I/O Performance Isolation Modeling And Evaluation For Cloud Container System

In recent years,container technology has attracted intensive attention due to the features of light-weight and easy-portability.However,the performance isolation between containers is becoming a significant challenge,especially in terms of disk throughput I/O.Compared with virtual machines(VMs),containers suffer from a worse I/O performance isolation as they share not only physical resources but also OS kernels.A couple of solutions have been proposed to enhance the I/O performance isolation between containers.However,there is a lack of study on how to quantify the I/O performance isolation between containers.In traditional virtual machine environments,the I/O performance isolation is usually calculated based on the performance loss ratio.In the container environment,the amount of resources occupied by the container,will also affect the containers' I/O performance.Therefore,the loss of containers' I/O performance is not only due to the influence of overload containers,but also because the resources occupied by the containers themselves have declined.Therefore,the measurement models that only take performance loss into consideration will be not accurate enough,and the concept of resources is also required.At the same time,I/O performance loss rate is not effective in reflecting the I/O performance changes in the process of decline,and also makes a certain error when measuring the performance decline.In order to solve the above problems,this paper studies the I/O performance isolation in container environment and proposes a new I/O performance isolation measurement model.At the same time,the method of calculating the I/O performance and resources changes is optimized in this model.Finally,this paper implements a container I/O performance evaluation framework based on the model.The main contents of this article are the following aspects:(1)In this paper,we propose an I/O performance isolation measurement model that combines the performance loss and resource shrinkage of containers,measurement model will be effective resource and performance of container occupies the combination through the dynamic measurement of all parameters decreased.This model reduces the error metrics,combined with the container resource volatility features.We introduce the concept of resources the measurement results,and the isolation is more accurate and effective.We conduct a group of I/O performance evaluation experiments based on an open-source container project Docker.Experimental results validate the effectiveness of our proposed model.Compared with the traditional I/O performance degradation rate measurement model,the results show that our metric model can better reflect the isolation difference in container system,and the result is more accurate.(2)We implement an evaluation framework for I/O performance evaluation of container system at last,and realize the evaluation function of measuring container I/O performance isolation,which reduces the working pressure of researchers.This framework realizes batch management and task distribution of containers.By abstracting the testing process of I/O performance isolation measurement model,it increases the convenience of performing tasks in batches in a container environment.Finally,the validity and accuracy of the framework are verified through experiments.