Traditional operating systems schedule processes by time slices and priority queues, while existing monitors rely on single-dimensional thresholds that miss complex pathologies such as handle leakage under sufficient memory. Cross-platform semantic heterogeneity prevents unified health comparison. This paper presents the 5DOS Architecture, a closed-loop framework with two layers: (1) a cross-platform Python probe that maps heterogeneous OS metrics into five dimensions (Boundary, Structure, Reserve, Direction, Intensity) and computes synergy coefficients for real-time diagnosis; and (2) a Linux kernel module that translates synergy coefficients into dynamic priorities via an Ontological Floor mechanism, preserving idle processes by preventing synergy collapse to zero. Experiments show the Perception Layer detects memory leaks, CPU bursts, handle leaks, and zombie spawns with 1.1 s delay, 100% recall, and 0.909 F1 on Windows 11 and Ubuntu 22.04, where threshold-based rules score F1 = 0. The Execution Layer classifies 307 entities into seven ontological states on Linux 6.8 with below 0.3% throughput overhead, demonstrating that ontological scheduling can be realized at the kernel level without perceptible performance penalty.
Guiru Zhao (Wed,) studied this question.