高阶间隔估计算法在网络流量监测中的研究

发布时间：2018-04-20 05:14

本文选题：抽样技术 + 高阶间隔　；参考：《昆明理工大学》2014年硕士论文

【摘要】：高带宽和高吞吐量是目前网络发展的重要方向。随着网络传输速率的不断增大,流量监测的技术得到不断完善,目前主要应用的技术是基于报文或时间间隔抽样的抽样,然后根据抽样结果进行均值和方差估算等进行网络流量评估。随着网络带宽的数量级提升,数据传输率和有效吞吐量大幅度上升,在相同情况下采用报文抽样所需要的抽样次数明显增加,导致系统付出的资源消耗比也随之上升；而采用时间间隔抽样将会导致相同时间间隔中的数据量和随机跃变概率增加,因此为了获得相对准确的估计值就需提高抽样频率,增加资源消耗。因此伴随网络流量的逐步增大,优化常规流量估计方法,实现在相对较少的抽样频率和资源消耗的基础上,获得更高的估计精度是一个较有意义的研究方向。为满足更加高速的数据传输网络流量监测的低资源消耗,高精度的需求,本文在对大量的抽样方式进行研究的基础上提出了新的适用于高带宽高吞吐量的流量监测方法--高阶间隔估计算法,它以减少系统内存资源占用及提高流量监测精度为前提,在理论算法及优化后的低阶间隔抽样数据的基础上对高速网络的流量监测方案进行改进。即以低阶间隔的抽样数据为基础,以高阶间隔估计算法为理论前提,进行高阶间隔信息的估算,并根据高阶间隔信息估算值的大小进行当前高速网络流量的评价与估计。针对报文抽样、时间间隔抽样和低阶抽样高阶间隔估算等方法,文章提出了利用KL散度理论对网络流量监测的精度进行评价,并在不同流量监测方案的基础上对系统资源消耗和精度等性能进行了仿真对比和说明。仿真结果表明,当网络带宽和吞吐量不断增大时,文章所提出的算法可以有效的解决资源消耗和精度之间的矛盾,确保网络流量监测的可行性。随着采样间隔的增大,高阶估计算法的综合效果更加明显,一方面有效的解决了抽样频率较高所引起的高资源消耗的问题,另一方面流量估计精度也相对明显提高。由此可以看出文章所提的方法在更高速的网络流量监测中拥有更好的使用价值。
[Abstract]:High bandwidth and high throughput are the important directions of network development. With the increasing of network transmission rate, the technology of traffic monitoring has been improved. At present, the main technology is based on the sampling of message or time interval sampling. Then the network traffic is evaluated according to the mean and variance estimates of the sampling results. With the increase of network bandwidth, the data transmission rate and effective throughput increase greatly, and the sampling times required for packet sampling increase obviously in the same situation, which leads to the increase of the resource consumption ratio of the system. The time interval sampling will increase the amount of data and the probability of random jump in the same time interval. Therefore, in order to obtain a relatively accurate estimate, it is necessary to increase the sampling frequency and resource consumption. Therefore, with the gradual increase of network traffic, it is a meaningful research direction to optimize conventional traffic estimation methods and achieve higher estimation accuracy on the basis of relatively low sampling frequency and resource consumption. In order to meet the demand of low resource consumption and high precision for more high-speed data transmission network traffic monitoring, Based on the research of a large number of sampling methods, this paper presents a new traffic monitoring method for high bandwidth and high throughput, which is called high-order interval estimation algorithm, which is based on reducing the memory consumption of the system and improving the accuracy of traffic monitoring. Based on the theoretical algorithm and the optimized low-order interval sampling data, the flow monitoring scheme of high-speed network is improved. Based on the sampling data of low order interval and the theoretical premise of high order interval estimation algorithm, the high order interval information is estimated, and the current high speed network traffic is evaluated and estimated according to the size of high order interval information. Aiming at the methods of packet sampling, time interval sampling and high order interval estimation of low order sampling, this paper presents a KL divergence theory to evaluate the accuracy of network traffic monitoring. On the basis of different flow monitoring schemes, the performance of system resource consumption and precision is compared and explained. The simulation results show that the proposed algorithm can effectively solve the contradiction between resource consumption and precision and ensure the feasibility of network traffic monitoring when the network bandwidth and throughput are increasing. With the increase of sampling interval, the synthesis effect of high-order estimation algorithm is more obvious. On the one hand, it solves the problem of high resource consumption caused by high sampling frequency, and on the other hand, the accuracy of flow estimation is improved. From this, we can see that the method proposed in this paper has better use value in higher speed network traffic monitoring.
【学位授予单位】：昆明理工大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TP393.06

【参考文献】