Research On Strategy Execution Mechanism For Self-Adaptive System

With the deep application of software technology in the real world,the scale and complexity of software is showing an ever-increasing trend,and the operating environment is also characterized by openness,dynamics,and changeable.At the same time,the user needs also continue to evolve.Therefore,the self-adaptive software has become a core research issues in the field of software engineering.Self-adaptive software cope with the above changes from user needs and the environment by constructing adaptive control loops,and adjusts software behavior and structure at run-time.Among them,the process of executing adaptive strategy is an important part of implementing software adaptation,and it is an important step who truly implement the adjustment of software behavior and structure in accordance with the planned strategy.However,most of the existing adaptive researches focus on the other three aspects of perception,analysis and decision-making,and the research on adaptive strategy execution is weak.In addition,existing methods lack the corresponding tools to support the adjustment to system structure and behavior at runtime,practicality is poor.Moreover,the adaptive strategy is planned by software,and bad strategy is possible.However,most of the existing adaptive policy execution mechanisms are assuming that the strategy it performs is safe and reliable,without analyzing the impact of the strategy on the reliability of the target system,therefore,it is difficult to guarantee the reliability of the adaptive software.For these problems,this thesis starts from the two aspects of self-adaptive adjustment mechanism and strategy reliability analysis.First of all,in order to support the execution of self-adaptive strategy and reliability analysis of the strategy,we analyze the classification of software connections,selects the data flow and process call as the structure adjustment object,and use the idea of service process modeling to establish a runtime model,describing the structure and parameter status of the target system.Then,a self-adaptive software parameter and structure adjustment mechanism is proposed.The adaptive strategy definition and software parameter extraction method are given.And the corresponding workflow is designed for the adjustment of structure and behavior.Finally,a self-adaptive software reliability analysis method based on Markov chain is proposed,it verifies the influence of the strategy on the reliability of the target system before the strategy is executed which could discover the bad self-adaptive strategy is in time and avoid a substantial impact on the target system.The above models,mechanisms,and methods form the strategy execution mechanism for adaptive software proposed in this thesis.In order to realize the above implementation mechanism,effectively organize,apply and implement the above models,mechanisms and methods,and provide practical software tools for the strategy execution.we design and implement a strategy execution framework for self-adaptive software.It is a three-tier structure of "Public Facilities Layer,Multi-Agent System Layer,Target System Layer".Among them,two key components in the adaptive software parameter and structure adjustment mechanism,message middleware and parameter control center are implemented in the " Target System Layer " to realize the adjustment of the target system structure and parameters;in the "Multi-Agent System Layer",The establishment of a multi-agent system make the agent act as a strategy executor and translate the strategy into specific operations on the "target system layer";the "Public Facilities Layer" reliability verification center is built,applied the self-adaptive software reliability analysis method,provide the function of strategy analysis and verification.Finally,this thesis designs and implements a web application system Bookstore,and uses this system as a target system to verify the correctness and effectiveness of the above mechanisms and frameworks.