Research On The Evaluation Methods Of Elastic Cloud Resource Scheduling

Nowadays, there are many kinds of cloud computing products, and they have been used by many people. Users can be both individuals as well as enterprises. How to choose a service that meets the performance requirements while maintaining the cost of cloud services become a problem plagued users of cloud services. To solve the problem, we need make an evaluation on cloud computing. According to the evaluation results, users can choose their own cloud services.Elasticity is one of the core aspects of cloud computing. When the workloads are overwhelming the ability of current system, the system can provision resources to deal with the situation in an autonomic way. On the other hand, when the ability of current system is redundant to the current workload, the system can deprovision resources to avoid wasting system resource. Thus the evaluation of elasticity of cloud computing system contributes to the choice of cloud services.This paper studies the method of evaluating the elasticity of cloud computing system. According the studies of the core of elasticity of the cloud computing sytem, we know that the evaluation to elasticity of cloud computing system is a multi-angle multi-objective problem. But the current mainstream testing methods often give the result on a certain view. Thus, this paper proposes an evaluation method of elasticity which can give comprehensive results based on AHP. The method include four main parts: workload parts, evaluation metrics and evaluation model, data collecting parts, data analysis parts. According the real scenarios, we design three types of workload in order to simulate the real scenarios.This paper also design and implement the under test system. The paper also design two kinds of system resource scheduling strategies to verify the evaluation metrics and model, and implement them in the system under test. Load Runner implement the workload on the system under test. Monitoring module collect the data which generated during the test process. Using evaluation metrics and model analysis the data to give out comprehensive result. Users can make their decision based on the result, and the provider can draft different service level according to the result.