• <tr id="yyy80"></tr>
  • <sup id="yyy80"></sup>
  • <tfoot id="yyy80"><noscript id="yyy80"></noscript></tfoot>
  • 99热精品在线国产_美女午夜性视频免费_国产精品国产高清国产av_av欧美777_自拍偷自拍亚洲精品老妇_亚洲熟女精品中文字幕_www日本黄色视频网_国产精品野战在线观看 ?

    Improving robustness of complex networks by a new capacity allocation strategy

    2021-01-21 02:13:06JunLiu劉軍
    Chinese Physics B 2021年1期
    關(guān)鍵詞:劉軍

    Jun Liu(劉軍)

    Department of Information Science and Technology,Tianjin University of Finance and Economics,Tianjin 300222,China

    Keywords: capacity,cascading failure,complex network,robustness

    1. Introduction

    In the past few years,complex network theory has developed rapidly and has been used to study robustness of infrastructure network against cascading failures,such as transportation networks,[1,2]logistics network,supply chain networks,[3]and power grids.[4,5]Cascading failure can make great damage to infrastructure networks,which can be described as follows. If there are some nodes fail, it can cause the redistribution of loads, which may make the loads of some nodes exceed their capacities, so the failures scale will be continuously expanded,eventually,resulting in collapse of the entire network. Cascading failure is fatal to the functioning of infrastructure networks. These infrastructure networks are important for us,once they are destroyed,it will seriously affect people’s production and life and cause great property losses.In June 2019,Argentina and Uruguay suffered massive power outages across the country.The blackout also affected cities in Paraguay,southern Brazil,and parts of Chile. What is worse,the power outages have led to water and gas shortages in five South American countries, as well as the closure of subway,bus and intercity railway, which have severely affected residents’ lives. So cascading failures[6–9]and robustness[10,11]of infrastructure networks are paid more attention by scholars because of their great importance. Chen et al.[12]researched controllable robustness on networks with considering the external or internal failure.Jing et al.[13]constructed a cascading failure model by considering time delay in failure propagation process and considered the effect of the node recovery probability on cascading failures.

    As we all know, there exists load on nodes in infrastructure networks.[14]So the failure of nodes can cause the redistribution of the load. And it will make the load of some nodes exceed their capacities,which can lead to further overload failures and ultimately spread over the entire network. In order to improve the robustness of complex networks against cascading failure, scholars have researched from different perspectives,such as the initial load of node, the capacity allocation strategy, and load redistribution strategy.[15–18]. Motter et al.[15]proposed a classical ML load model,which consider that network information was exchanged along the shortest path between nodes, so they defined the betweenness of a node as its initial load. At the same time, they assume that the capacity of a node is proportional to its initial load due to the cost limit,this capacity allocation strategy is adopted in many researches.[18–21]The ML model requires each node to master the whole information of the network. So,it is appropriate for small networks. In fact the global information is hard to obtain,especially for large scale networks.And in the actual network system the load may not be transmitted along the shortest path,such as public transport network. Wang et al.[22]defined the initial load of node as the product of the degree of node and the sum of the degree of its neighbors,which is in positive correlation with betweenness centrality.[23]Considering both local and global information,Liu et al.[24]defined the load of a node by combining the node degree with the betweenness. We all know that cascading failures do not occur as long as the capacity of the node is large enough.However,capacity is costly,the capacity of a node cannot be infinite. So it is very meaningful to study the capacity allocation strategy under limited resources, which has great effect on the network robustness.However,there is not much literature on the dynamical capacity allocation procession to resist the cascading failure. Wang et al.[25]proposed an improved capacity model based on ML model, and defined the cost to protect vertices. According to the definition of cost,their model has always a lower value of the cost than ML model. However, in this model, there will be no fault-tolerant capacity for some nodes if β /=0, which is unreasonable. Because the load on a node usually fluctuates in real network, and the network will not work properly due to the presence of these nodes without fault-tolerant capacity. Li et al.[26]also proposed an improved capacity model based on ML model. Different from ML model, the faulttolerant capacity and topological structure are interrelated in their model,which can achieve better network robustness. But the model considered that network information was exchanged along the shortest paths between nodes. In fact, the loads in some real networks are not always exchanged along the shortest paths.[17]To solve this problem,Wang et al.[17]proposed a load redistribution strategy based on local information. Since then,more and more scholars[20,21,27]began to study the load redistribution strategy based on local information or to use the load local redistribution strategy to study the cascading failure. And most of them considered that the capacity of the node is proportionate to its initial load which was proposed in ML model.

    Based on the above analyses, we put forward a new capacity allocation strategy based on the community structure,meanwhile we use the load local redistribution strategy to research the robustness of complex networks. And in order to demonstrate the effectiveness of the new strategy, a comparison was made with the capacity allocation strategy of ML model.

    2. Capacity allocation strategy and cascading failure model

    Cascading failures of the network can be described as follows, each node has an initial load L(0) and load capability C, under normal circumstances, L(t)<C, the network is in the steady state. After the node is removed (attacked or mechanical failure),the load on the failed node is redistributed to its neighbors. If the capacity of the neighbor is not enough to handle the extra loads,the neighbor would collapse,which can lead to load-redistribution and further overload failures and finally result in the collapse of the entire network. Assume that the potential cascading failure is caused by the removal of one single node, then we pay attention to the following dynamic process of the network. Cascading failure model used in this paper is briefly described as follows.

