Cocktail: A New Tool for Audit by Performance Evaluation on Local Machine

Cloud computing is committed to provide users with convenient, economic and high scalable computing resource. However, according to the delivery mode of "pay-as-you-go", how to evaluate the trustworthiness of the accounting of cloud computing service has become a vital problem need to be solved. From the perspective of the users, this paper proposed a novel user application performance evaluation tool called Cocktail. Cocktail includes application decomposition, static analysis and dynamic analysis of three modules. An application is divided into many evaluation units on the basis of its resource density. Based on this condition, through local small data set, we could quantitative estimate user application time overhead and resource overhead under the cloud environment with large data set. We evaluate the accuracy of Cocktail by doing with four classical algorithm programs, and the result shows that Cocktail has practical relevance.