Thermal and Power Optimization Service for Cloud Computing Software

Registered by Nishi Ahuja on 2015-07-28

Towards to software defined infrastructure, the power and thermal intelligence is being adopted to optimize workload scheduling, which can help improve efficiency, reduce power, as well as to improve data center PUE and lower down operation cost in data center. The architecture will utilizing Power, thermal, Compute telemetries to optimize cloud workload scheduling to achieve more efficient datacenter management and better business continues.
The architecture will support the following feature:
• Thermal and power data telemetry : Ceilometer can provide information
• Real time updated thermal and power data
• Inlet temperature
• Outlet (Exhaust Air) temperature
• Volumetric airflow
• System Power
• CUPs (Compute Usage per Second)
• Fan speed

• Work load migration based on power and thermal data.
leveraging the live-migration feature in OpenStack, the architecture provides a policy to migration the virtual machines based on current thermal conditions. One of the thermal awareness migration policy migrates the virtual machines from the host with high outlet temperature to other hosts with healthy thermal status, in order to avoid hotspot in the datacenter. The standalone energy package provides multiple thermal awareness migration policies, including the policies keep uniform outlet temperature, uniform airflow, migration based on computing resource utilization (CUPS).
• Work load placement based on thermal and power status.
With the scheduler, the newly launched virtual machine can be scheduled to a host with relatively good thermal condition in cluster. For example, the thermal awareness scheduler will select the host with lowest inlet temperature to put the new work load and avoid hot spot in the cluster. Also, the placement based on power consumption curve and simple utilization balance is provided.
• Active/standby management of cluster hosts.
The architecture can manage the power status of the hosts in cluster, power off or on hosts according to the work load status, thermal or power conditions.

Blueprint information

Status:
Complete
Approver:
Antoine Cabot
Priority:
High
Drafter:
Nishi Ahuja
Direction:
Needs approval
Assignee:
junjie.huang
Definition:
Obsolete
Series goal:
Accepted for mitaka
Implementation:
Implemented
Milestone target:
milestone icon mitaka-2
Started by
Antoine Cabot on 2016-01-12
Completed by
Antoine Cabot on 2015-12-02

Related branches

Sprints

Whiteboard

this BP must be split up in 4 BPs

(?)

Work Items

This blueprint contains Public information 
Everyone can see this information.