    2.1. Definition of initial load

    The initial load of node j is set as

    where Lj(0) is the initial load of node j, kjand Γjare the degree of node j and the set of neighbors of node j, respectively. θ is a tunable parameter and governs the strength of the node load. The definition of initial load is reasonable because the product of the degree of node and the sum of the degree of its neighbors is in positive correlation with betweenness centrality.[28,29]

    2.2. Capacity allocation strategy

    In the existing model, one model should be paid special attention: the capacity allocation strategy in Motter–Lai(ML)model,[15]which is widely used to study cascading failures in complex networks. The ML model assumes the capacity of node j is proportional to its initial load as

    where β (≥0)is a tolerance parameter representing the extra capacity.

    In this paper, we put forward a new capacity allocation strategy against cascading failures,as folows:

    In the new capacity allocation strategy model proposed in this paper,the cost is

    where siis the number of nodes in community i.

    Different from ML model,the fault-tolerant capacity and community structure are interrelated in the current model,and the intensity of correlation is controlled by γ.Considering that each component of the network may have different fault tolerance capabilities,this work proposes a more general allocation strategy to allocate additional capacity.

    2.3. Load-redistribution strategy

    Considering the time-varying load and the time-varying residual capacity, we adopted the method in Ref.[31], which can be defined as

    where ΔLjiis the extra load that node j should obtain from the failure node i,Cjis the capacity of node j,Lj(t)is the load of node j at the time step t,q is the number of time steps before the cascading process stops. j is the nearest neighbor of node i,Γiis the set of the nearest neighbors of node i.

    So at the time step t+1,if Lj(t)≤Cj,the load of node j is

    where Γiis the set of the nearest neighbors of node j.

    3. Simulation and analysis

    According to Ref. [31], Piis used to evaluate the size of the cascade due to the removal of node i

    where Fiis the number of failure nodes caused by the removal of node i. N is the total number of nodes in the complex network.

    The average size of the cascade due to the removal of one node can be described as

    In this paper, we focus on cascading failure triggered by the removal of a single node. And to measure the overall level of the network robustness against cascading failures due to the removal of a single node,we randomly remove only one node(such as node i)each time,and calculate the proportion of failure nodes(Pi)induced by removing node i.Remove a different node each time until all nodes have been removed once. Then the average proportion of failure nodes P is calculated according to Eq.(11)to evaluate the overall level of the network robustness against cascading failures. Each result is the average value of more than 20 realizations for the same topology.

    Two artificial network models are tested, i.e., Erd¨os–R′enyi model and the Barabasi–Albert (BA) model.[32]The two typical complex networks used in this paper are all generated by Pajek with network size N=1000 and average degree〈k〉=8. Furthermore,US power grid network as real complex network is considered. US power grid network is the highvoltage power grid in the Western States of the United States of America. There are 4941 nodes and 6594 edges in the power grid of the western United States. The nodes are transformers, substations, and generators, and the ties are high-voltage transmission lines.

    Many literatures assume that node load is constant when studying network robustness,but in fact,node load is variable.Hence, in order to verify the effectiveness of our method, we conduct experiments from two aspects,i.e.,the load in the network is constant during normal operation and the load in the network is time-varying during normal operation.

    The time-varying load Li(0)of node i can be defined as

    where Δl is the disturbance variable,it can be defined according to the variation trend of the load. Here,in order to explore the effectiveness of the model for simplicity,Δl=βL0iεi,εiis a random number between-1 and 1.

    First of all,we compare and analyze the new strategy and ML method based on two typical networks.

    Figure 1 shows the experimental results in BA network.From Fig.1 one can see that the new strategy(as shown in the line) can effectively reduce the scale of cascade failure compared with ML model(as shown in the dotted line)in BA network regardless of whether the load is variable or fixed. And with other parameters unchanged, the cascading failure scale decreases as the value of γ increases. γ is an adjustable parameter in Eq.(3),the smaller the value of γ is,the more homogeneous the capacity of node is. With the increase of γ,capacity resources tend to be allocated to communities with high average node degree. The experimental results indicate that allocating more capacity resources to the community with high average node degree is beneficial to reduce the scale of cascading failure. In Fig. 1, with the increase of β, the scale of cascading failure decreases,and one can see that cascading failures can be avoided if the capacity is large enough,which is consistent with the actual situation. Meanwhile,we can find that the scale of cascading failure decreases with the increase of θ for the same β, γ, which means that the more heterogeneous the initial load of the network nodes is, the stronger the overall destruction resistance of the network will be. But this kind of network can be vulnerability to deliberate attack with the increase of θ (shown in Fig. 7). Furthermore, comparing Figs.1(a),1(c),and 1(e)with Figs.1(b),1(d),and 1(f),we can find that the scale of cascading failure under constant load condition is lower than that under time-varying load condition, which indicates that BA network is more vulnerable under time-varying load condition.

