• <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 ?

    Name Relevance and Contact Opportunity-Based Routing Strategy for Mobile Content Sharing

    2018-10-10 06:27:26ChaoLiWeiHeandHuimeiLu

    Chao Li, Wei He and Huimei Lu

    (School of Computer Science and Technology, Beijing Institute of Technology, Beijing 100081, China)

    Abstract: With the rapid advancement of smart devices and mobile computing technologies, content sharing in dynamic wireless networks is in substantial demand. To address this problem, in our former work, we presented the Delay/Disruption-tolerant Mobile Content Sharing Network (DMCSN). In this paper, we further examine the routing strategy under the DMCSN framework. We first design the composite interest for organizing user interests. Then, an interest routing strategy based on incremental name relevance (IRINR) and a content routing strategy based on incremental contact opportunity (CRICO) are devised to correspond to the two stages of the content acquisition process: the spreading of the interests and the returning of the contents. Our simulation results validate the following: first, the composite interest can reduce redundant dissemination of content; second, matching of interests and contents are accomplished in a more cost-effective manner under the guidance of IRINR compared with baseline methods; and finally, the integration of IRINR and CRICO achieves a good balance between cost and performance.

    Key words: opportunistic routing; content-based; name relevance; contact opportunity; composite interest

    Mobile devices have become indispensable in people’s common daily lives because of their portability, versatility and computational capability. However, most mobile users still rely on cellular or Wi-Fi connection to obtain content and rarely share it among peers. On the one hand, free public hotspots are seldom available outdoors, so users usually have to be charged for a large amount of data traffic. On the other hand, a considerable amount of content may reside in devices of neighboring ones, and non-private contents could be made public for sharing, such as news, weather reports, announcements, and so forth. Currently, these massive and diversified resources are not fully exploited. In addition, the limited communication range and inherent mobility of mobile devices make the connectivity in the network constructed by such devices highly intermittent, which is quite challenging for content routing and exchanging. Moreover, although users can utilize device-to-device communication methods, e.g. Bluetooth and Wi-Fi direct, to contact with neighboring people, procedures of content discovery and connection establishment are not automatic.

    To provide a solution to this issue, in our previous work[1], by borrowing properties inherited from the Delay/Disruption Tolerant Network (DTN) as a remedy to highly dynamic network conditions and combining the emerging concept of content-centric networking, we proposed the Delay/Disruption-tolerant Mobile Content Sharing Network (DMCSN) to enhance collaboration among multiple devices in content sharing. The key features of DMCSN are listed below.

    ①Each piece of content possesses a name. Users can request the content by name, thus eliminating the need to know the explicit owner of the content.

    ②Content body and information in bundle header are stored separately without independent bundle encapsulation to avoid redundancy and alleviate pressure on the buffer.

    ③Previous transmissions can reduce costs for later requests, because they leave replicas on intermediate nodes. Requesters can fetch one of the replicas nearby, not necessarily from the source. In this way, contents in the buffer are far more valuable and users could expect to spend less time on waiting.

    ④Two phase routing methods, namely, interest and content routing, are more appropriate and effective for content request, matching, and delivery.

    ⑤Integration of transmission inits is supported to save link capacity.

    In this paper, we further develop an enhanced routing protocol for the DMCSN framework. Our chief focus is to incorporate and utilize the characteristics of content naming and two-phase routing to achieve better performance. To this end, we devise two separate strategies: the first is interest routing based on incremental name relevance(IRINR), which is responsible for matching content based on interest at low cost, and the second is content routing based on incremental contact opportunity (CRICO), whose task is to effectively pass content back to the requester.

    The remainder of this paper is organized as follows. In Section 1, related works are outlined and discussed. In Section 2, we first describe the naming convention based on attribute-value pairs and demonstrate how it works in representation and matching; then we elaborate the IRINR and CRICO routing strategies. Experimental evaluations and correlated analysis are given in Section 3. Finally, we end the paper with the concluding remarks in Section 4.

    1 Related Works

    Content sharing has attracted significant attention from scholars over the past few years. As reflected in these works, most proposals on the Internet or MANET agree on some assumptions, such as stable connections and periodic information exchanging for maintaining status and knowledge of the network. But these assumptions are often violated in mobile wireless scenarios in which, due to active mobility of participating nodes, the publisher and consumer of a content cannot be expected to establish an end-to-end connection under all circumstances. So the “publish/subscribe” paradigm, which decouples the request and content dissemination both temporally and spatially[2], is more suitable for mobile environments where the contacts of nodes are sporadic and intermittent.

