Balancing the Load in a Multi-Paradigm Era: A Comprehensive Survey of Algorithms from Cloud to Quantum

Abstract

Load balancing is a pivotal facet of distributed computing systems that significantly influences the performance of the system by achieving optimal resource utilization, reduced response time, and enhanced system reliability through the even distribution of workloads to computing nodes. The rate at which the computing paradigms are changing has resulted in the increased complexity of the load balancing problem, which is evident in the cloud, fog, edge, grid, and quantum computing since they are particularly characterized by the issues of latency sensitivity, energy efficiency, heterogeneity, scalability, and quantum decoherence. This paper surveys load balancing algorithms implemented in the five computing paradigms, highlighting the main operating principles, architectural differences, scheduling strategies, and performance evaluation criteria applicable to each context. The cloud computing section is devoted to the classification of static and dynamic algorithms, and we also touch on the weighted round-robin, honeybee foraging, and VM migration strategies. The paper also surveys the load balancing of the fog and edge computing wherein we pinpoint the latency-aware and mobility-aware load balancing strategies that are best for the resource-limited, geographically distributed infrastructures. In the grid computing section, peer-to-peer, and decentralized scheduling methods are discussed, which are best suited to loosely coupled networks. Quantum computing is also discussed in the paper as an early stage of load balancing in hybrid classical-quantum systems, the partitioning of quantum jobs, and the coherence limitation of qubit. In a unified taxonomy of algorithms, the paper describes how various algorithms can be mapped depending on their design philosophy, decision criteria, and applicability to different paradigms. Further, the paper lists evaluation charters that are widely utilized in the literature and underscores open issues like cross-paradigm interoperability, scheduling with context awareness, and the trade-offs between energy and performance. By bringing together understanding from these different places, this survey acts as a base for researchers and practitioners who want to create load balancing strategies that are not only adaptable but also paradigm-specific, and cross-platform, and that fit into the changing face of distributed computing.