随着移动通信、人工智能等技术的发展,智能设备和数据呈现爆炸式增长的态势,物联网(Internet of things,IoT)场景对算力、时延和能耗提出了更高的要求.算力网络通过对计算节点进行互联,基于统一的算力度量标准和任务调度策略,实现算力资源的共享和高效利用,为提升物联网系统的计算性能提供了新思路.但是由于物联网设备种类繁多、网络连接方式各异且对功耗敏感,当前以计算能力为主的算力度量方法无法满足物联网设备协作的需求.此外,目前算力网络的计算任务调度方法普遍依赖中心化的网络路由节点或管理平台,不能适应物联网设备分布离散和资源受限的特点.针对上述问题,提出了一种面向物联网端侧设备的新型算力度量架构,为异构物联网端侧算力资源提供计算、存储、通信、功耗和电源的统一度量.在此基础上,提出了一种分布式的任务调度策略,实现离散异构算力资源与业务场景需求的智能匹配,支持物联网端侧设备的资源管理和任务调度.选取智慧家庭场景对提出的算力度量架构进行评估,结果表明,该架构可以有效地实现端侧设备算力资源的共享和调度,提升物联网计算效率,减少能源消耗.
Computational measurement and task scheduling:a study on IoT edge device strategies
The rapid advancement of mobile communications and artificial intelligence has catalyzed an exponential increase in intelligent devices and data generation. This surge necessitates enhance the computational resource capa-bilities,particularly in the Internet of things (IoT) environments,where there are pressing demand for improved re-source management in terms of computation,latency,and energy efficiency. The concept of a computility network,which leverages interconnected computing nodes for resource sharing and optimization based on a unified measure-ment standard and task scheduling strategy,offers a promising solution for augmenting IoT systems' computational performance. However,the current models for computing resource measurement,predominantly focused on computa-tional capacity,fall short in addressing the diverse and collaborative needs of various IoT devices. These devices often differ in network connectivity modes and exhibit sensitivity to power consumption. Moreover,prevalent task schedul-ing methods in computility network predominantly rely on centralized network routing nodes or management plat-forms. Such approaches are not well-suited for the unique characteristics of IoT devices,which are typically dispersed and constrained in resources. To address these challenges,a novel architecture for computing resource measurement tailored to IoT devices was introduced. A comprehensive and unified framework for measuring diverse aspects of computing resources in heterogeneous IoT environments was provided,including computation,storage,communica-tion,power consumption,and power supply metrics. Building on this foundation,a distributed task scheduling strat-egy that intelligently aligned the disparate computing resources with specific business scenario requirements was pro-posed,thereby facilitating efficient resource management and task scheduling for IoT devices. To validate the effec-tiveness of the proposed architecture,it was applied to a smart home scenario. The empirical results demonstrate that the proposed architecture significantly enhances the sharing and scheduling of computing resources among IoT de-vices. It elevates the overall efficiency of IoT computing while concurrently reducing energy consumption,thereby of-fering a robust solution to the evolving demands of IoT systems.
万方数据
维普
