Research On The Data-Driven Method To Promote The Efficient Utilization Of Network Resources In Cloud Service

Because of the unique charm,cloud computing has become the focus of the industry."Internet plus initiative" impels enterprises to migrate into the "cloud".At the same time,with the rapid development of live online,e-commerce video and online games,users pay more attention to the service quality of experience.These trends have brought challenges to cloud service providers.How to efficiently use network resources to provide high quality services for more and more users,which has become an urgent problem to be solved.At present,cloud service providers mainly improve the utilization of network resources within their system,and lack market means to guide the efficient utilization of network resources.For this purpose,we propose a way to solve the above problem by means of market information feedback mechanism,and the mechanism has been implemented in the CDN(Content Delivery Network)industry.The key point of the method is to provide information services for cloud service providers and users.For this reason,firstly,a data-driven information service platform needs to be established and supports the operation of market mechanism.Secondly,network basic information,such as,network topology and basic information of website,will be obtained to provide various network information services through the platform.Finally,industry data is collected and analyzed to get industry information to provide various industry information services and assist all parties in the industry to choose.The selected results are regarded as feedback and provided to all parties in the industry again for promoting the optimal allocation of network resources.The main research contents and contributions are as follows:1.Research on the data-driven network information service platform.Based on the idea of big data,first of all,we put forward the design goal and structure of the platform.Secondly,we elaborate the difficult problems in the construction of the platform and give the corresponding solutions.Finally,we establish the network information service platform,which could be seen as a support platform for the implementation of market information feedback mechanism.2.Research on the method for acquiring information of router-level network topology.The method for obtaining router-level network topology information based on incremental measurement mechanism is proposed.The IP alias resolution algorithm based on analytical techniques has been improved from three aspects: subnet segmentation computing,verification of IP alias pair and the extraction of IP.The goal of getting more complete and accurate router-level network topology information under the premise of cost control is realized.Through the method,we get the router-level network topology information in China.3.Research on the CDN recommendation method based on the neural network.The CDN recommendation method based on the neural network is proposed and implemented.First,a set of data acquisition and analysis method for CDN service is established.The domestic CDN information is obtained from three aspects: network,performance and market share,and displayed by visual analysis method.Secondly,the method of extracting ICPs'(Internet Content Provider)keywords by computing the weight of words is proposed.These keywords are used to match ICPs by type.Finally,based on the ICPs' keywords and CDN basic information,we use the neural network to establish a CDN recommendation model,improve ICPs' efficiency in selecting CDN service resources and accumulate experience in information recommendation services for other cloud services industry.