Balancing Energy Efficiency and Infrastructure Knowledge in Cloud-to-Edge Task Distribution Systems

In the rapidly evolving landscape of distributed computing, maintaining energy efficiency in edge and data center infrastructures has become critical. While the problem has been faced with centralized approaches assuming knowledge of the available underlying infrastructure resources, this paper introduces a distributed task allocation framework emphasizing energy awareness without requiring infrastructure knowledge. The framework is designed to optimize energy consumption in heterogeneous computing environments, leveraging a distributed consensus algorithm that allows nodes to maximize individual or global goals. Each private custom utility function enables a node to carefully determine whether executing a task is efficient, thus ensuring flexibility in the task allocation process based on local preferences. While showcasing the energy efficiency of our framework, we also illustrate that it is not necessary to disclose the underlying infrastructure resources status, ensuring the preservation of potentially sensitive local resources information. Experimental results demonstrate the framework's ability to achieve optimal power consumption outcomes while maintaining privacy, offering a significant advancement over traditional centralized allocation policies and Kubernetes-like scheduling algorithms.