    Routing protocols based on the pub/sub model in opportunistic wireless networking are well studied by numerous researchers. Context-based methods, as represented by Refs.[3-4], emphasize the social or contact relationship of users and consider them as the solid bases for determining a better carrier. However, context-oblivious methods, such as the one discussed in Ref.[5], do not need much knowledge about surroundings or the neighboring nodes, nor do they require a data collection phase. Meanwhile, past works like Ref.[6] and Ref.[7] use content popularity as the central metric. Each interest explicitly specifies a demanded piece of content, and this one-on-one relationship is the foundation of calculating content popularity. As it can be seen, the relationship of multiple interests and contents are not taken into consideration. In our work, a more flexible relevance between interests and contents can be quantified and exploited. Of these, Ref.[8] proposes an agent-assisted method, which utilizes nodes that could discover or retrieve content more easily to provide services for others. We can learn that the majority of the aforementioned routing protocols still rely on address to perform routing, i.e., they always try to track the specific nodes instead of seeking desired contents, which is proven to be inefficient for distributed content sharing[9-10].

    Haggle[11]is a search-oriented framework, in which content and interest can be described by users by attaching flexible “attribute-value” labels to them. All labels form a relation graph as the foundation of local search operations. The proportion of matched labels represents the relation strength. Here, the interests of nodes can be learned by far-away counterparts through dissemination. Clearly, the relationship between nodes is quite concentrated on content, and this represents a major difference from traditional routing proposals. In the current work, we are interested in following this new content-centric approach to design an appropriate routing strategy.

    Despite numerous advantages achieved through the principle of “content from anywhere”, such as fast fetching of popular contents and the optimization of bandwidth consumption, the new approach also makes the network susceptible to new cache- and content-based attacks. This problem, also known as “content poisoning”, jeopardizes the safety of users and causes harmful effects to the network. Although it is beyond the scope of this study to solve this problem, reviewing related solutions on attack mitigation is still a noteworthy subject as it may be a chief concern of readers and a future working direction of us.

    A basic and passive method for addressing such a problem is by incorporating a signature in packet[12]. A user can clearly be aware of whether or not the received content is trustworthy by simply verifying the signature using the public key of the producer. However, this method only protects users, to some extent, by enabling them to distinguish and discard poisoned contents. Due to the caching nature of the networking, the contents still exist in the intermediary caches and, if not handled properly, these will be served continuously to more victims. Therefore, more proactive methods are proposed. In Ref.[13], upon detecting poisoned content, nodes will propagate a feedback message upstream to warn the others. Then, the path to the poisoned content is treated as a bad choice. In Ref.[14], a ranking algorithm for cached contents is applied based on statistics collected from consumers. Apart from user reports, other studies theorized that routers or intermediary nodes can undertake inspection jobs along with forwarding. However, performing on-the-fly checking of every packet is a computationally heavy and time-consuming task. To reduce the overhead, Ref.[15] only performs verification on popular content and Ref.[16] employs probabilistic verification during the data packet forwarding process. Furthermore, Ref.[17] notes that well-designed caching strategies also have effects on relieving poisoned content attacks. In delay-tolerant networks, where the network topology is highly dynamic, it is even more difficult to control poisoned contents. In Ref.[18], a node tries to detect poorly behaving nodes by observing the suspicious activities of its neighbors. If a node is considered to be malicious, it will be quarantined, i.e. others will cease to connect with it in the future.

    2 Name Relevance and Contact Opportunity-Based Routing

    2.1 Composite interest

    Interest is an important indicator in content dissemination, as it allows nodes to be aware of the needs of others and forward contents accordingly. In our design, users can express their interests or describe features of contents via attribute-value pairs, and the attribute field can be left empty. The name of interest or content consists of a series of such pairs. We do not assign priorities to pairs and treat them equally for simplicity. We take this flat naming scheme mainly because of two reasons. First, it can provide the flexibility allowing users to show their requirements through arbitrary keywords. In this way, the well-known taxonomy of contents does not become a mandatory prerequisite. Second, the compactness and conciseness of the flat naming scheme are expected, because the name information are carried in each bundle header in forwarding process. Once a name is assigned to a piece of content at the source node, it will stay unchanged unless some modifications are made to the content body. In this way the purity of names is kept and the computational complexity is reduced. This aspect is different from Haggle in which anyone can put labels on content at will.

