Add support to launch different instances on heterogenous architectures
Modify nova to support launching instances on heterogenous architectures (i.e. x86_64, i386, gpu, sgi, tilepro).
1) Add architecture type in the Services and Instances table in models.py.
2) Specify the compute node's architecture as a flag and set for each compute node in the service table.
3) Upon creating an instance in compute/api.py:
a. extract the architecture from the image.
b. verify the image's architecture and the INTSTANCE_TYPE match.
c. set the instance's architecture in the Instance table.
d. send architecture to the scheduler.
4) Create a new scheduler to find all available hosts for that specific architecture. Select one of those hosts to instantiate the instance.
The backend (hypervisor level for each architecture) support will be a separate blueprint.
Blueprint information
- Status:
- Complete
- Approver:
- Vish Ishaya
- Priority:
- Low
- Drafter:
- None
- Direction:
- Approved
- Assignee:
- USC-ISI
- Definition:
- Obsolete
- Series goal:
- None
- Implementation:
- Unknown
- Milestone target:
- None
- Started by
- Completed by
- Jinwoo "Joseph" Suh
Whiteboard
I would love to see this stuff get merged into nova proper. Any plans for this? --Vish
USC/ISI is still planning on getting this merged up. -- Lorin