Elastic Autoscaling for Distributed Workflows in MEC Networks

With the recent advancements in computing technologies, new paradigms have emerged enabling users to access a large variety of distributed resources, overcoming several limitations of localized applications and information storage. Among these paradigms, Mobile Edge Computing (MEC) places storage and computing capabilities at the edge of the network, significantly decreasing congestion and service response times, at the cost of limited capacities. Within this context, the emergence of novel computationally intensive services has triggered the necessity to design algorithms that adaptively scale resources, achieving solutions tailored to traffic demand. In this paper, we present a preliminary scaling method to determine the resource provisioning of complex workflows of web services that are distributed on a MEC infrastructure, with the intent of improving the distribution of the end-to-end response time of the workflow. The method is designed to run compositionally, exploiting a structured hierarchical workflow representation, enabling efficient top-down determination of the resource provisioning. The method is also formalized to act considering the inherent limitations and complexities of an MEC network landscape. In so doing, we demonstrate the applicability of the approach on two synthetic application scenarios, confirming the validity of the proposed elastic scheme in optimizing resource management within a resource-constrained MEC network.