    However, as the number of requests increases, the interest units are produced in greater quantity. Even after a content is delivered to the requester, corresponding interests would not disappear if they are not yet expired or removed deliberately. This will continuously incur more redundant transfers. It would be quite difficult and implausible to locate all of them and inform owners to remove them immediately under a distributed situation. To alleviate this problem, we propose the concept of “composite interest”. From the viewpoint of form, as Figs.1-2 shown, this is an aggregation of all valid single interests of a node. Using the composite interest, we are able to express multiple individual interests at one time. Hence, the transmission times of interests can be reduced and the local management of interests is much easier. More importantly, the composite interest can function implicitly as a notification message of invalidation. For example, if composite interestP1, which consists of single interestI1andI2, comes at timeT1, then later,P2comes containingI1andI3atT2(T1

    Fig.1 Structure of single interest

    Fig.2 Structure of composite interest

    Fig.3 Structure of response content

    2.2 Content matching

    We define the metric as “name relevance” to measure the matching degree between two objects, one of which can be either an interest or a content. For example, if the name of interestIaisnaand the name of contentCbisnb, then the name relevance is calculated as

    (1)

    which yields a value between 0 and 1. The bigger the value is, the stronger the relation between the interest and content. The cardinality of the name is the number of “attribute-value” pairs it includes. Meanwhile, the result of the intersection is the number of pairs that the two names have in common. Next, we extend the equation to calculate the relevance between two objects containing multiple name entries as Algorithm 1.

    Algorithm1Calculation of relevance between two composite interests

    repeat

    rAB←rAB+(1-rAB)ri,?ri∈T

    T←T-{ri}

    untilTis ?

    The setRis defined as

    R={rij|rij=f(ni,nj),?ni∈A,?nj∈B}

    (2)

    2.3 Interest routing based on the incremental name relevance

    To acquire content, a node should initially announce its interests to let others know what it actually demands. In our strategy, we provide two ways to disseminate interest to avoid flooding interests to the regions where required contents do not likely exist. The first is by broadcasting beacons which include the interests of the node itself. The beacons are forwarded within 1-hop for neighbor discovery and connection establishment. If the desired contents are found at the direct neighbor, then they can be delivered immediately. The second way is by storing and forwarding the interests of other nodes in a multi-hop manner to largely expand the reachable scope of interests.

    With the aim of helping improve the proportion of successful matching and controlling the dissemination of interests within a reasonable area, we propose interest routing based on incremental name relevance (IRINR). Our primary goal is directing interests to nodes, including but not limited to, those with demanded contents. The nodes that have highly similar interests or contents are deemed as good forwarders as well, because they share common preferences and also have high possibility to obtain demanded contents in the near future in their neighborhoods. The composite interest can facilitate the finding of such nodes as it can reflect the overall requirements of the source node. Moreover, with the name relevance metric, the interests have more opportunities to approach the regions, in which the contents needed possibly reside, along the direction of increasing name relevance without the location knowledge of node or content.

    For example, we presume the names of composite interestI1andI2aren1andn2, respectively. Then, according to Eq.(1), the relevance betweenn1andn2is given by

    (3)

    If contentC3, whose name isn3, completely matchesI2, it means that

    (4)

    Then, we can have

    n2∩n3=n2

    (5)

    Furthermore,

    n1∩n2∩n3=n1∩n2

    (6)

    From Eqs.(1) and (6), we have

    Therefore, we can deduce that

    p3=r13≥p1=r12

    (7)

    We can conclude that, if the source ofI2getsC3, then the probability thatC3can satisfyI1is greater than the relevance ofI1andI2. This indicates that interests should be forwarded along the direction of higher relevance to approach the expected contents.

    The procedures of how nodes forward interests to neighbors and how they react to the incoming interests are given in Algorithms 2 and 3.

