The Research Of Iaas Cloud Service Modeling Method Based On Fault Tolerant Aware

As an emerging technology in the world, Cloud computing is in rapid development stage. With the development of cloud computing, the demands from the users are increasing more than ever, more and more users require cloud computing center to provide quality assurance services. Being an important part of cloud computing, IaaS distributes the computer resources to the users according to their needs. Therefore, it is been an important significance to the cloud service provider as well as the users, about how to effectively accurate assessment of the IaaS service quality. However IaaS cloud computing center is very complex, the service performance is affected by many factors, such as physical machine performance, the number of physical machines, available capacity, the characteristics of virtual infrastructure, scheduling policy and so on, so accurately assessing the cloud IaaS service quality is a very challenging job. Along with the development of the queuing theory, the technology has been applied to computer network modeling research, become an indispensable part of research in the performance of cloud computing.Google Cluster Dataset data shows that 64% of the tasks is a single destination, so the modeling and analysis based on a single destination task is of great significance. Building the model by means of simulating the task, queuing, services and leave with computer simulation technology, and then analyze the performance of the queuing system. Based on IaaS cloud’s single arrival task this paper proposed two analytical models for different scheduling strategies, system modeling was carried out and constructed the corresponding state transition matrix according to the effective state transition event of each model, through the state transition matrix can obtain the steady-state probability of analytical model. Because there is a corresponding relationship between steady-state probability and the system state, we could obtain the probability of denial of service, the average response time and so on a series of key performance indicators, so as to accurately assess the quality of IaaS services. The experimental results show that the usability of the two kinds of analytical model, finally, comparing to two kinds of analytical model, the advantages and disadvantages of each model are discussed by the experimental results.This paper mainly does the following work for the analysis model.(1) Due to the complexity of cloud computing center, many models haven’t been included physical machine migration, but in this paper, the model takes it into account and combines it with fault tolerance, then construct the whole model;(2) The whole model was introduced in detail on this paper, two analytical models for different scheduling strategies are constructed, it summarizes the effective state transition event of each model, gives a detailed explanation about effective state transition event, combined with effective state transition event model for modeling and analysis;(3) Because of the state explosion, this model building by the numerical analytical method only can describe small-scale cloud computing system. In order to be able to make the model describe the cloud computing system of medium-sized and above, Petri network is proposed to solve this problem, and constructs the analytical model based on it, so it can describe the cloud computing system of medium size and above, the experimental results show that analytical model that based on Petri network has better practicability.