    Fig.1. The average proportion of failure nodes P as a function of β and γ based on ML model and new strategy in BA networks for different θ. (a)Constant load, θ =0.4; (b) time-varying load, θ =0.4; (c) constant load, θ =1; (d) time-varying load, θ =1; (e) constant load, θ =1.6; (f) timevarying load,θ =1.6. The dotted curves reflect the relationship between P and β for different θ and γ based on ML model. The solid curves present the relationship between P and β for different θ and γ based on new strategy.

    Fig.2. The average proportion of failure nodes P as a function of β and γ based on ML model and new strategy in ER networks for different θ.(a)Constant load,θ =0.4;(b)time-varying load,θ =0.4;(c)constant load,θ =1;(d)time-varying load,θ =1;(e)constant load,θ =1.6;(f)time-varying load,θ =1.6. The dotted curves reflect the relationship between P and β for different θ and γ based on ML model. The solid curves present the relationship between P and β for different θ and γ based on new strategy.

    Figure 2 shows the experimental results in ER network.As can be seen from the experimental results,compared with the capacity allocation strategy proposed in ML model, the new strategy also can reduce the scale of cascading failure most of the time even though it is not obvious. For the random network,the average degree of each community does not differ much,so the effect of the new strategy is not obvious.In Fig.2,P is decreasing with the increase of β for the same θ,γ,which is consistent with the basic theory that there will be no cascading failures as long as the network has sufficient capacity. And focus on the effect of θ on the robustness of complex network,one can see that in ER network P is decreasing with the increase of θ. But the effect is not obvious in ER network compared with that in BA network. Moreover, considering the comparative analysis under constant node load and timevarying load, almost all the results (Figs. 1 and 2) show that the network is more vulnerable under time-varying load condition. Although the load redistribution strategy adopted in this paper has considered the time-varying characteristics of the load, the random variation of the load makes the distribution of the residual capacity in the network uncontrollable.For example,large load nodes may have little surplus capacity while low load nodes have relatively high surplus capacity. So capacity cannot be fully utilized due to the unreasonable distribution of residual capacity, which increases the network’s vulnerability.

    In order to show the effect of the new strategy more intuitively,we introduce indicator w,as

    where PMLis the average size of the cascade under the ML model, PNewis the average size of the cascade under the new capacity allocation strategy.

    Considering that the overall destruction resistance of the network is stronger with θ =1.6,we analyze the effects of β and γ on w with θ =1.6 in two typical networks.

    From Fig. 3 one can see that the new strategy improved the robustness of BA network compared with the strategy of ML model,and the robustness of BA network is gradually enhanced with the increase of γ regardless of whether the load is variable or fixed. The performance of the BA network against cascading failures has been improved by nearly 90%when β =0.13 and γ =1.9 (as shown in Fig. 3(a)). Meanwhile, we find that wincreases as βincreases, but wpresents a downward trend when β >1.3, which indicates that there is a β =βmaxthat maximizes w for each γ. A similar conclusion can be drawn on random networks,which can be seen in Fig. 4. However, the performance of the ER network against cascading failures has been improved by nearly 10%based on the new strategy,which is much less than that in BA network.

    Although the cost for the two methods (i.e., ML model and the new strategy)can be calculated by Eq.(5),cost cannot be used to judge the merits of the two methods.And the impact of capacity cost on network destructiveness is more suitable as an evaluation index. So we introduce contribution rate of unit capacity to network robustness Q,which is defined as

    The larger Q is,the higher the capacity utilization is.

    Fig.3. Indicator w as a function of β and γ for fixed θ =1.6 in BA network,with N=1000 and〈k〉=8. (a)Constant load and(b)time-varying load.

    Fig.4. Indicator w as a function of β and γ for fixed θ =1.6 in ER network,with N=1000 and〈k〉=8. (a)Constant load and(b)time-varying load.

    From the results shown in Fig.5,one can see that Q goes up and then goes down with the increase of β in BA network.In other words,there is a β =βmaxthat maximizes Q for each θ. It indicates that it is not economical to increase the cost of capacity when β >βmax. Meanwhile, the scale of cascading failure has been reduced to a very low level when β =βmax,which can be seen in Fig. 1. So β ≤βmaxis a more reasonable range of β in the actual networks. Compared with ML model, our method can achieve a higher Q for the sameβ,when β ≤βmax. It indicates that the new strategy is more efficient in using network capacity. Figure 6 shows the experimental results in ER network.As shown in Fig.6,at the beginning Q increases with the increase of β,but once β reaches a certain value,Q stops growing as β goes up.Meanwhile,compared with ML model, the new strategy can achieve a higher Q for the same β in ER network even though it is not obvious.Comparing Fig.5 with Fig.6, one can see that the new strategy is more advantageous in the rational utilization of network capacity in BA scale-free network.

    Fig.5. Q as a function of βand γ for different θ under different strategies in BA networks: (a)constant load,θ =0.4;(b)time-varying load,θ =0.4;(c)constant load,θ =1; (d)time-varying load,θ =1; (e)constant load,θ =1.6; (f)time-varying load,θ =1.6. The dotted curves are the results of ML model. The solid curves are the results of the new strategy.

