I'm considering an analytical model for the cloud computing composed of:
- input load balancer (LB) with M/M/1/K
- P (PH) physical machines with hypervisors, each able to host V (VM) virtual machines.
So the total number of processing units N = P*V.
My question is shall I then put an M/M/1/K queue for each VM or for each PH? In other words, should there be P or N queues? What are the disadvantages and advantages of both approaches? Literature review shows there are both.
Thanks in advance for any comments,
P.