Authors:

Much of this developer documentation provides historical context but may not reflect the current state of the project.
If you see outdated content please navigate to the page footer and click "Report an issue on GitHub".

It is not user documentation and should not be treated as such.

User Documentation is available here.

cpu-and-memory

ovirt flow: Use case1:

   Assign Resource proportion(e.g.a customer buy 20%CPU)
   1.find a VM
   2.limited the CPU consumption to be 30% host CPU

Use case2:

   Cluster Policy based Cgroup control
   1.Config a cluster's policy is to be: best utilization/best performance/best balance
   2.Engine config MOM to be different policy:None/medium/aggressive about all resources

Use case3:

   Resource tendency:
   1.Config a VM to be:compute node/ftp server/mail server/http server/general...
     Will have different resource tendency
   2.Engine will translate some resource(mem/cpu/io/bandwith) as rigid level of every vm
   e.g.:VM1(mem:strong/cpu:medium/bandwith:weak),VM2(mem:medium/cpu:weak/bandwith:strong)
   3.Engine config Mom of these policy of different resources
   4.Mom will tune VMs according to the policy

Use case 4:

   User discrimination
   1.Config VM to have different credit on engine
   2.Engine config MOM
   3.restrain resource according to different credit
   e.g.:Gold VM restrain 10% CPU resource, Silver VM restrain 20% CPU resource

Relevant oVirt Projects

  • oVirt engine
  • VDSM
  • MoM

CPU

Stats vdsm needs to be collected:

   1.host cpu usage: flags of host cpu pressure and flag of tuning, when cpu utilization rise, action will be tune or migrate.via libvirt
   2.host perspective guest usage: guest use of host resource,  flag of how much has allocated actually via libvirt
   3.guest perspective guest usage: flag of how much potential the overcommitment can be

Controlls vdsm needs to perform:

   1.via cpu cgroup
   2.via pin cpu
   3.via hibernate/stop vm
   4.via numad

Policies:

   1.QOS: 
     (1)high prior guests demand should be satisefied
     (2)VM's feature demand should be satisefied(compute node's cpu demand has higher priority)
     e.g.:Golden Vm are assigned to a larger quota as original value,
      compute node are pinned to a specific cpu.

   2.Overcommit:
   When:
     (1)host cpu pressure rise,
     (2)import guest or cpu demanding guest demands cpu resource
    it tries to get cpu from:
     (1)non important guest
     (2)non cpu demanding guest
     (3)guest already got very high host usage but with comparatively low priority

Numad Integration


vdsm API should perform: (1)Add domain processes to Numad and the Numactl (2)Remove cpu pin from vdsm

Should MOM control numad parameters? OR use default parameter for numad?

Memory

int virDomainSetMemoryParameters (virDomainPtr domain,

                      virTypedParameterPtr params, 
                      int nparams, 
                      unsigned int flags)
  • Guarenteed
  • Hard Limits
  • Soft Limits