IE网络测速的实现及操作

发布时间：2018-03-18 12:08

本文选题：流量生成　切入点：软硬件协同　出处：《湖北工业大学》2017年硕士论文　论文类型：学位论文

【摘要】：网络测试又是保证整个网络系统的设备与功能以及运行状态的手段。在网络测试中,流量生成是至关重要的一项技术,结果的正确性误差程度均决定于此。网络技术不断发展,日新月异,网络数据交换速率越来越快,对精度的要求越来越高,流量越来越大,对于一个万兆级的网络来说,现在已有的测试方法都难以满足需要。所以,面对新形势下的新需求,研究针对以万兆级网络为代表的高速网络的测试方法,是十分迫切和必须的。在进行网络测试时,流量可以来源于回放,真正的流量是被根据不断回放历史捕捉到的一些数据来实现的;同时,还可以对流量进行模拟,这种并非还原真实,而是根据统计到的特点,自主的模仿出一些流量。两种方法没有很确定的优劣之分,只是在某些给定的条件下,某一方会有优势。在分析了上述目前广泛应用的两种方法,论文对基于FPGA对两种方案进行了修正与改进,详细的描述如下:1、根据已有的回放的方案,加入硬件因素,形成软硬件联动的方案。该方案理论上解决了以万兆级网络为代表的高速网络的流量生成问题。此方法将回放中的文件大小提高到了GB级,现有的方法只有64MB,这有赖于FPGA与主机存储优势的结合。为提高底层数据传输速度,采用了轮询的方式,并利用映射优化内存访问。同时从新设计了传输引擎,实现网络与主机本地数据交换速度达到16Gbps。此外,结合随机存取硬盘和主机内存的特点,还设计了几种新的回放方式:无拷贝利用静态缓冲的回放,基于随机存取硬盘的高速GB级回放模式和已经达到TB级的针对硬盘的低俗回放。对内存的优化使得数据拷贝有了将近70%的性能提高。在误差控制方面,利用FPGA设计基于令牌的方案,大大提高精度,相对误差不超过10E-5.2、结合硬件特点改进流量模拟的方法,先分析网络中的数据的特征,根据各部件自身所特有的功能,最终形成一种标志各部件的特征序列流,最后根据已有的特征序列,利用FPGA模拟生成流量数据。并且设计了对用户友好的借口,流的配置根据具体的动态与静态的特征描述确定。用户可以自主的配置数据头部不超过256字节的基于各种网络协议的数据包格式。相关测试结果表明此方案能正确稳定的运行,实现了模拟万兆级网络流量数据的预期功能。
[Abstract]:Network testing is also a means to ensure the equipment, function and running state of the whole network system. In network testing, traffic generation is a very important technology, and the degree of correctness and error of the result is determined by it. With each passing day, the network data exchange rate is faster and faster, the demand for accuracy is higher and higher, the traffic is increasing, for a network of ten thousand megabytes, the existing testing methods are difficult to meet the needs. Under the new situation, it is very urgent and necessary to study the test method of high-speed network, which is represented by the multi-megabyte network. In the network testing, the traffic can come from playback. Real traffic is realized based on some data captured in the history of playback; at the same time, traffic can be simulated, not to restore reality, but according to the characteristics of statistics. Independently mimic some of the flow. There is no definite difference between the two methods, but under certain given conditions, one party will have an advantage. This paper modifies and improves the two schemes based on FPGA. The detailed description is as follows: 1. According to the existing playback scheme, the hardware factor is added. The scheme solves the problem of traffic generation in high speed network represented by ten thousand megabytes network theoretically. This method increases the file size in playback to GB level. The existing methods are only 64MB, which depends on the combination of FPGA and host storage advantages. In order to improve the speed of data transmission, polling is adopted and memory access is optimized by mapping. At the same time, a new transport engine is designed. The speed of local data exchange between network and host can reach 16Gbps.In addition, combining the characteristics of random access hard disk and host memory, several new playback methods are designed: the playback with no copy and static buffering. High speed GB playback mode based on random access hard disk and low standard playback for hard disk that has reached TB level. The optimization of memory improves the performance of data copy by nearly 70%. In error control, Using FPGA to design token based scheme, the accuracy is greatly improved, the relative error is not more than 10E-5.2. the method of traffic simulation is improved by combining the characteristics of hardware. Firstly, the characteristics of data in network are analyzed, according to the special function of each component itself, Finally, a characteristic sequence flow is formed to mark each component. Finally, according to the existing feature sequence, the traffic data is generated by FPGA simulation, and a user-friendly excuse is designed. The configuration of the stream is determined according to the specific dynamic and static characteristic description. The user can independently configure the data packet format based on various network protocols with data header not exceeding 256 bytes. The related test results show that the scheme can run correctly and stably. The expected function of simulating the network traffic data of ten thousand megabytes is realized.
【学位授予单位】：湖北工业大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP393.06

【相似文献】