Maximizing Profits with ON/OFF Multi-Server Configurations in Distributed Edge Clouds

The Internet of Things (IoT), smart devices have generated large volumes of data and requests, surpassing the capabilities of traditional cloud computing paradigms to meet user requirements. Edge computing, enabled by distributed edge clouds (ECs) in urban areas, offers low-latency and bandwidth-efficient services for IoT devices or end-users. By implementing a multi-server configuration scheme that transitions servers between sleep mode and active state through ON/OFF operations, energy consumption costs can be significantly reduced while maintaining the profit based on service level agreements (SLAs). In this way, we develop a mathematical model for optimizing the multi-server ON/OFF schedule problem within the context of distributed ECs to maximize profits for the edge service provider (ESP). To address this problem with the exponentially high computational complexity and coupling characteristics, we propose a QMIX-based method for configuring multiple servers. Our method is capable of adapting to time-varying and fluctuating user request loads in different areas while making real-time decisions regarding multi-server ON/OFF operations. Experimental results obtained under realistic environmental conditions using real-world user request loads demonstrate that our approach can significantly increase the ESP’s profit.