Analysis of User Behavior in Cloud Broker

Cloud broker, which procures cloud services from multiple cloud providers and resells them to users, has been put forward to facilitate the delivery of cloud resources. However, user behaviors are dynamic over time. Hence it is challenging for cloud broker to schedule cloud resource in a suitable way. In this paper, we conduct a study on user behaviors to help cloud broker to make scheduling decisions. We choose job arrival rate and job service time as the primary measurements of user behaviors. We assume that job arrival rate is comparable to the Poisson distribution, while job service time conforms to the Pareto distribution. Through a detailed study of Google trace, we confirm that our assumptions are valid, and job arrival rate and job service time are qualitatively matching the Poisson distribution and Pareto distribution, respectively. Overall, we can consider these characteristics to promote cloud resources scheduling.