Yes, you can:

nova boot --image IMAGE_ID --flavor FLAVOR --availability-zone nova:server2

Try reading this

If you have a particular disk image (let's say vmdk) and you want to spawn on vmware hosts only, you must add metadata to glance images:

glance image-update --property hypervisor_type="vmware" IMAGE_ID

Compute hosts can be grouped using host aggregates and availability zones are mapped to host aggregates. Host aggregates are an admin-only concept, availability zones are for the user. Users aren't supposed to know about what servers are in the AZs (or cloud). The aggregates can have metadata associated with them though. As well as the image can have properties (metadata) and flavors can have metadata (extra specs). Those are all combined in the scheduler to map a given request to a given host aggregate. For example, if you have a compute node with ppc64 architecture, and you have a ppc64 image, the scheduler should send the request with that image and that architecture (defined via the image properties) to the proper compute in the aggregate with matching metadata.

Yes, you can force controller to create VM on particular compute node

nova boot --flavor <flavor-id> --image <image-uuid> --nic-net-id=<network-uuid> <name_vm> --availability-zone nova:<compute node="" name="">