Bioinspired Model for Edge-based Task Scheduling Applications

Abstract

Optimization of task scheduling for edge devices requires extensive analysis of task computational requirements & capacities of edge VirtOptimizingual Machines (VMs). Scheduling models are responsible for mapping edge tasks with these VMs for minimization of make-span, with higher deadline-hit-ratio, and lower computational complexity. To perform this task, a wide variety of bioinspired models are proposed by researchers, and most of them use static learning rates, which limits their task-to-edge mapping performance. Moreover, adaptive learning rate models are highly complex, which increases their convergence delays, thereby reducing scheduling efficiency under heterogeneous task types. To overcome these limitations, this text proposes design of an Improved & Adaptive Bioinspired Model for Edge-based Task Scheduling Applications, which uses League Championship Model (LCM). It optimizes its learning performance via an adaptive stochastic process. The model sets-up an initial learning rate for scheduling tasks to edge VMs, and then uses a Grey Wolf Optimizer (GWO) to continuously update learning rate for better mapping performance. The GWO Model uses low-complexity wolf behavioral functions in order to determine optimum learning rates for different VM & task types. This allows for faster convergence, and lower service delays when evaluated on standard scheduling datasets. The model was compared in terms of scheduling efficiency, make-span, computational complexity, and delay needed for scheduling, with various state-of-the-art models. It was observed that the proposed model outperformed them w.r.t. these evaluation metrics under most use cases, wherein it showcases 8.5% lower make-span, 1.5% higher efficiency, 5.9% lower computational complexity, and 8.3% lower computational delay, thereby making it useful for a wide variety of real-time use casesthey had developed.