Research And Implementation Of Self-adaptation Planning Optimization Mechanism Based On Spark Parallel Search

With the rapid development of computer technology,the scale of software systems is also increasing and tends to be complex,which also leads to artificially adjusted software behavior becomes more difficult,error-prone,time-consuming,effort and cost,and lack of robustness of software.The software system in the Internet environment needs to deal with complex software changes and continuously meet the needs of users.Therefore,the software system needs to be adaptive in the running process,which can dynamically adjust its behavior,attributes,structure,etc.during the operation process to adapt to changes in the environment or user needs,and improving the adaptability and extend the use of software.The Self-adaptive systems realizes the adaptive control of the software by establishing a selfadaptive control loop(MAPE),in which the Self-adaptation planning link is mainly responsible for generating an adaptive strategy.Therefore,the ability of Self-adaptation planning directly affects the quality attributes and external services of Self-adaptive systems,and is a hot and difficult problem in the field of Self-adaptive systems.The essential definition of Self-adaptation planning is to take change as a goal,and find or generate the adaptive strategy that best meets the goal from several alternatives.Therefore,it is not only a problem of state transition but also an optimization problem.From the optimization point of view,it will be possible to establish a new Self-adaptation planning method.Therefore,my research group combines the search-based software engineering with the Self-adaptive systems.At the same time,according to the advantages of the genetic algorithm in global search and scalability,the genetic algorithm is used to implement software Self-adaptation planning.According to the idea of search-based software engineering,different software changes can be modeled as different optimization goals.In the running process,the software system can select the optimal adaptive strategy in the solution space through the search genetic algorithm to adjust its behavior,which is suitable for software Self-adaptation planning.However,genetic algorithms have the disadvantages of poor search performance and large time overhead in implementing Self-adaptation planning,which directly affects the efficiency of Self-adaptation planning.Therefore,how to improve the efficiency of Self-adaptation planning based on search and reduce the time for Self-adaptation planning is extremely important.This paper aims at solving the efficiency problems in the process of Self-adaptation planning based on search-based software engineering,and researches a Self-adaptation planning optimization mechanism based on parallel search of Spark to improve the efficiency of selfadaptive decision-making based on search and ensure the time-effectiveness of self-adaptive adjustment.Firstly,a Self-adaptation planning optimization framework based on Spark parallel search is established.The search-based Self-adaptation planning optimization process is described.It guides the optimization process establishment.Then it analyzes the Spark platform resource scheduling methods and processes,and combines software adaptation requirements with Self-adaptive task features,designing a dynamic task scheduling strategy based on Spark.Then,based on the natural parallelism of genetic algorithm,a parallel coarse particle-size parallel genetic algorithm based on Spark is used to realize the parallel execution of adaptive tasks.The parallel search process based on adaptive tasks is designed.At the same time,from the realization point of view,an adaptive decision optimization center is established,and the concrete implementation of the adaptive decisionmaking optimization mechanism based on Spark parallel search proposed in this paper is explained.Finally,a prototype system was set up.Experiments were performed by using Book Store as a target system and the test results were analyzed to verify the effectiveness of the above mechanism.In the end,summarizing all the work of this paper,at the same time,it also looks forward to some of the directions that need to be improved and in-depth research.