Research On Measuring Performance For Cloud File Systems

With the rapid development of cloud computing,the file system in the cloud environment plays a more and more important role in cloud infrastructure.At present,the industry produces a lot of performance testing tools for cloud file systems.However,most of these performance testing tools focus on traditional indicators such as CPU,memory,network bandwidth,IOPS and throughput.These traditional indicators cannot meet the requirements of cloud users and providers.Because of the diversity and complexity of I/O loads from the top applications,heterogeneous and dynamic characteristics are highlighted and proposed more stringent requirements for cloud file systems.There are increasing users who need to know system performance like scalability,stability,availability,and isolation between applications.The existing evaluation tools and metrics model can not fully meet the needs of cloud users when to evaluate and measure the new type of performance indicators in cloud environments.How to help users to select the optimal cloud platform or cloud file system and provide the performance comparison and optimization of cloud file system accurately for the user has become a hot topic of current research.With the fast-growing user data and applications,how to assess and measure the degree of performance interference between each application to improve the quality of services for users in cloud file systems and how to efficiently expand the scale of the system are becoming more and more important.The isolation measurement model can help users quantify the isolation of target systems conveniently.Then,the model can instruct users to optimize isolation performance of systems and select the best isolation scheme to improve the system.Scalability measurement model is a multi-indexes based on radar chart to analysis scalability of target systems for cloud file system,the model of extended performance can evaluate comprehensively cloud file system under the kinds of load scenarios for the different extended resource indexes.Based on the load characteristics of cloud file system as the test basis,to make the test scenes and the real scenes without deviation,this paper proposes two general measurement models,which are taken as examples,to quantify the performance isolation between mulit-tenants and scalability respectively.With performance test data from the bottom,the author establish measurement model for the characteristic index of the upper clouds,which can accurately measure cloud file systems and provide valuable data for cloud users.At last,the validity,usability and accuracy of these two models are verified by experiments.It is hoped that the two models in this paper can be used to help other researchers to further study the performance evaluation and measurement of cloud file systems.