    Although typical model networks are useful in understanding how real networks behave, they cannot catch many of the structural properties of the real systems. Therefore,we consider a real complex network,i.e.,US power grid network.Here we mainly study its robustness against deliberate attacks.So we selected the top 10 highest degree nodes in US power grid network. And each time only one of the ten nodes is removed, then the size of the cascade Pidue to the removal of the node is calculated. The average size of the cascade due to the removal of one node can be described as

    where Γ represents the set of the top 10 highest degree nodes.

    As you can see from Fig. 7, our new strategy still performs well against deliberate attacks on real network,whether the load is fixed or time-varying. When parameter θ is small,cascading failure scale is more sensitive to the change of β. it can be seen in Fig.7(a),When β increases from 0.09 to 0.21,P10can decrease from 1 to 0. However,with the increase of θ,the effect of the increase of β on improving the robustness of the network against deliberate attacks becomes less obvious.It indicates that the increase of θ can reduce the robustness of the network against deliberate attacks if the capacity cost is fixed.Moreover,the increase of γ can improve the robustness of the network against deliberate attacks when other parameters are fixed.

    Fig.6. Q as a function of β and γ for different θ under different strategies in ER networks: (a)constant load,θ =0.4;(b)time-varying load,θ =0.4;(c)constant load,θ =1; (d)time-varying load,θ =1; (e)constant load,θ =1.6; (f)time-varying load,θ =1.6. The dotted curves are the results of ML model. The solid curves are the results of the new strategy.

    Fig.7. P10 as a function of β and γ for different θ under different strategies in US power grid networks. (a)Constant load,θ =0.4;(b)time-varying load,θ =0.4;(c)constant load,θ =1;(d)time-varying load,θ =1;(e)constant load,θ =1.6;(f)time-varying load,θ =1.6. The dotted curves are the results of ML model. The solid curves are the results of the new strategy.

    Furthermore,we research the contribution rate of unit capacity to network robustness Q in US power grid network.From the experimental results, we can see that our new strategy has a higher capacity utilization. And from Fig.8,one can see that Q is decreasing with the increase of θ. It is reasonable because the heterogeneity of the network becomes more obvious with the increase of θ. And heterogeneous networks are vulnerable to deliberate attacks by high degree nodes. So the contribution rate of unit capacity decreases as θ increases.When θ is small, the value of Q increases rapidly with the increase of β,which indicates that the network with homogeneous load distribution can effectively improve its robustness against deliberate attacks by increasing capacity cost. When θ is big, the value of Q grows slowly with the increase of β,which indicates that network with heterogeneous load distribution cannot resist the deliberate attack on generous nodes simply by increasing capacity cost.

    Fig.8. Q as a function of βand γ for different θ under different strategies in power grid networks: (a)Constant load,θ =0.4; (b)time-varying load,θ =0.4; (c)constant load, θ =1; (d)time-varying load, θ =1; (e)constant load, θ =1.6; (f)time-varying load, θ =1.6. The dotted curves are the results of ML model. The solid curves are the results of the new strategy.

    4. Conclusions

    The cascading failure is a crucial process in many real systems,which can cause significant damage to network functions. Limited network capacity is a necessary condition for cascading failures, but the capacity of real system cannot be infinite. Changing the topology of the network can effectively improve the robustness of the network, but the topology of many existing network systems is not easy to change. So it is very important to allocate the limited network capacity reasonably without changing the network topology.

    In this paper, we developed a new strategy to allocate the limited capacity for suppressing cascade failure. Through comparison experiments,one can see that the new strategy can effectively reduce the scale of cascading failure. And it works better in scale-free networks than in random networks. Furthermore,we find that capacity efficiency is not monotonically increasing as the cost of capacity increases,that is to say,there is a cost threshold βmax. It is not economical to improve network robustness by increasing capacity costs when the cost is higher than βmax. And the capacity utilization of the new strategy is higher than that of ML model when the cost is lower than βmax. Meanwhile,the increase of γ can improve capacity utilization when β ≤βmax. Furthermore, it is found that cascading failures are more likely to occur when the network load is time-varying through comparative analysis.

