Research On Generation Method Of Cloud Resource Elasticity Testing Based On Iterative Feedback

With the development of cloud computing technology,further improving the efficiency of resource allocation is significant to cloud resource providers and users.However,the current way of predicting users' future demand through the model needs to allocate a fixed amount of resources in advance,which is easy to cause idle and waste of resources during off-peak periods.The way of dynamically allocating resources by using the elasticity service of the cloud platform,optimizing the allocation strategy of cloud resources through algorithms is mainly implemented from the perspective of cloud platform providers.Users still have difficulties in how to reasonably allocate resources.Therefore,this thesis aims to screen the rule-set that can achieve the best elasticity level by means of elasticity testing,and provides users with cloud resource allocation strategies according to the rule-set.Thus,achieving the goal of improving cloud resource allocation efficiency.At present,the generation of test cases in elasticity tests is based on empirical values or random selection.And the generated test cases usually cannot represent the optimal elasticity level of the cloud platform.So,they cannot be used as the basis for the cloud resource allocation strategy.Therefore,this thesis improves the elasticity testing process and proposes an optimal generation method for elasticity testing.The generation of test cases drives the iterative execution of elasticity testing.Finally,the best rule-set is selected from the generated test cases.According to the improved testing process,this thesis designs relevant evaluation metrics to evaluate the elasticity level.And extracts the important attributes of elasticity as a rule-set.Then the genetic algorithm idea is introduced to optimize the test case generation method.The initial population is designed to perform the test,and the new population is generated according to the iterative feedback results to guide the heredity and mutation.After several iterations,all test cases are generated.Experimental results show that the elasticity testing generation method proposed in this thesis can obtain test cases with high elasticity level after a few iterations.Its elasticity level is obviously better than the initial test cases set according to empirical values and is better than the average level of all test cases.The rule-set represented by the final best test case sets up cloud resource allocation policies,which can significantly improve the elasticity level of the cloud platform and improve the allocation efficiency of cloud resources.