Fund Project:Project supported by the 2020 Military-Civilian Integration Development Research Project of Hebei Province, China (Grant No. HB20JMRH008), the Humanities and Social Sciences Fund of the Ministry of Education of China (Grant No. 19YJAZH069), the National Natural Science Foundation of China (Grant Nos. 61672206, 61572170), the Science and Technology Support Program of Hebei Province, China (Grant Nos. 18210109D, 20310701D), and the High Level Talent Funding Project of Hebei Province, China (Grant No. A2016002015)
Received Date:10 August 2020
Accepted Date:09 October 2020
Available Online:17 March 2021
Published Online:05 April 2021
Abstract:The transmission performance of the network depends to a certain extent on the topology of the network. This article analyzes the traffic dynamics of complex networks from the perspective of structural information, and looks for information structure measurement indicators that affect network traffic capacity. Existing research shows that the communicability sequence entropy of complex networks can effectively quantify the overall structure of the network. Based on this measurement, the difference between networks can be effectively quantified, and the connotation of sequence entropy of communicability can be explained. Communication sequence entropy can effectively quantify the overall structure information of the network. In order to characterize the overall traffic capacity of the network, the communication sequence entropy is introduced into the phenomenon of complex network congestion, the correlation between the network communication sequence entropy and the transmission performance is studied, and the internal mechanism of this correlation is analyzed. Simulations in BA scale-free network model and WS small-world network model show that the communication sequence entropy of the network is closely related to its traffic capacity. As the communication sequence entropy increases, the uniformity of the network topology will increase, and the traffic capacity will increase significantly. The traffic capacity of the network is a monotonically increasing function of the entropy of the communication sequence, and is positively correlated with the entropy of the communication sequence. The communication sequence entropy of the network can effectively evaluate the traffic capacity of the network. This conclusion can provide a theoretical basis for the design of a high traffic capacity network and help provide an effective strategy for the design of the high traffic capacity of the network, which can be optimized by increasing the communication sequence entropy. Keywords:complex network/ communication sequence entropy/ traffic capacity/ congestion
其中, $ \sigma _{sd} $表示源节点$ s $到达目的节点$ d $之间的最短路径条数, $ \sigma _{sd}\left ( i \right ) $表示从源节点$ s $到达目的节点$ d $的最短路径中经过节点$ i $的最短路径条数. 虽然介数的定义基于最短路径路由算法, 但是有很多已经存在的路由算法并不是以最短路径为基础. 研究人员将介数的定义扩充为有效介数, 用以评估实际路由策略下网络的容量情况, 有效介数一般定义为[53]
图1(a)和图1(b)分别给出了BA无标度网络和WS小世界网络的通信序列熵$ S_{\rm N} $和网络的度分布$ P(k) $之间的关系. 可以看出, 两种类型网络的通信序列熵在增大的时候, 网络的度值范围增大, 但网络中节点的度分布曲线从陡峭化向平缓状态发展, 网络愈加均匀. 分析如下: 网络向均匀网络发展时, 通信序列元素集合B中的元素会变得更加均匀, 进而通信序列熵值会增大. 随着网络通信序列熵的增大, BA无标度网络的度分布并没有改变整体的幂律分布形状, WS小世界网络的度分布则更加趋近于正态分布. 图 1 网络的通信序列熵SN与网络的度分布P(k)之间的关系图 (a) BA无标度网络; (b) WS小世界网络 Figure1. Rrelationship between the network’s communication sequence entropy SN and the network’s degree distribution P(k): (a) BA scale-free network; (b) WS small world network.
图2和图3给出了网络的通信序列熵与传输容量的关系. 可以看出, 网络传输容量随着通信序列熵的增大而增大. 理论分析如下: 网络的均匀性会随着网络节点之间的连接情况的改变而改变. 当网络变得越来越稠密时, 网络通信序列熵中的元素数值彼此接近且均匀, 进而导致网络的通信序列熵增大. 从网络的拓扑结构角度来看, 网络中许多非邻节点随着通信序列熵的增大或减小而进行连接或断开, 这些节点的度值会因此增大或减小, 使网络中原有的核心节点的影响降低或增大. 网络的核心节点比例程度极大地影响网络的均匀性. 当均匀性较低时, 网络中核心节点影响力高, 网络向非均匀性网络发展, 根据传统流量模型随机选取的两个节点之间的数据包传输路径经过网络的核心节点的概率高, 数据包将会通过网络的核心节点, 核心节点的处理能力是固定的, 那么数据包在核心节点的缓存队列中需要等待更长时间, 随着时间的延长, 网络就会处于一个拥塞状态. 反之, 当均匀性程度较高时, 网络中核心节点影响力低, 网络向均匀性网络发展, 根据传统流量模型随机选取的两个节点之间的数据包传输路径经过网络的核心节点的概率低, 数据包传输将会经过更多的普通节点, 舒缓了核心节点上的负载压力(数据包负载数目), 网络不易拥塞. 且本次路由策略选取的是有效路由的策略, 合理有效地避开了部分核心节点, 再一次降低了核心节点的影响力. 对以上分析我们采取负载在节点分布的方法进行验证. 图 2 (a)不同的通信序列熵的BA无标度网络下, 有序参数H(R)与数据包生成率R的关系; (b) BA无标度网络通信序列熵$S_{\rm N}$与传输容量$R_{\rm c}$的关系. 采用的路由策略为有效路由策略 Figure2. (a) Relationship between the order parameter H(R) and the packet generation rate R under BA scale-free network with different communication sequence entropy; (b) relationship between BA scale-free network communication sequence entropy$S_{\rm N}$ and traffic capacity $R_{\rm c}$. The routing strategy adopted is an effective routing strategy.
