Performance Evaluation Technology For OpenStack Based Cloud Platform

Intelligentization has become a new development direction of high-speed railway information systems with the synergetic development of big data technology and artificial intelligence applied in China high-speed railway.More and more intelligent application and devices have been access to cloud computing platform.It enables cloud computing platform to become the basics of intelligent services.Recently,private cloud has developed rapidly which lays the foundation for its implementation in high-speed railway.There exit many difficulties in the migration of railway systems including accurate evaluation for the performance of private cloud on railway and determination of appropriate baseline standard.Due to the specific characteristics of railway service,like real time computing,massive data through put.Exiting performance evaluation techniques for cloud platform cannot be directly applied in railway private cloud systems.This paper aims to study a plan to evaluate performance of railway private cloud platform and verify the proposed theory through experimental results.To solve the problem that cloud platform's performance will decrease due to the overload of physical nodes,we propose a load balance solution to improve the performance of load balance and service stability of railway private cloud.Firstly,we analysis the architecture and characteristic of railway private cloud system and propose a three-layer test model according to the needs of railway service including virtualization performance test in the bottom layer,ability test to resource management in the middle layer and ability test to service delivery in the top layer.We also set test items,index and tools for each layer.In addition,we quantify all the test indexes and determine the theoretical value on best performance.Secondly,we build OpenStack based private cloud test environment on Liaoning Section of Beijing-Shenyang high-speed railway test line.We compare the performance test results of railway private cloud platform and Alibaba cloud platform.The result shows that the performance of railway private cloud platform can meet the requirement of railway service.So we can validate the correctness and feasibility of our plan.We also proposed the baseline standard for railway private cloud performance.Finally,we propose a back propagation neural network based loading balance algorithm to deal with the decreased performance with overloaded physical machines which found in the test.We perform contrast experiment in the realistic scenarios,we found that the load of the physical machine is significantly reduced and the service response time is kept within the required range after the application of load balance algorithm.So we can verify the great improvement of platform performance stability utilizing the loading balance algorithm which indicates the feasibility of the proposed scheme.