    猜你喜歡
    劉軍
    劉軍作品
    大眾文藝(2022年10期)2022-06-08 02:33:28
    彩泥大變身
    The energy-saving advantages of burst-and-glide mode for thunniform swimming *
    Using spanwise flexibility of caudal fin to improve swimming performance for small fishlike robots *
    Numerical and experimental studies of hydrodynamics of flapping foils *
    映像畜牧業(yè)
    搶電視
    愛你(2017年35期)2017-11-24 18:30:41
    搶電視
    圖說
    看圖作文寫作指導(dǎo)
    午夜影院在线不卡| 丁香六月欧美| 欧美xxⅹ黑人| 国产欧美日韩一区二区三区在线| 伦理电影免费视频| 91aial.com中文字幕在线观看| 日韩,欧美,国产一区二区三区| 久久久久久免费高清国产稀缺| 国产成人av激情在线播放| 麻豆乱淫一区二区| 成人三级做爰电影| 日韩av免费高清视频| 久久久久视频综合| 操美女的视频在线观看| 亚洲欧美精品自产自拍| 久热这里只有精品99| 国产爽快片一区二区三区| 亚洲精品日本国产第一区| 日韩大片免费观看网站| 制服诱惑二区| 永久免费av网站大全| 观看av在线不卡| 亚洲欧洲国产日韩| 99香蕉大伊视频| 国产av码专区亚洲av| 久久久久精品人妻al黑| 亚洲av日韩精品久久久久久密 | 亚洲一码二码三码区别大吗| 99re6热这里在线精品视频| 一个人免费看片子| 国产精品久久久久久精品电影小说| √禁漫天堂资源中文www| 久久久久人妻精品一区果冻| 狂野欧美激情性xxxx| 日韩伦理黄色片| 国产精品一国产av| 亚洲av中文av极速乱| 精品酒店卫生间| 国产精品国产av在线观看| 91成人精品电影| 在线亚洲精品国产二区图片欧美| 国产黄色视频一区二区在线观看| 色婷婷久久久亚洲欧美| 久久青草综合色| 男女高潮啪啪啪动态图| 最新在线观看一区二区三区 | 老熟女久久久| 久久久久久久精品精品| 中文精品一卡2卡3卡4更新| 婷婷色综合大香蕉| 视频在线观看一区二区三区| 你懂的网址亚洲精品在线观看| 亚洲久久久国产精品| 国产亚洲av片在线观看秒播厂| 侵犯人妻中文字幕一二三四区| 韩国av在线不卡| 婷婷成人精品国产| 精品国产一区二区三区四区第35| 少妇被粗大的猛进出69影院| 国产精品.久久久| 国产极品天堂在线| av又黄又爽大尺度在线免费看| av天堂久久9| 亚洲婷婷狠狠爱综合网| 精品久久久精品久久久| 久久精品亚洲熟妇少妇任你| 国产免费一区二区三区四区乱码| 亚洲国产欧美网| 亚洲自偷自拍图片 自拍| a级毛片黄视频| 亚洲欧美精品综合一区二区三区| 久热爱精品视频在线9| 韩国高清视频一区二区三区| 另类精品久久| 国产精品99久久99久久久不卡 | 秋霞伦理黄片| 久久久国产欧美日韩av| 一级a爱视频在线免费观看| 色播在线永久视频| av在线播放精品| 欧美日韩视频高清一区二区三区二| 亚洲精品国产一区二区精华液| 久久久久国产一级毛片高清牌| 精品一区二区免费观看| 国产福利在线免费观看视频| 精品国产露脸久久av麻豆| 久久久久久久久久久免费av| 1024香蕉在线观看| 欧美精品高潮呻吟av久久| 老熟女久久久| 伊人久久国产一区二区| 亚洲成人一二三区av| 美女视频免费永久观看网站| 最近2019中文字幕mv第一页| 三上悠亚av全集在线观看| 亚洲美女搞黄在线观看| 亚洲美女搞黄在线观看| 在线观看人妻少妇| 欧美精品高潮呻吟av久久| 日韩 欧美 亚洲 中文字幕| 日本91视频免费播放| 国产精品久久久久成人av| 欧美少妇被猛烈插入视频| 亚洲综合色网址| 伦理电影免费视频| 国产1区2区3区精品| 亚洲国产精品一区三区| 欧美日韩亚洲国产一区二区在线观看 | 久久久精品免费免费高清| h视频一区二区三区| 亚洲国产欧美日韩在线播放| 日日啪夜夜爽| 国产在线视频一区二区| 亚洲欧美一区二区三区黑人| 国产 一区精品| 精品午夜福利在线看| 精品第一国产精品| 亚洲,一卡二卡三卡| 91老司机精品| 男人爽女人下面视频在线观看| 在线观看人妻少妇| 欧美日本中文国产一区发布| 国产精品国产三级专区第一集| 中文字幕色久视频| 一二三四中文在线观看免费高清| 国产精品一二三区在线看| 搡老岳熟女国产| 