EleVMate — A data-driven approach for ‘on-the-fly’ horizontal small datacentre scalability and VM starvation

Abstract

The volume of data exchanged over the network has increased exponentially over the past few years, driving the need for a new solution. To address this issue, a new layer for computation and storage between the cloud and the user has been introduced to reduce the load on the network. To facilitate this solution, small datacentres have emerged characterized by their limited capacity in terms of computation and storage resources. These datacentres are managed by small fog service providers (SFSPs) and are used to host virtual machines (VMs) with QoS requirements that vary based on application and user subscription. Being limited in capacity, these SFSPs are unable to accept all incoming VM requests, forcing them to reject some VM requests based on QoS requirements. This causes an issue for VMs with minimum QoS requirements as they are kept waiting for unlimited time causing them to starve. VM starvation in SFSPs is a critical issue that has not been addressed in the literature. To address this issue, in this paper, we propose to support the SFSP by horizontally scaling up its resources using volunteer nodes. We propose, EleVMate, a novel management framework to manage the allocation of VMs to the SFSP nodes. Our solution works to prioritize the placement of priority VMs in these datacentres. We validated the solution using a publicly available VM dataset and we concluded that priority influences VM placement in SFSP. Last, our framework has improved the performance of the SFSP by reducing rejection rate by at least 4% with existing solutions.