Development Of Parts System Based On Cloud Platform And Research Of Mass Small File Storage

With the wide application of cloud computing,cloud services have become a main alternative in IT infrastructure and applications.The flexible delivery model,on-demand selfservice,and broad network access have changed the business model of traditional IT services tremendously.The company focuses on business only and customizes personalized services on demand.This thesis is a combination of The National Science and Technology Support Plan Project and Application of Technology with Business Cloud Platform(2015BAH19F02),which aims at adjusting quick response of marketing service to sales services of automobile and auto parts industry supply chain and,developing and building a cloud-based auto parts sales system,which can provide integrated services to Auto Part Dealers,4S Shops and Service Providers.With the help of different business modules such as basic data,parts procurement,parts sales,and parts inventory on cloud platform,the manual operations in business management can be reduced,work efficiency can be improved,and some business goals,which are overall planning,resource sharing,and cost savings,can be achieved as well.This platform proposes a storage scheme that is based on Hadoop for the massive pictures generated by the parts sales system of the cloud platform.The scheme integrates the technologies of classifying,merging,retrieving and distributed storage of massive pictures.The content details of the thesis are as follow:(1)Analyzing and understanding the concepts of cloud computing,cloud services,cloud platforms and their features.Focusing on the cloud computing delivery model,multi-tenant architecture and it's storage mode implementation methods and application scenarios.Analyzing the business demand of the Parts Dealers,4S Shops,and Service Providers to understand the needs of the system's functions,behaviors,performance,and design.(2)Using the above business demand and combining the characteristics of the cloud platform,the overall solution of the system is provided and the functional structure of the system is designed.Using Multi-tenant data storage model to design data structures to enable data sharing and isolation between tenants.Analyzing the business relationship between different modules such as basic data,procurement,sales,inventory,designing and developing an auto part sales system based on cloud platform.(3)The parts sales system that is based on cloud platform oriented towards massive tenants and it is bound to generate a large amount of data,including structured data(i.e,order data)and unstructured data(including pictures,logs,bills,statements,etc.).For massive data,different kind of methods,which are hybrid data storage model,storing order data with relational database,saving massive unstructured data with distributed file system,are used.This thesis focused on Hadoop's distributed framework and storage technologies.(4)According to the requirements of the system for storing massive pictures,this thesis proposed a massive picture storage scheme based on Hadoop,which consists of three methods: picture classifying,merging,and retrieving.For grouping of massive pictures,the principle of access time and space locality is used to store different types of pictures in different groups.For merging massive pictures,this thesis provides a method called Local Indexed SequenceFile(LISF).Comparing the LISF with the Hadoop's Sequence File and MapFile methods,it does not only guarantees the lowest memory consumption of NameNode,but also improves file access efficiently.For retrieval of massive pictures,the picture Index File is used to search for pictures from the merged file.Retrieval method involves random reading of pictures.Finally,using the picture classifying,merging,and retrieval methods enabled the implementation of massive picture storage in the auto parts sales system.