Advancements in Edge Computing Architectures: Optimizing Resource Allocation for IoT Applications

Abstract

This market trend of Internet of Things devices has resulted in an avalanche of data generation which causes significant issues of low latency, bandwidth and real time data analysis requirements to traditional paradigms of cloud-based data processing. Edge computing also has become a new paradigm that allows making computers proximate holistically to the place where the data is being generated, and takes off the load on cloud-based data servers and interrogates data more efficiently and expedites decision making. Since the amount of IoT usage in various fields (i.e. smart city, healthcare, industrial automation and autonomous vehicle) is growing, the need to assign resources efficiently and dynamically to the edge computing architectures is growing as well. We present a review of recent state of the art on edge computing architectures in this article, paying special attention to the optimization of the allocation of resources in the context of an IoT deployment. We then give a summary of the current development of edge architectures to those which can support decentralization in processing, collaborative edge-cloud paradigm and distributed-AI paradigm as well as scalability, reliability and energy efficiency. This encompasses adaptive resource management techniques for dynamically sharing the processing power, memory and bandwidth in real-time depending on the user requirements, the network characteristics and the user priorities. Machine Learning Algorithms anticipating Resources and anomaly detection. These intelligent methods allow edge-systems to estimate possible resource bottlenecks and automatically re-plan computing tasks at a variety of levels (device, edge node and cloud). Mobility: Edge applications are becoming increasingly more modular and mobile through containerization: and microservice architecture. An outline of the new trends in hierarchical and federated edge structures and the enabling distributed decision-making and load-balancing among geographically dispersed edge nodes is also given in the paper. These architectures are essential in applications that need low responsiveness and latency. We as well examine the energy aware models of resource allocation that minimize the calculation burden in order to lengthen the battery life, primarily in remote environments and mobile eco systems. In line with the fact that the IoT data is so sensitive and the risks of cyberattacks are also growing, resource allocation issues characterized by security and privacy concerns are also researched accordingly. In order to enable the allocation processes that one can trust and verify, we introduce blockchain and secure multi-party computation methods. © 2025 IEEE.