Research And Implementation Of Highly Scalable Microkernel Rtos For Low-end IoT Devices

Real-time operating systems have the advantages of high customization,high real-time,and low resource consumption,and are widely used in smart homes,smart agriculture,smart cities,and other fields.Microkernel is to isolate the operating system's file system,network protocol stack and other components outside the kernel,the kernel only retains task scheduling,memory management,inter-process communication and other components,which can improve the scalability,flexibility and stability of the device and provide a modular design.With the continuous development of the Internet of Things(IoT),the shortcomings of the traditional RTOS are also emerging,making it difficult to use in applications that require high scalability and flexibility(e.g.,online upgrade of IoT device programs,fast loading of tasks in computing offload,etc.).This paper designs and implements a microkernel real-time operating system that supports dynamic loading and linking of software on processors without memory management units(MMU),which can solve the problem of low scalability and flexibility.The main innovations in this paper are as follows.1)A Dynamic Loading and Dynamic Linking(DL~2)method is proposed.The traditional DL~2technique requires the processor to support MMU,which requires high hardware resources of the system.In this paper,we design a DL~2technique that can run on devices with small memory and provides good support for fast loading of programs.The dynamically loaded files in this paper are 10.4%?68.4%of the size of traditional ELF loadable files compared to traditional ELF files,with almost no loss in execution performance.When this DL~2technique is applied to remote program upgrade,its upgrade speed is increased by 30.6 times in the test routines written.2)A Micro-kernel Real Time Operating System(MKRTOS)is designed.Based on the DL~2technology proposed in this paper,MKRTOS is designed and implemented with a highly simplified kernel function,system components can be loaded dynamically,memory reuse and fast boot-up are achieved by DL~2technology with the advantage of micro-kernel,and system stability is improved by modular development.The system has reduced the boot time by 57.59%,52.55%and 47.59%in the boot time tests of the three functional sets,respectively.3)An Edge Micro-kernel Real Time Operating System(EMKRTOS)is designed.At present,there are few operating systems for IoT that support fast loading and unloading of tasks for edge computing,and most studies focus on offloading tasks to edge servers with higher computational power.For this reason,this paper designs EMKRTOS based on MKRTOS,which can achieve fast task loading and unloading using DL~2technology,enabling IoT devices to offload computing tasks to other IoT devices for acceleration,thus realizing collaborative computing among devices.In the scenarios of video decoding,video coding,and large prime number generation,the improved computation model and auction algorithm are used to allocate computational tasks.In the simulation experiments,the acceleration ratios are 816.11%,884.28%,and 834.87%.In the experiments using hardware devices for real scenarios,the speedup ratios are 41.01%,94.14%,and2511.08%.The microkernel real-time operating system designed and implemented in this paper is very suitable for applications with less hardware resources.The advantages of the microkernel make the system highly modular and improve the robustness of the system,and the program can be loaded and executed as a dynamic library,which improves the flexibility and scalability of the system.