欧美最新免费一区二区三区| 国产精品一国产av| 国产欧美日韩综合在线一区二区| 午夜老司机福利片| 久久久精品国产亚洲av高清涩受| 日本wwww免费看| 亚洲欧洲精品一区二区精品久久久 | 久久精品亚洲av国产电影网| 精品国产乱码久久久久久小说| 久久久久视频综合| 久久ye,这里只有精品| netflix在线观看网站| 欧美成人午夜精品| 欧美成人精品欧美一级黄| 亚洲国产看品久久| 国产不卡av网站在线观看| 国产一区二区三区综合在线观看| 国产一区二区激情短视频 | 欧美日韩视频高清一区二区三区二| 无限看片的www在线观看| 国产精品一区二区精品视频观看| 一二三四在线观看免费中文在| 中文字幕亚洲精品专区| 中文字幕制服av| 日韩 欧美 亚洲 中文字幕| 蜜桃在线观看..| 久久女婷五月综合色啪小说| 亚洲七黄色美女视频| 日韩av不卡免费在线播放| 午夜日本视频在线| 日韩精品有码人妻一区| 成人国语在线视频| 高清av免费在线| 精品少妇久久久久久888优播| 日韩免费高清中文字幕av| 侵犯人妻中文字幕一二三四区| 国产一级毛片在线| 亚洲精品一区蜜桃| 欧美黑人精品巨大| 亚洲,欧美,日韩| 亚洲中文av在线| 母亲3免费完整高清在线观看| 99香蕉大伊视频| 免费不卡黄色视频| 国产精品三级大全| 一二三四中文在线观看免费高清| 一区二区三区四区激情视频| av不卡在线播放| 麻豆精品久久久久久蜜桃| 精品亚洲成a人片在线观看| 久久久欧美国产精品| 色网站视频免费| 两性夫妻黄色片| 中文字幕色久视频| 色视频在线一区二区三区| 亚洲欧美清纯卡通| 精品少妇一区二区三区视频日本电影 | 精品少妇久久久久久888优播| 王馨瑶露胸无遮挡在线观看| 国产片特级美女逼逼视频| 999精品在线视频| 亚洲美女搞黄在线观看| 免费少妇av软件| 国产又色又爽无遮挡免| 90打野战视频偷拍视频| 国产成人精品无人区| 国产 一区精品| 2018国产大陆天天弄谢| 咕卡用的链子| 99热全是精品| 午夜福利视频在线观看免费| 捣出白浆h1v1| 母亲3免费完整高清在线观看| 亚洲一区二区三区欧美精品| 少妇的丰满在线观看| 无限看片的www在线观看| 婷婷色综合大香蕉| 久久这里只有精品19| 一本久久精品| 欧美av亚洲av综合av国产av | 国精品久久久久久国模美| 欧美人与性动交α欧美软件| 无遮挡黄片免费观看| 伊人亚洲综合成人网| 国产男人的电影天堂91| 丝袜脚勾引网站| 亚洲男人天堂网一区| 十八禁人妻一区二区| 黑人欧美特级aaaaaa片| 狠狠精品人妻久久久久久综合| 国产高清不卡午夜福利| 国产精品人妻久久久影院| 国产免费视频播放在线视频| 男人舔女人的私密视频| 国产男女超爽视频在线观看| 99久久精品国产亚洲精品| a级毛片黄视频| av视频免费观看在线观看| 男的添女的下面高潮视频| 伦理电影免费视频| 久久久久久免费高清国产稀缺| 人体艺术视频欧美日本| 啦啦啦中文免费视频观看日本| 美女中出高潮动态图| av福利片在线| 亚洲成人一二三区av| 高清黄色对白视频在线免费看| 亚洲精品av麻豆狂野| 无遮挡黄片免费观看| 欧美日韩亚洲综合一区二区三区_| 一级片'在线观看视频| 少妇 在线观看| 国产又色又爽无遮挡免| 国产亚洲欧美精品永久| 免费黄频网站在线观看国产| 看非洲黑人一级黄片| 亚洲欧美精品综合一区二区三区| 一级毛片黄色毛片免费观看视频| 久久97久久精品| 午夜福利乱码中文字幕| 可以免费在线观看a视频的电影网站 | av.在线天堂| 观看美女的网站| 日韩 欧美 亚洲 中文字幕| 日韩精品有码人妻一区| 青春草国产在线视频| 一级毛片黄色毛片免费观看视频| 中文字幕高清在线视频| 在线天堂中文资源库| 国产亚洲午夜精品一区二区久久| 97在线人人人人妻| 日韩不卡一区二区三区视频在线| 蜜桃国产av成人99| 国产精品偷伦视频观看了| 欧美国产精品va在线观看不卡| 国产在线免费精品| 精品一区二区三区av网在线观看 | 99热网站在线观看| 女的被弄到高潮叫床怎么办| 亚洲精品,欧美精品| 亚洲精品日本国产第一区| 黄色一级大片看看| 最近中文字幕高清免费大全6| 欧美黑人精品巨大| av网站在线播放免费| 亚洲av在线观看美女高潮| 如日韩欧美国产精品一区二区三区| 一边摸一边抽搐一进一出视频| 欧美最新免费一区二区三区| 午夜福利视频精品| 大香蕉久久网| 最近中文字幕高清免费大全6| 赤兔流量卡办理| 亚洲免费av在线视频| 老汉色∧v一级毛片| 伊人亚洲综合成人网| 美国免费a级毛片| 19禁男女啪啪无遮挡网站| 国产人伦9x9x在线观看| 日韩一区二区视频免费看| 最近最新中文字幕免费大全7| 