Research And Implementation Of "Cloud-Edge-Device" Integrated Rapid Development For IoT Applications

The recent years have witnessed the rapid growth of IoT(Internet of Things)technology.It is predicted that the number of IoT devices in the world will grow to 125 billion by 2030,and the IoT trend will affect almost every stage and field of the entire industry.For the past few years,not only a large number of domestic and foreign scholars have conducted extensive research in IoT,but also domestic and foreign top Internet companies and manufacturers have already entered the IoT industry.It helps to bring IoT technology into all aspects of manufacturing and daily lives.Due to increasingly fragmented IoT scenarios and IoT development techniques,it is of great difficulty to develop IoT applications.Currently,there exists three major challenges: 1)The development chain of IoT application is long.It involves cloud,client(i.e.,smartphone)and device,resulting in significant development difficulty and long development cycle;2)Hardware platform design for IoT devices is difficult.Developers need to consider suitable hardware components,their compatibility,and the co-design of the software and the hardware,which makes the development of IoT devices difficult;3)IoT devices lack computation capability for dealing with complex signal processing,e.g.,image recognition and speech recognition,etc.It rises the problem of high computation latency and high energy consumption.If we can rapidly develop integrated IoT applications and integrate cloud,client,device,and edge-side development technologies,we can perform rapid prototyping,gain first-mover advantage,reduce development costs,and improve application performance.To this end,this paper studies key techniques of rapid development for IoT applications towards integration of cloud,edge and device.The implemented system can cope with the aforementioned challenges.(1)Integrating cloud,client and device development for IoT applications.A typical IoT application usually contains the cloud,the client,and the device side,which involves deep interactions between the hardware and the software.We propose an integrated rapid development approach and implement the system which can cope with the first challenge.Developers use the integrated programming language to easily express diverse IoT application logic for the cloud,the client,and the device side,which greatly accelerates the development process.In this paper,we formulate the problem of automatically generating task schedules of IoT devices from the application logic into an optimization problem,and propose a heuristic algorithm to minimize the energy consumption of all task schedules.We evaluate the performance from multiple perspectives by using case studies,user studies,and benchmarks.The results show that the technique can greatly accelerate the development process,while maintaining the expressiveness and reducing energy consumptions of sensing and transmission.(2)Construction of IoT device application.We propose a device application construction technique and implement the system that can cope with the second challenge.It changes the original bottom-up development model and adopts a top-down development model.Developers no longer need to deal with the underlying hardware details.They just write hardware-independent application code.Taking the code as input,the implemented system automatically builds the IoT device application,including the hardware configuration and the software program,which greatly accelerates the development process.This technique has been evaluated by using benchmarks,detailed case studies and user studies.The results show that it realizes the rapid development of IoT device applications,and generates optimal solutions of hardware platforms,as well as the software programs for the target platforms,while incurring acceptable overhead.(3)Edge-side computation offloading.We propose an efficient computation offloading technique and implement the system that can cope with the third challenge.Complex event processing of IoT devices can be automatically offloaded to nearby edge nodes or the cloud transparently.Developers need not write notations in the application code.This can reduce the computation time and improve the application responsiveness.By considering the dynamic workloads and the multi-thread execution capability of edge nodes,it can schedule multiple processing tasks more efficiently than the state-of-theart.Experimental results show that it reduces the total offloading latency of multiple tasks under dynamic workloads of edge nodes,while incurring acceptable overhead.