图 3 (a)不同的通信序列熵的WS小世界网络下, 有序参数H(R)与数据包生成率R的关系; (b) WS小世界网络通信序列熵$S_{\rm N}$与传输容量$R_{\rm c}$的关系. 采用的路由策略为有效路由策略 Figure3. (a) Relationship between the order parameters $H(R)$and the packet generation rate R under the WS small world network with different communication sequence entropy; (b) relationship between WS small world network communication sequence entropy$S_{\rm N}$ and traffic capacity $R_{\rm c}$. The routing strategy adopted is an effective routing strategy.
图4(a)和图4(b)所示为两种网络的节点度值上的数据包负载 (traffic load)分布. 可以看出, 随着通信序列熵的增大, 数据包负载在各个不同度值大小的节点间的分布愈加缓和及均匀. 核心节点和普通节点的数据包负载数目随着通信序列熵的增大渐渐地接近于相同的数值. 数据包负载在传输过程中的分布愈加均匀, 进而使网络的传输容量整体提升, 所以BA无标度网络及WS小世界网络的传输容量随着网络的通信序列熵的增大而增大, 成正关联关系. 通信序列熵亦可成为评估同种类型但不拓扑同结构网络的传输容量的指标. 图 4 数据包负载 (traffic load) 在网络中节点度值上的分布情况 (a) BA无标度网络; (b) WS小世界网络 Figure4. Distribution of traffic load on degree value of nodes in the network: (a) BA scale-free network; (b) WS small world network.
图5(a)和图5(b)给出了网络的平均路径长度随通信序列熵的变化. 可以看出, 网络的平均路径长度随着通信序列熵的增大反而减小. 这是因为路由策略考虑的数据包在传输时尽量是避开核心节点, 对于数据包选择的路径而言, 节点之间间隔较短的路径上有度值较大的核心节点, 节点之间间隔远的路径上有度值较小的普通节点. 由于随着网络通信序列熵的增大, 网络从稀疏状态向稠密状态发展, 使节点之间间隔远的路径上的非邻节点之间进行连接, 出现了许多的度值较大的节点, 使核心节点的地位降低. 源节点和目的节点之间的数据包传输会经过更多新出现的度值较大节点, 经过更少的边. 从而导致节点与节点之间的传输路径长度变小, 进而导致整个网络的平均路径减小, 会使数据包到达各个目的节点的平均时间整体降低, 提高了数据包整体的传输效率. 图 5 网络的通信序列熵$S_{\rm N}$与平均路径长度$\left\langle {L} \right\rangle $的关系 (a) BA无标度网络; (b) WS小世界网络 Figure5. Relationship between communication sequence entropy SN and average path length$\left\langle {L} \right\rangle $ in the network: (a) BA scale-free network; (b) WS small world network.
根据2.3节得知, 网络中介数值最大的节点最先发生拥塞, 与网络传输容量成反比. 通过仿真实验计算网络的传输容量非常耗时, 因此可以通过计算在不同路由策略下网络中所有节点的有效介数的最大值来间接评估网络的传输容量. 通过对图6(a)和图6(b)的观察可以看出, 基于有效路由算法下, 同等规模的网络通信序列熵值变化较快时, 网络的最大有效介数变化随之增快, 可见网络的节点最大介数敏感于通信序列熵的变化. 当通信序列熵最大时, 网络中最大介数值最小, 传输容量最大. 当通信序列熵最小时, 网络中最大介数值最大, 传输容量最小. 图 6 网络的通信序列熵$S_{\rm N}$与节点的最大介数$B_{\rm {max}}$的关系 (a) BA无标度网络; (b) WS小世界网络 Figure6. Relationship between communication sequence entropy $S_{\rm N}$ and the maximum betweenness of nodes $B_{\rm {max}}$ in the network: (a) BA scale-free network; (b) WS small world network.