基于流量倾斜分类的网络调度算法仿真

发布时间：2018-05-20 03:03

本文选题：数据倾斜 + 任务调度　；参考：《计算机仿真》2013年11期

【摘要】：网络流量数据具有突发性、非线性和时变性等等特点,造成在一些复杂网络流量在时域内存在流量特征分形弱化的趋势,可预测特征出现了差异淡化的特征,使得传统的网络调度方法出现局部最优等缺点,从而导致调度精度低和难问题。为优化调度性能,在充分研究传统方法的基础上,引入了流量倾斜度的概念,利用流量数据中的流量特征进行一种属性倾斜程度的判断,将信道中的流量特征进行一种相似度计算,计算流量异常总的出现次数时,对异常流量的子概念进行累加。在调度的过程中,充分考虑流量区域之间的倾斜性,减少不必要的繁琐调度程序。最后在Matlab平台采用网络流量数据对模型进行了测试,实验结果表明,算法提高了网络流量的预测精度。相对于传统的网络流量预测方法,改进算法更适合于复杂的网络流量数据预测。
[Abstract]:Network traffic data have the characteristics of sudden, nonlinear and time-varying, which lead to the weakening of fractal characteristics of traffic in some complex network traffic in time domain, and the characteristics of different desalination of predictable features. The shortcomings of traditional network scheduling methods such as local optimization lead to low precision and difficult problems. In order to optimize the scheduling performance, the concept of traffic inclination is introduced on the basis of studying the traditional methods, and the traffic characteristics in traffic data are used to judge the degree of attribute tilt. The traffic characteristics in the channel are calculated by a similarity measure. When the total number of traffic anomalies is calculated, the subconcept of the abnormal traffic is accumulated. In the process of scheduling, the inclination between traffic regions is taken into account, and unnecessary scheduling procedures are reduced. Finally, the model is tested on Matlab platform using network traffic data. The experimental results show that the algorithm improves the accuracy of network traffic prediction. Compared with the traditional network traffic prediction method, the improved algorithm is more suitable for complex network traffic data prediction.
【作者单位】：南阳理工学院软件学院;
【分类号】：TP393.06

【共引文献】