Structuring the Continuum

This paper addresses challenges arising from inherent heterogeneity in the rapidly evolving landscape of computing infrastructures, spanning diverse environments across the compute continuum. As a result of the current transition from centralized architectures to contemporary distributed and edge computing models, the demand for innovative programming paradigms becomes increasingly apparent. These paradigms must efficiently harness available resources while seamlessly accommodating their heterogeneity, thus relieving programmers of management burdens. Our work encompasses various resource types within the computational environment to maximize efficiency, responsiveness, and availability. At its core, our contribution introduces a structured programming approach that facilitates the integration and exploitation of dynamically enlisted resources. A dynamic allocation policy and a communication abstraction layer orchestrate resources and workloads adaptively. To address the heterogeneity of the execution environment, we adopt a fat binary that encapsulates diverse executable formats in a single deployment package. Empirical evaluation validates the efficacy of our programming pattern in meeting real-world application needs within the compute continuum. Our approach demonstrates significant promise in navigating the complexities posed by the diverse and dynamic nature of modern computing environments.