Research And Implementation Of Techniques To Improve The Performance Of Multi-pollint Tasks With Limited Resources

The proliferation of information technology has been spurred on by the advancement of computer hardware and the growth of the Internet.Nowadays,the 5G era has opened up a plethora of opportunities for the Internet of Things,and intelligent information systems have been designed and utilized to capitalize on these possibilities.Most intelligent information systems on the market today collect data through the physical device side,process the data initially in the intelligent information system,and then transfer the processed data to the network center and cloud computing cluster.Intelligent information systems often need to create multiple task threads to cope with the processing of different input data,and the demand for hardware resources increases as the number of task threads rises,thus increasing the hardware cost while ensuring system performance.In this thesis,we propose a new solution to improve the performance of multi-polling task systems in a limited resource environment.Three main research works will be conducted in this thesis as follows:(1)In terms of improving the system data processing speed,this thesis proposes a solution based on a concurrent thread model with user-state threads.The solution aims to solve the problem that the mainstream edge computing system is designed with a kernel-backed thread model,in which the thread creation and scheduling operations are done in the kernel state,and thus the concurrency performance is degraded by switching between the user state and the kernel state in the case of multiple threads.The solution also hopes to solve the problem of mismatch between the time complexity of task functions and the ratio of allocated CPU time slices brought about by the fair allocation of CPU time slices provided by the operating system for kernel threads in the kernel support thread model by performing thread regulation in the user state.In this thesis,we will verify the feasibility of the threading model proposed in this thesis by comparing experiments with the traditional threading model,and validate that it is beneficial to improve the computational performance of the system under restricted CPU conditions.(2)In terms of improving the speed of cache queries,this thesis proposes a solution for a memory-based storage model that can be applied to fuzzy queries.The storage model introduces a word splitter design to pre-process the input data,split the input data into words and retain the keyword relevance information of each word,assign a unique ID to the input data and then store it in an ordered ordered database to speed up the fuzzy query of the data.Experimentation comparing the storage model has demonstrated its capacity to accelerate fuzzy query of data,and its practicality and usability have been confirmed.(3)Based on the above two solutions,a video data-based traffic event monitoring platform is built in this thesis.The platform’s video data-based traffic event detection platform system is oriented towards vehicle-road cooperation and vehicle-city integration to analyze the status of traffic participants and provide auxiliary decision-making information needs for different traffic subjects.At the conclusion,the platform is evaluated for its primary functions and the produced data is shown to validate the viability of the two solutions mentioned above.