Open Stack

Hi All,

Can I have one compute that acts as a proxy for multiple logically pooled
resources? i.e the proxy compute driver will add one entry in the DB for
each logical resource. The proxy will also update the statistics in the
resource tracker for each of the resource added so that the scheduler can
choose the right resource.
The baremetal provisioning was similar to this. Should the same be used or
there are other implementations?

The approach we were thinking was :
1. the nova compute driver that acts as a proxy will create DB entries in
ComputeNode table for each logical resource.
2. Modify the scheduler so that the scheduler first  determines the correct
node after filtering and weighing . The scheduler as a part of the rpc_cast
will send additional information of the compute node so that proxy compute
driver will know which resource to use.

Thanks
VK Kumar
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstack.org/pipermail/openstack-dev/attachments/20130117/1588f52a/attachment.html>