Abstract :
Medium-scale private clouds are being widely used in enterprises and universities. While, these clouds have a relatively small pool of resources, diversity of those resources, users, and their needs are still comparable with public clouds. We present a resource and policy aware Virtual Machine (VM) scheduling solution for such medium-scale clouds. The proposed scheduler enables the deployment of VMs based on a predefined set of policies and user priorities, while being aware of the resource utilization of the cloud. This is achieved by periodically polling resource statistics from the cloud nodes, enforcing a set of predefined policies, taking into account the priority levels of users and VM requests, and then scaling, migrating, and preempting VMs based on available resources and policies. Such resource and policy aware scheduling improves resource request acceptance rate and increases the utilization of cloud resources. A proof of concept solution is implemented using Apache CloudStack and validated against a carefully crafted set of resource requests.
