A Framework for Adaptive Resource Management in Edge Computing Environments

Thesis Abstract

Edge computing has emerged as a vital paradigm for processing data at or near the source, thereby reducing latency, conserving bandwidth, and enhancing real-time data analytics. However, the dynamic and distributed nature of edge environments presents significant challenges in effectively managing limited computational, storage, and network resources. Variability in workload demands, heterogeneity of edge devices, and fluctuating network conditions necessitate adaptive resource management frameworks capable of optimizing performance, energy efficiency, and Quality of Service (QoS). This study aims to develop a comprehensive, modular framework that facilitates adaptive resource allocation and load balancing within edge computing environments, thereby addressing the prevalent issues of resource underutilization and overload. The primary objectives include (1) to analyze existing resource management strategies in edge environments; (2) to identify the key parameters influencing resource allocation efficiency; (3) to design an adaptive framework incorporating real-time monitoring, predictive modeling, and dynamic policy adjustment; (4) to implement the framework within a simulated edge environment; and (5) to evaluate its effectiveness through experimental analysis. The research seeks to answer questions concerning the adaptability and scalability of current resource management techniques, the impact of predictive analytics on resource optimization, and the overall improvement in system performance and energy consumption. Methodologically, the study adopts a mixed-methods approach combining quantitative simulation-based experiments with qualitative assessments of framework adaptability. The population of the study comprises a simulated edge environment configured with 50 heterogeneous devices, including IoT sensors, mobile devices, and micro-data centers, modeled to reflect real-world operational variability. A purposive sampling technique selects representative workload scenarios categorized as high, medium, and low demand. Data collection instruments include simulation tools such as CloudSim Extended and EdgeCloudSim, integrated with custom modules for real-time monitoring, workload prediction, and resource allocation algorithms. The framework’s parameters will be validated through synthetic benchmark datasets, while reliability will be ascertained using test-retest methods, ensuring consistency across repeated simulations. Data analysis will employ analysis of variance (ANOVA) to compare performance metrics—such as task completion time, resource utilization efficiency, and energy consumption—across different workload scenarios, complemented by regression analysis to explore predictive relationships between workload parameters and resource allocation outcomes. The framework's effectiveness will also be assessed via key performance indicators (KPIs) including latency reduction, throughput, and system scalability, analyzed through statistical process control charts and sensitivity analysis. It is expected that the proposed framework will demonstrate significant improvements in resource utilization efficiency, reduction in latency, and energy consumption, thereby validating its potential to enhance the operational efficacy of edge environments. The findings will contribute to the theoretical understanding of adaptive resource management, grounded in systems theory and dynamic resource allocation models, extending the current models with real-time predictive capabilities. Moreover, the research will offer practical insights and a replicable framework for developers and practitioners seeking to optimize edge infrastructure performance amidst operational variability. The study concludes that implementing adaptive strategies based on real-time monitoring and predictive analytics can substantially mitigate resource management challenges in edge environments. Recommendations include the integration of the framework into existing edge infrastructure through modular deployment, further development of machine learning models for workload prediction, and exploration of hybrid approaches combining centralized and decentralized management paradigms. Future research directions suggested include longitudinal validation in real-world deployments, exploration of AI-driven autonomic management systems, and extension to multi-cloud orchestration environments, thereby paving the way for resilient and autonomous edge computing ecosystems.