久久久久久久大尺度免费视频| 最新的欧美精品一区二区| 亚洲精品日本国产第一区| 最近中文字幕2019免费版| 一区二区三区四区激情视频| 老汉色av国产亚洲站长工具| 天美传媒精品一区二区| 午夜福利,免费看| 国产亚洲午夜精品一区二区久久| 国产淫语在线视频| 久久久国产一区二区| 国产亚洲av片在线观看秒播厂| 99热国产这里只有精品6| 性色av一级| 日韩av在线免费看完整版不卡| 亚洲自偷自拍图片 自拍| 亚洲av日韩精品久久久久久密 | 久久97久久精品| 黄片小视频在线播放| xxx大片免费视频| 男女午夜视频在线观看| 国产成人免费观看mmmm| 婷婷成人精品国产| 午夜日本视频在线| 欧美成人午夜精品| 只有这里有精品99| 在线观看www视频免费| 日韩一本色道免费dvd| www日本在线高清视频| 精品国产一区二区三区久久久樱花| 中文字幕av电影在线播放| 国产成人一区二区在线| 色视频在线一区二区三区| 亚洲精品一二三| 高清av免费在线| 欧美精品高潮呻吟av久久| 成人国产麻豆网| 蜜桃国产av成人99| 国产 一区精品| 日韩 亚洲 欧美在线| 久久99一区二区三区| 亚洲综合精品二区| 免费观看av网站的网址| 老司机深夜福利视频在线观看 | 日韩,欧美,国产一区二区三区| 丰满迷人的少妇在线观看| 日韩欧美一区视频在线观看| 午夜久久久在线观看| 搡老岳熟女国产| 一级毛片 在线播放| 精品亚洲乱码少妇综合久久| 国产成人91sexporn| 韩国精品一区二区三区| 色婷婷av一区二区三区视频| 日日啪夜夜爽| 日韩大片免费观看网站| 免费黄频网站在线观看国产| 日韩中文字幕视频在线看片| 日韩视频在线欧美| www.精华液| 国产又爽黄色视频| 精品午夜福利在线看| 亚洲国产毛片av蜜桃av| 少妇的丰满在线观看| 免费黄频网站在线观看国产| av又黄又爽大尺度在线免费看| 久久久国产欧美日韩av| 777米奇影视久久| 亚洲第一av免费看| 2018国产大陆天天弄谢| 80岁老熟妇乱子伦牲交| av一本久久久久| av视频免费观看在线观看| 夫妻午夜视频| 亚洲人成网站在线观看播放| 亚洲国产精品一区三区| 国产探花极品一区二区| 国产成人免费无遮挡视频| 大香蕉久久网| 国产免费一区二区三区四区乱码| 国产有黄有色有爽视频| 在线 av 中文字幕| 在线 av 中文字幕| 亚洲激情五月婷婷啪啪| 亚洲熟女毛片儿| 一区二区三区精品91| 天堂俺去俺来也www色官网| 国产精品国产三级专区第一集| 99热全是精品| 一区在线观看完整版| 亚洲精品成人av观看孕妇| 午夜福利免费观看在线| 亚洲av电影在线观看一区二区三区| 熟妇人妻不卡中文字幕| 亚洲欧美一区二区三区黑人| 亚洲av电影在线观看一区二区三区| 精品亚洲成国产av| 国产黄频视频在线观看| 一本大道久久a久久精品| 亚洲国产精品999| 色婷婷av一区二区三区视频| 老司机在亚洲福利影院| 国产探花极品一区二区| 亚洲美女视频黄频| 一级a爱视频在线免费观看| videosex国产| 激情五月婷婷亚洲| 亚洲av成人不卡在线观看播放网 | 人人妻,人人澡人人爽秒播 | 丝袜美腿诱惑在线| 青春草亚洲视频在线观看| 涩涩av久久男人的天堂| 国产精品成人在线| 纯流量卡能插随身wifi吗| 美女主播在线视频| 人人妻,人人澡人人爽秒播 | 我的亚洲天堂| 女人久久www免费人成看片| 大片免费播放器 马上看| 叶爱在线成人免费视频播放| 热99国产精品久久久久久7| 男女免费视频国产| 亚洲精品成人av观看孕妇| 青草久久国产| 嫩草影视91久久| 日韩制服骚丝袜av| 美女午夜性视频免费| 国产亚洲午夜精品一区二区久久| 免费在线观看完整版高清| 日本av手机在线免费观看| 日本黄色日本黄色录像| 丰满少妇做爰视频| 最近最新中文字幕大全免费视频 | 国产不卡av网站在线观看| 久久综合国产亚洲精品| 国产成人精品福利久久| 亚洲欧美清纯卡通| 日韩人妻精品一区2区三区| 亚洲三区欧美一区| 99精品久久久久人妻精品| 亚洲伊人色综图| 国产精品亚洲av一区麻豆 | 综合色丁香网| 伦理电影大哥的女人| av电影中文网址| 人成视频在线观看免费观看| 在线观看www视频免费| 精品国产乱码久久久久久小说| 亚洲精品一区蜜桃| 高清av免费在线| 国产精品香港三级国产av潘金莲 | 精品亚洲乱码少妇综合久久| 久久久精品免费免费高清| 美女扒开内裤让男人捅视频| 国产xxxxx性猛交| 中文字幕色久视频| 亚洲国产成人一精品久久久| 亚洲熟女精品中文字幕| 欧美成人午夜精品| 男女下面插进去视频免费观看| 咕卡用的链子| www.av在线官网国产| 