Optimal Placement of Edge-to-Cloud AR/VR Services with Reconfiguration Cost
Mohammadsadeq Garshasbi Herabad, Javid Taheri, Bestoun S. Ahmed, Calin Curescu · 2024
With the emergence of edge-to-cloud computing, determining the optimal service placement in these networks has become increasingly complex, and the problem is classified as NP-complete. This complexity is further amplified by the dynamic nature of the system and the high computational demands of multimedia processing. Maintaining optimal service placement under varying loads requires dynamic reconfiguration, which involves adjusting resource allocation or redeploying service components on the same resource. Although reconfiguration may cause disruptions to services and increase overhead, maintaining services in their current location can lead to substantial operational costs. This study investigates the optimal placement of service components with reconfiguration cost within the edge-to-cloud system when a portion of resources are already allocated to currently running services. We propose a Parallel Multi-Objective Genetic Algorithm (PMOGA) to determine an optimal service placement strategy, balancing placement and reconfiguration costs with efficient execution time. The results obtained from the implementations show that the PMOGA finds service placement strategies closer to optimal as the reconfiguration rate increases, but the reconfiguration cost also becomes more significant. Additionally, we parallelized the algorithm and significantly reduced its execution time. We believe that the findings in this study equip network providers with a valuable resource that allows them to plan optimal reconfigurations, leading to substantial improvements in service quality and agility.