PMMJC: A preference-based multi-stage matching-mechanism for JointCloud environments

Abstract

With the rise of data-intensive applications, the demand for cloud services has increased significantly, driving the emergence of JointCloud, a novel cloud 2.0 architecture. JointCloud facilitates collaboration among Cloud Service Providers (CSPs) to meet global computational demands. However, as consumer needs become increasingly diversified, the challenge of service matching has grown more complex, particularly in balancing user preferences with CSP resource attributes, such as reputation and data relevance. To address this challenge, this paper proposes a preference-based multi-stage matching mechanism (PMMJC). This mechanism integrates user preferences, CSP reputation, data relevance, risk factors, and Quality of Service (QoS) metrics, employing multi-dimensional optimization methods for service matching. First, a rule-based filtering method is used to quickly eliminate CSPs that do not meet basic resource requirements, narrowing the search space. Next, Uniform Manifold Approximation and Projection (UMAP) dimensionality reduction and the Maximal Information Coefficient estimator (MICe) are combined to assess data relevance and optimize computational efficiency. Then, a coverage decision-making method is applied to derive the Pareto optimal solution set, ensuring balanced performance across multiple dimensions for candidate CSPs. Finally, weighted methods and entropy-weighted fuzzy comprehensive evaluation are used to dynamically adapt to user preferences and generate personalized matching results. Experimental results demonstrate that compared to benchmark methods such as AHP-IOWA and Fuzzy-ETDBA, PMMJC excels in matching efficiency, data relevance accuracy, multi-objective balance, and user satisfaction, significantly enhancing service matching quality in the JointCloud environment.