TenonOS: A Self-Generating Intelligent Embedded Operating System Framework for Edge Computing

Abstract: The rapid evolution of edge computing has exposed fundamental limitations in traditional operating system and hypervisor architectures, particularly in managing heterogeneous platforms and meeting the constraints of limited resources. Existing solutions often rely on monolithic or layered combinations of hypervisors and guest OSes, which are difficult to tailor for the diverse and dynamic requirements of edge scenarios. To address these challenges, we propose TenonOS, a demand-driven, self-generating, and lightweight operating system framework that fundamentally rethinks and reconstructs both the hypervisor and OS architectures. TenonOS introduces a novel LibOS-on-LibOS approach, in which both virtualization and OS functionalities are modularized into fine-grained, reusable micro-libraries. A dynamic orchestration engine composes these modules on demand to construct customized, application-specific runtime environments. At the core of TenonOS are two key components: Mortise, a minimal, modularized hypervisor, and Tenon, a real-time LibOS. Mortise provides low-overhead resource isolation, fast inter-VM communication, and manages the full lifecycle of Tenon instances - including on-demand creation, suspension, and termination - enabling TenonOS to flexibly adapt its runtime layout to workload variations. Tenon delivers deterministic scheduling and multi-process support for time-critical applications. Through this unified and modular architecture, TenonOS eliminates redundant layers, reduces system overhead, and enhances scalability, security, and maintainability. Extensive evaluations demonstrate that TenonOS achieves superior real-time scheduling (40.28% improvement), a compact memory footprint (361 KiB), and high adaptability to dynamic edge workloads, making it an ideal foundation for heterogeneous, resource-constrained edge systems.