Analytical Model For Priority-Aware Heterogeneous Cloud Services

IaaS (Infrastructure-as-a-Service) is the bottom model in the three-tier cloud service model. IaaS uses the underlying infrastructure (computing, storage, network, etc) to serve the consumers , such as the virtual service typically. IaaS often manages and configures computing resources by structuring virtual machines. SLA (Service Level Agreements) is a contract which helps IaaS cloud service providers and the consumers to determine the parameters such as QoS (Quality of Service). A better SLA can help cloud service providers to seize the initiative in the cloud computing market. Therefore,how to evaluate the performance of cloud services is a very challenging and complex task. The numerical analysis model is an effective performance analysis tool. The characteristics of large scale cloud lead to the necessity of building scalable model. At present, researchers have proposed a scalable and interactive analysis model to reflect the cloud center operation. The accuracy of the stand-alone model is the premise of the accuracy of the interactive model. Some researchers utilized the existing and mature queuing model directly. However, these models have some huge limitations. To make it worse, many results of existing cloud service model do not take the heterogeneity of the resources and the priority of the task into account.This paper aims to study the stand-alone cloud service model which contains heterogeneous resources and tasks with priority. In addition, the model is supposed to be integrated with the existing interactive analysis model seamlessly to reflect the cloud center behavior.This paper introduces the cloud computing, queuing system, Markov chain and other aspects of the background knowledge first. Through the actual operation of the cloud center, we take the heterogeneity of the resources and the priority of the task into account.Then we use the M /M /m/m + q1,q2 queuing system to describe the active physical machine in the IaaS. Above all, we choose the continuous Markov chain to construct the numerical analysis model for the system.We have a detailed introduction of system behavior, relevant assumptions and variable formula. Then we use the transfer rule to derive the transfer matrix of the model in detail and satisfy an important condition that the sum of steady-state probability is 1.Finally we obtain the steady-state probability of the system accurately. Then we obtain a series of performance through the derivation formula of the performance index. Such as the average queue length,the average response time of the tasks and the probability of the service being rejected.Then we use the method of Monte Carlo to design simulation for an active and heterogeneous single physical machine systems of cloud services. The experiment is highly consistent with the actual IaaS cloud service. We design the simulator by assuming that the task has priority and the number of requests for tasks is heterogeneous. The design of the simulator of the cloud service model and the calculation method of the performance index are introduced in detail. Then we obtain the accurate performance index of the IaaS cloud service system which working in a long enough working time.Finally, the accuracy of the model is verified by comparing the results of simulation experiment and numerical analysis model. The experimental results show that our model can reflect the operation of multi-core single server system accurately.