Research And Implementation Of Data-driven Self-adaptive Monitoring Method

With the rapid development of software development and Internet technology,software systems have been widely used and people's dependence and requirements on software systems are becoming higher and higher,which makes the software scale show an exponential growth trend.The dynamically developed operating environment and changing user requirements make software systems need to adjust their structure or behavior during the running process.However,the huge software scale and the intricate relationships between heterogeneous software units make the process of artificially adjusting software behavior difficult,error prone,and costly.Therefore,adaptive software came into being,which can adjust its behavior and structure according to software changes.Generally,adaptive software guides the software system's adaptive adjustment process through adaptive control loops of Monitor,Analysis,Plan,and Execute.Among them,the perception link is responsible for acquiring software changes and serving as the input of the entire adaptive process,which directly affects the quality of the adaptive software.Therefore,the perception of efficiency and accuracy is a key and difficult issue in the field of adaptive software.The essence of the perception link is to collect system state data related to software changes in real time,so there are two difficulties in ensuring the efficiency and accuracy of perception On the one hand,the dynamic and open operating environment and heterogeneous system resources make the accuracy of the perception data It is difficult to guarantee that there will be missing data and noise data,which will affect the accuracy of software change recognition On the other hand,the heterogeneous software units and complex interactions of the software system make there exist a large number and diverse types of perception objects,and comprehensive perception will make The perception efficiency is low,and the processing,storage,and analysis of a large amount of generated perception data will bring huge system overhead and affect the accuracy and real-time of perception.However,most of the existing research work focuses on solving the latter through data processing and other methods,and less consideration is given to perceived objects.The main goal of this work is to design and implement a data-driven adaptive sensing method to address the above issues,while ensuring the accuracy and efficiency of sensing while reducing system overhead,and by mining and analyzing operational data,achieving adaptive sensing.Method feedback optimization.The main work of this article includes the following five parts:First,according to the structure and operating characteristics of current complex software systems,establish a perception index system to clearly identify the objects,and analyze the types and characteristics of the objects to establish a hierarchical perception framework to ensure the efficiency and accuracy of perception Sex.Second,introduce complex networks to model software systems,analyze key nodes based on degree-centricity methods for key monitoring,and analyze associations of perceived objects based on operational data analysis.At the same time,consider user needs and perceived costs to establish multi-factor analysis methods to screen and Refined perception objects achieve targeted perception.Third,in response to the problem that the fixed-period sensing method cannot sense changes in real time and generate a large amount of redundant data,a data feedback-based sensing period correction method is run to dynamically modify the sensing period,and considers the changes in sensing needs in different scenarios.Data analysis of the sensing object update method runs in real time to update the sensing object to improve the accuracy of sensing and reduce system overhead.Fourth,an adaptive sensing engine is designed and implemented to support the practical use of the aforementioned targeted sensing and feedback optimization methods.Fifth,the adaptive sensing method researched and implemented in this paper is verified by experiments,which proves the effectiveness of the proposed method.