Design And Implementation Of A Private Cloud Scientific Research Management Platform Based On OpenStack

In recent years,many research institutes have questioned whether the public cloud platform can protect its internal business data.The data center architecture of the private cloud is inside the enterprise firewall,which is more secure than the public cloud.Therefore,how to quickly and conveniently deploy a highly reliable private cloud research management platform has become an urgent problem to be solved.This paper plans to deploy a private cloud platform for the college’s research management needs.Through this platform,the infrastructure of college can be uniformly deployed,and experimental data and related documents can be controlled to ensure the security of the internal information and resources of the college.The paper proposes practical solutions to the problems of the current OpenStack,such us complex deployment,low availability,single scheduling algorithm and slow transmission speed of file.The main research results obtained are as follows:(1)According to the specific needs of scientific research projects,three functions,including user management,cloud hosting and cloud storage,as well as performance requirements such as reliability,security and scalability are proposed.Based on the analysis of two popular cloud platforms and the combination of scientific research needs,OpenStack is selected as the underlying cloud platform.Aiming at the difficult problems of OpenStack service and complex deployment,a distributed deployment scheme of three control nodes +four computing nodes + three storage nodes is designed,and various functions are realized based on the scheme.(2)In view of the low availability of OpenStack,the high-availability solutions popular in recent years at home and abroad are deeply studied,and the Pacemaker+Corosync+Haproxy+Swift solution is designed.Through verification,when a single point of failure occurs in the control node cluster under the scheme,the service on the node can be quickly transferred to the standby node.This solution improves the system’s high availability.(3)For the shortcomings of the built-in scheduling algorithm of cloud host module,the popular scheduling algorithm is compared and the improved multi-objective scheduling strategy based on ant colony algorithm is proposed.Through verification,the improved algorithm can improve the resource equalization rate of the server while ensuring high resource utilization.(4)For the problem that the cloud storage module file transfer rate is slow,the solution of segment transmission is proposed.The whole file is divided into multiple small files for transmission.The results show that the fragment upload is 9.3% faster than the overall upload,and the clip download is 29.1% faster than the overall download.Based on the above design and optimization scheme,the scientific research management platform was successfully realized,and the actual results were achieved in small-scale laboratories.