分布式大型网络数据库的信道属性权重分配方法

为解决传统方法在所需分配信道量很多的情况下，信道分配准确性低，忽略了对信道分配公平性的考虑，导致通信过程中干扰增加的问题。为此，提出一种新的分布式大型网络数据库的信道属性权重分配方法。通过建立分布式大型网络数据库干扰模型。依据建立的干扰模型，结合流量负载与干扰对链路权重进行设置，将属性权重当成依据确定信道优先级，令权重高的链路优先选择信道。属性权重的分配非常关键，依据信道属性权重的分配主要受属性关键性影响的特性，给出信道属性权重分配公式，获取不同信道属性分配的权重。当前得到的权重为静态权重，为了改善实用性和动态性，提出属性权重的动态迭代学习方法，在原有方法的基础上通过梯度下降法实现信道属性权重调整，达到信道属性权重分配的目的。结果表明，所提方法的信道利用率可达到92.31%。可见所提方法分配结果公平、干扰少，整体性能优。

Research on channel attribute weight assignment method for distributed large scale network database

In order to solve the traditional method, when the required amount of allocated channels is large, the channel allocation accuracy is low, and the consideration of channel allocation fairness is neglected, resulting in an increase in interference during communication. To this end, a new method of channel attribute weight assignment for distributed large-scale network databases is proposed. By establishing a distributed large network database interference model. According to the established interference model, the link weight is set according to the traffic load and the interference, and the attribute weight is used as the basis to determine the channel priority, and the link with the high weight is preferentially selected. The assignment of attribute weights is very important. According to the characteristics of channel attribute weights, which are mainly affected by attribute criticality, the channel attribute weight assignment formula is given, and the weights of different channel attribute assignments are obtained. The current weights are static weights. In order to improve the practicability and dynamics, a dynamic iterative learning method for attribute weights is proposed. Based on the original method, the channel attribute weights are adjusted by the gradient descent method to achieve the purpose of channel attribute weight distribution. The results show that the channel utilization of the proposed method can reach 92.31%. It can be seen that the proposed method has fair distribution, less interference and excellent overall performance.