Multi-Objective Dynamic Registry Provisioning in Edge Computing Infrastructures

Abstract

The rise of mobile devices and IoT sensors has led to the development of latency-sensitive, data-oriented, and resource-intensive applications. These devices often lack sufficient processing and storage capabilities, but cloud computing can provide the necessary support. However, the distance to cloud data centers causes communication delays. Edge computing addresses this by bringing computational resources closer to users, and improving latency and bandwidth. Container-based virtualization further enhances QoS and QoE due to its lightweight nature, enabling faster application provisioning and reduced resource overhead. Despite these benefits, container provisioning faces significant overhead during content downloads from registries. Existing solutions often focus only on reducing provisioning time, neglecting critical metrics like latency and resource usage. Some approaches use complex techniques like registry migration, which can harm network utilization. We introduce MODyn, a multi-objective dynamic provisioning strategy for container registries. MODyn allocates container registries to edge servers based on infrastructure analysis, eliminating the need for resource migration. Our evaluation shows that MODyn balances provisioning time, latency, and resource usage, optimizing container-based application deployment in edge computing.