智慧服务解析平台的设计与实现

发布时间：2018-04-12 13:54

本文选题：智慧协同网络 + 服务注册与解析　；参考：《北京交通大学》2016年硕士论文

【摘要】：随着互联网的发展,用户数量呈现爆发性增长,应用多样性不断增加,传统网络体系的弊端日益突显,如安全性差、移动性支持不足、资源利用率低、可扩展性不足等。传统互联网“三重绑定”的特性,以及相对“静态、僵化”的设计,是导致各种严重弊端的原始设计根源。针对这一根本原因,北京交通大学下一代互联网互联设备国家工程实验室提出了智慧协同网络体系架构。本文依托国家973项目“智慧协同网络理论基础研究”,设计并实现了智慧协同网络下的智慧服务解析平台,完成了服务注册与解析机制的功能实现。同时,针对大规模网络情况设计并实现了快速服务查询与解析方案。该方案对智慧协同网络原型系统的验证研制,以及该网络体系的正确性和可行性具有重要意义。首先,本文分析了当前互联网存在的问题与不足,阐述了智慧协同网络的设计思想以及“三层”“两域”总体框架模型,简要介绍了服务注册与解析机制的研究现状,说明了本文的主要工作以及论文的组织结构。其次,本文详细描述了智慧协同网络中的服务注册与解析机制,具体包括智慧协同网络服务注册机制、服务解析与转发机制以及包格式设计方案,并介绍了本文实现方案所用开发平台(OpenFlow、Click Router和DPDK)的相关技术概念。接着,本文根据智慧协同网络中服务注册与解析机制的设计,完成了域内和域间服务注册与解析机制的实现方案,重点阐述资源管理器和边界路由器两部分功能节点的模块设计与实现。然后,基于OpenFlow平台和Click平台完成了不同自治域的域内、域间服务注册与解析机制,并利用实验室试验平台,进行了相关功能验证测试。最后,本文针对大规模服务请求、海量注册信息的情况,设计了智慧服务解析平台的实现方案。在DPDK平台实现优化之后的高并发Cuckoo Hashing算法,并结合DPDK特性对算法进行大页内存、批处理及预处理等优化,提升数据包转发吞吐量。进行方案测试,包括裸转发测试、算法优化分析测试、表项数量影响测试以及CPU数量影响测试。测试结果表明,当服务标识数量为5亿,服务请求发包速度为网卡极限速度20.49Mpps时,智慧服务解析平台能够达到DPDK的数据包极限转发速度为18.30Mpps。
[Abstract]:With the development of the Internet, the number of users is increasing explosively, and the diversity of applications is increasing. The disadvantages of traditional network systems are increasingly prominent, such as poor security, insufficient mobility support, low utilization of resources, insufficient scalability and so on.The characteristic of "triple binding" of traditional Internet, and the relative "static, rigid" design, are the original design root of all kinds of serious malpractices.Aiming at this fundamental reason, the National Engineering Laboratory of next Generation Internet Interconnection equipment of Beijing Jiaotong University put forward the intelligent cooperative network architecture.Based on the national 973 project "the theoretical foundation of intelligent cooperative network", this paper designs and implements the intelligent service analysis platform under the intelligent collaborative network, and completes the function realization of the mechanism of service registration and analysis.At the same time, a fast service query and resolution scheme is designed and implemented for large scale network.This scheme is of great significance to the verification and development of intelligent collaborative network prototype system and the correctness and feasibility of the network system.First of all, this paper analyzes the problems and shortcomings of the current Internet, expounds the design idea of intelligent collaborative network and the general framework model of "three layers" and "two domains", and briefly introduces the research status of service registration and resolution mechanism.The main work of this paper and the organization structure of the paper are explained.Secondly, this paper describes the mechanism of service registration and parsing in intelligent collaborative network in detail, including the registration mechanism of intelligent collaborative network service, the mechanism of service resolution and forwarding, and the design scheme of packet format.The technical concepts of OpenFlowCick Router and DPDK are introduced.Then, according to the design of service registration and resolution mechanism in intelligent collaborative network, the implementation scheme of intra-domain and inter-domain service registration and resolution mechanism is completed.This paper focuses on the module design and implementation of resource manager and boundary router.Then, based on the OpenFlow platform and the Click platform, the service registration and resolution mechanism of different autonomous domains is completed, and the related function verification tests are carried out by using the laboratory test platform.Finally, this paper designs a solution of intelligent service parsing platform for large-scale service requests and massive registration information.The high concurrency Cuckoo Hashing algorithm is implemented on DPDK platform, and the algorithm is optimized for large page memory, batch processing and preprocessing combined with the DPDK characteristics to improve the throughput of packet forwarding.Test the scheme, including naked forwarding test, algorithm optimization analysis test, table item number impact test and CPU quantity impact test.The test results show that when the number of service identifiers is 500 million and the speed of service request outputting is 20.49Mpps, the intelligent service parsing platform can achieve the maximum forwarding speed of 18.30 Mppsof DPDK data packets.
【学位授予单位】：北京交通大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：TP393.09

【相似文献】