Research On Intelligent Online Experimental System Based On Cloud-Edge Collaboration

With the continuous rise of the digital transformation of teaching in colleges and universities,the use of online experimental systems as the teaching platform of C programming language courses has also become a new trend.However,such systems still have some problems in the actual teaching process.First of all,the system mostly adopts a single cloud computing center processing mode.With the rapid increase in the number of people online in the system,the disadvantages of this mode gradually appear.For example,system function feedback is no longer timely,data processing delay is high,and hardware resource utilization is low.Secondly,most systems are not very intelligent,which is manifested in the low level of automation in the grading process,students’ programming question assessment and program grading cannot be carried out continuously,and the test results still need to be printed out and manually reviewed by the teacher.In order to solve the problems of prolonged task response time,low resource utilization,and high load of cloud servers in the single cloud computing center mode of the system,this paper uses the cloud-side collaborative task processing mode to replace the original mode.The cloud-side collaboration model mainly introduces edge computing（Edge Computing）technology,which provides network,storage and computing services nearby on the side close to the data source.Because the geographic location of the edge computing platform is closer to the terminal device,it can greatly shorten the execution delay of the task and promote the efficient use of resources.At the same time,this paper proposes a joint resource deployment and task scheduling scheme in a cloud-side collaborative environment,aiming to reasonably schedule some computing tasks in the central cloud server to the edge server for execution,so as to shorten the overall task processing delay and improve resource utilization and ease the high load of cloud servers.The main research contents of this paper are as follows:（1）Joint resource deployment and task scheduling algorithm based on task prediction.According to the long-term monitoring data of the experimental system,it can be known that there is a strong correlation between the time period and the change of user tasks.At the same time,the cloud computing center server has relatively sufficient computing power.Therefore,we can use the massive computing power of the central cloud to predict the changing trend of the number and types of user tasks.In this paper,a joint resource deployment and task scheduling algorithm based on task prediction is proposed.First,the central cloud server predicts the changing trend of user tasks through a two-dimensional time series based on periodic statistics and trend analysis;then,the central cloud server further guides the edge based on the predicted results and the resources required for task operation The server deploys its hardware resources;finally,the cloud server reasonably schedules user tasks to the edge server for execution according to the task resource deployment set returned by the edge server.The final experimental results show that the algorithm can effectively reduce the task processing delay and improve the overall performance of the system.（2）An automatic scoring scheme for programming questions based on the combination of dynamic detection and static analysis is proposed.This program focuses on the lexical analysis operations,syntax analysis operations,abstract syntax tree construction methods of the GCC compiler in the compilation process,as well as the intermediate code conversion form of the source program and the node traversal method;using GCC based on the keyword Trie tree The abstract syntax tree eliminates the redundancy algorithm to optimize the AST;formulates reasonable abstract syntax tree standardization rules,mainly applicable to the AST standardization of the selection structure and cyclic structure;uses the KMP algorithm to perform keyword matching in the dynamic detection stage;uses the based on the static analysis stage The tree edit distance algorithm of the weight of the node matches the control structure subtree,calculates the similarity,and gives a comprehensive score result.（3）The design and realization of the experimental system.The experimental system design part and concrete realization are introduced in detail.The system structure design part mainly explains the design of functional modules and the design of database.The main functional modules of this system include user login module（management end user,student end user,and teacher end user）,experimental system management module,test taker answer module,system competition module,and program automatic scoring module.In Chapter 5,these modules will be summarized in turn and the corresponding implementation interfaces will be shown.In the automatic scoring part of the experimental system,the implementation of the scoring module and the internal scoring steps of the system will be emphatically explained,and the comprehensive scoring results of the experimental system will be displayed.