcpu of the virtual machine into multiple nodes

<cpu mode='host-passthrough' check='none'>
    <topology sockets='1' cores='8' threads='1'/>
    <numa>
      <cell id='0' cpus='0' memory='2' unit='GiB'/>
      <cell id='1' cpus='1' memory='2' unit='GiB'/>
      <cell id='2' cpus='2' memory='2' unit='GiB'/>
      <cell id='3' cpus='3' memory='2' unit='GiB'/>
      <cell id='4' cpus='4' memory='2' unit='GiB'/>
      <cell id='5' cpus='5' memory='2' unit='GiB'/>
      <cell id='6' cpus='6' memory='2' unit='GiB'/>
      <cell id='7' cpus='7' memory='2' unit='GiB'/>
     </numa>
</cpu>

# 创建了8个node，每个node的memory为2GiB
vcpu 0 ：绑定在node0
vcpu 1 ：绑定在node1
vcpu 2 ：绑定在node2
vcpu 3 ：绑定在node3
vcpu 4 ：绑定在node4
vcpu 5 ：绑定在node5
vcpu 6 ：绑定在node6
vcpu 7 ：绑定在node7

Query the CPU node allocation en in the virtual machine
lscpu | grep -i numa

, or you can not separate it. If not, the xml is as follows

<cpu mode='host-passthrough' check='none'>
    <topology sockets='1' dies='1' cluster='1' cores='8' threads='1'/>
    <numa>
      <cell id='0' cpus='0-7' memory='16' unit='GiB'/>
     </numa>
</cpu>

---

Bind memory according to node (can be omitted)

<numatune>
	<memory mode="strict" nodeset="0-7"/>
	<memnode cellid="0" mode="strict" nodeset="0"/>
	<memnode cellid="1" mode="strict" nodeset="1"/>
	<memnode cellid="2" mode="strict" nodeset="2"/>
	<memnode cellid="3" mode="strict" nodeset="3"/>
	<memnode cellid="4" mode="strict" nodeset="4"/>
	<memnode cellid="5" mode="strict" nodeset="5"/>
	<memnode cellid="6" mode="strict" nodeset="6"/>
	<memnode cellid="7" mode="strict" nodeset="7"/>
</numatune>

If there is no distinction

<numatune>
	<memnode cellid="0" mode="strict" nodeset="0"/>
</numatune>