精品第一国产精品| 秋霞伦理黄片| 国产亚洲最大av| 在线精品无人区一区二区三| 99久久综合免费| 国产免费福利视频在线观看| 麻豆精品久久久久久蜜桃| 日韩伦理黄色片| 一个人免费看片子| 国产精品一区二区在线不卡| 久久久久久久久久久免费av| 亚洲一区中文字幕在线| 91国产中文字幕| 成年av动漫网址| 色吧在线观看| 亚洲视频免费观看视频| 久久国产亚洲av麻豆专区| 美女脱内裤让男人舔精品视频| www日本在线高清视频| 汤姆久久久久久久影院中文字幕| 亚洲国产成人一精品久久久| 精品午夜福利在线看| 亚洲成人国产一区在线观看 | 久久久久久久大尺度免费视频| 亚洲av男天堂| 免费看不卡的av| 一二三四中文在线观看免费高清| 满18在线观看网站| 妹子高潮喷水视频| 国产97色在线日韩免费| 免费看不卡的av| 国产1区2区3区精品| 久久精品国产亚洲av高清一级| 韩国av在线不卡| 亚洲欧洲国产日韩| 午夜福利乱码中文字幕| 啦啦啦在线免费观看视频4| 香蕉丝袜av| 一级黄片播放器| 成人国产麻豆网| 男人添女人高潮全过程视频| 日本av手机在线免费观看| 精品国产露脸久久av麻豆| 狂野欧美激情性bbbbbb| 看非洲黑人一级黄片| 国产日韩欧美在线精品| 色视频在线一区二区三区| 香蕉丝袜av| 麻豆乱淫一区二区| 国产精品秋霞免费鲁丝片| h视频一区二区三区| av在线播放精品| 制服丝袜香蕉在线| 欧美亚洲日本最大视频资源| 成年女人毛片免费观看观看9 | 国产精品av久久久久免费| 久热爱精品视频在线9| 久久久久精品久久久久真实原创| 国产精品香港三级国产av潘金莲 | 久久人人97超碰香蕉20202| 在线观看三级黄色| 精品人妻熟女毛片av久久网站| 一本大道久久a久久精品| 国产黄频视频在线观看| 午夜福利视频在线观看免费| 国产日韩一区二区三区精品不卡| 美女脱内裤让男人舔精品视频| 亚洲精品第二区| 国产麻豆69| 国产av一区二区精品久久| 午夜福利视频精品| 一级,二级,三级黄色视频| 亚洲美女搞黄在线观看| 激情五月婷婷亚洲| 成人亚洲欧美一区二区av| 亚洲欧美日韩另类电影网站| 少妇人妻久久综合中文| 国产午夜精品一二区理论片| 嫩草影院入口| 18在线观看网站| 国产乱人偷精品视频| 久久ye,这里只有精品| 黑丝袜美女国产一区| 99热全是精品| 色吧在线观看| 亚洲,欧美,日韩| 99久国产av精品国产电影| 欧美日韩视频精品一区| 搡老岳熟女国产| 免费观看性生交大片5| 亚洲一码二码三码区别大吗| 久久久久久久国产电影| 国产精品免费大片| 精品酒店卫生间| 狂野欧美激情性bbbbbb| 精品一区二区免费观看| 亚洲国产精品999| 老司机深夜福利视频在线观看 | 又大又爽又粗| 丁香六月天网| 天美传媒精品一区二区| 亚洲精品aⅴ在线观看| 亚洲专区中文字幕在线 | 五月开心婷婷网| 国产黄色视频一区二区在线观看| 亚洲国产av影院在线观看| 亚洲精华国产精华液的使用体验| 欧美乱码精品一区二区三区| 两性夫妻黄色片| 日韩 欧美 亚洲 中文字幕| 啦啦啦视频在线资源免费观看| svipshipincom国产片| 成人亚洲精品一区在线观看| 国产日韩欧美视频二区| 狂野欧美激情性xxxx| 国产在线视频一区二区| 欧美精品亚洲一区二区| 色婷婷av一区二区三区视频| 高清视频免费观看一区二区| 老司机影院成人| 亚洲一卡2卡3卡4卡5卡精品中文| 久久婷婷青草| 伦理电影免费视频| 国产精品二区激情视频| 日韩大片免费观看网站| 一本久久精品| 精品国产乱码久久久久久男人| 国语对白做爰xxxⅹ性视频网站| 香蕉丝袜av| 国产 精品1| 少妇人妻久久综合中文| 亚洲,欧美,日韩| 曰老女人黄片| 操美女的视频在线观看| 男女床上黄色一级片免费看| 老司机亚洲免费影院| 美女主播在线视频| 国产午夜精品一二区理论片| 亚洲精品国产av成人精品| 69精品国产乱码久久久| 亚洲人成网站在线观看播放| 又黄又粗又硬又大视频| 美女主播在线视频| 国产不卡av网站在线观看| 久久精品亚洲熟妇少妇任你| 成人国产麻豆网| 免费在线观看完整版高清| 亚洲精品日本国产第一区| 一区二区av电影网| a级片在线免费高清观看视频| 高清欧美精品videossex| 日韩伦理黄色片| 宅男免费午夜| 成年人午夜在线观看视频| 亚洲国产av新网站| 91精品三级在线观看| 狂野欧美激情性bbbbbb| 可以免费在线观看a视频的电影网站 | 少妇被粗大猛烈的视频| 欧美变态另类bdsm刘玥| 2021少妇久久久久久久久久久| 视频区图区小说| 欧美黑人欧美精品刺激| 国产 精品1| 爱豆传媒免费全集在线观看|