本文介绍了一种通过绑定VCPU到特定CPU来提升虚拟机(VM)性能的方法，并提出了将其作为OpenStack扩展API的想法。文章详细描述了设置和获取VCPU绑定状态的API请求与响应格式。
As we tested, when VCPU is less than CPU, if we bind vcpu to cpu， the perfromance of VM will be better. So we want to add this into Openstack
as extended API.
The following memo is what in my mind now:
API :
1. Set instance vcpupin api：
PUT v2/{tenant_id}/vcpupin/{instance_id} Set instance vcpupin
Request parameters JSON:
{
"vcpupin" : [
{
vcpu : 0 # VCPU
cpulist : "1,2" # CPU List , This sample means VCPU0 can run in CPU1 and CPU2
},
{
vcpu : 1 # VCPU
cpulist : "3" # CPU List , This sample means VCPU1 can run in CPU3
}
]
}
2. Get instance vcpupin api：
GET v2/{tenant_id}/vcpupin/{instance_id} Get instance vcpupin status
Response parameters JSON:
{
"vcpupin" : [
{
vcpu : 0 # VCPU
currentcpu : 1 # VCPU is running on which CPU
state : "running" # State
cputime : "946.4" # cpu time, 946.4 means 946.4sec
cpulist : "1,2" # CPU List , This sample means VCPU0 can run in CPU1 and CPU2
},
{
vcpu : 1 # VCPU
currentcpu : 3 # VCPU is running on which CPU
state : "running" # State
cputime : "1001.2" # cpu time, 1001.2 means 1001.2sec
cpulist : "3" # CPU List , This sample means VCPU1 can run in CPU3
}
]
}
Function：
1. Basic call flow：
api/CLI client ----HTTP call---->nova-api ----MQ:asyc call----> nova-compute
2. Set instance vcpupin ：
In nova-compute, call API/CLI do the thing like following:
Before cpupin:
[root@ci13sjcmp03 ~]# virsh vcpuinfo 1 <---1 is instance domain
VCPU: 0
CPU: 1
State: running
CPU time: 1.3s
CPU Affinity: yyyyyyyy <-- VCPU0 can run on CPU0-CPU8
Run cpupin:
[root@ci13sjcmp03 ~]# virsh vcpupin 1 0 1
After cpupin:
[root@ci13sjcmp03 ~]# virsh vcpuinfo 1
VCPU: 0
CPU: 1
State: running
CPU time: 1.3s
CPU Affinity: -y------ <-- VCPU0 can only run on CPU1
3. Get instance vcpupin ：
In nova-compute, call API/CLI do the thing like following:
[root@ci13sjcmp03 ~]# virsh vcpuinfo 1
VCPU: 0
CPU: 1
State: running
CPU time: 1.3s
CPU Affinity: -y------
扫一扫
5973
被折叠的 条评论
为什么被折叠?
到【灌水乐园】发言