    Algorithm2Procedure of forwarding the interests of other nodes in IRINR

    localInterestList=getLocalInterestList()

    for eachiin local Interest List do

    ifi.source ≠ localNode then

    neighborInterest = getInterest(currentNeighbor)

    selfInterest = getInterest(localNode)

    if relevance(i, neighborInterest) > relevance(i, selfInterest) then

    sendInterest(i)

    end if

    end if

    end for

    Algorithm3Procedure of receiving the interests in IRINR

    incomingInterest = receiveInterest()

    if localExistInterest(incomingInterest.source) then

    localInterest = getInterest(incomingInterest.source)

    if incomingInterest.timestamp > localInterest.timestamp then

    updateLocalInterest(incomingInterest)

    updatelocalMatchingRecords()

    end if

    end if

    2.4 Content routing based on incremental contact opportunity

    When two nodes initiate a connection, all local matching records, content digest, and predictions of contacting other nodes will be exchanged. The exchanging of content digest will clarify to both sides what useful contents they could obtain from each other. The content digest is valid only during the current session and will not be passed to the third party. Next, both sides will take the same operations. Here we just take one side to explain. The node will first receive the matching records from the neighbor. For the interest in each record, if the current node has not seen it before or has an older version of it, then an update should be applied. Otherwise, if the interest the current node has is relatively new and the content in the record cannot satisfy any of the interests of the current node, this indicates the matching information provided by the neighbor may be outdated, so the content will not be retrieved. In this way, only contents that can satisfy the latest version of interests can be forwarded. Next, if the content is not stored locally, and if the current node has a higher probability to meet the destination of compared with the neighbor, the content will be pulled back. For those contents that are already present locally, only corresponding matching records will be updated. Matching will also be performed upon the arrival of new content to satisfy the needs of more awaiting interests. To calculate contact opportunity, prediction-based routing methods, which have been extensively studied in DTN research[19-21], can be applied to our framework. However, our method differs from that proposed by Haggle in terms of routing in that Haggle uses an identical method for forwarding interest and content, but here we use separate methods for different purposes. In addition, Haggle considers contents as interests and may misguide them to sources of similar content instead of to the actual requesters. By contrast, we make content gradually come closer to the nodes that really need them. The complete procedure of CRICO is given in Algorithm 4.

    Algorithm4Procedure of CRICO

    matchingRecords = receiveMatchingRecords()

    check ← false

    for eachrin matchingRecords do

    incomingInterestInfo = r.getInterestInfo()

    incomingContentInfo = r.getContentInfo()

    if localExistInterest(interestInfo.source) then

    localInterestInfo = getInterest(interestInfo.source)

    if localInterestInfo.timestamp > incomingInterestInfo.timestamp then

    check ← true

    else

    updateLocalInterest(incomingInterestInfo)

    end if

    end if

    acceptContent ← true

    if check then

    if not matching(incomingContentInfo, localInterestInfo) then

    acceptContent ← false

    end if

    end if

    if localExistContent(incomingContentInfo) then

    acceptContent ← false

    end if

    if acceptContent and hashigherContactProbability(currentNeighbor, incomingInterestInfo.source) then

    receiveContent(receivedContentInfo)

    matchingLocalInterests(incomingContentInfo)

    end if

    updateLocalMatchingRecords()

    end for

    3 Experiments and Results

    3.1 Experimental setup

    In our experiments, we adopt a modified version based on ONE[22](version 1.4.1) as the simulator. The main function we implemented is the support of content naming. The global parameters and settings are given in Tab.1. Here, we give a brief explanation on the reasoning for these parameters. We assumed that the communication interface that each node is equipped with is a widely used IEEE 802.11g Wi-Fi interface. IEEE 802.11g works at a 2.4 GHz band; it can provide transmission speed of up to 54 Mb/s and transmit at a maximum range of 100 m[23]. However, in reality, the speed and range of the wireless radio are often significantly influenced by the surrounding environment. In order not to lose generality, we set the transmission speed to 25 Mb/s and the transmission range to 30 m. The moving velocity of the nodes is set to 0.5-1.5 m/s, which is the normal speed for walking pedestrians used by numerous past studies[24-26]. The default map of ONE simulator is used, which depicts a section of downtown Helsinki, Finland. Nodes move along roads on the map and Dijkstra’s algorithm is used to calculate the shortest path from the current location to a randomly selected destination according to the shortest path map-based movement model.

    Tab.1 Experiment parameters and settings

    We compare our design with two baseline methods (i.e., the direct method and the flooding method)in order to exclude the impact of other algorithms and to discover the real defects and benefits of our work tentatively. In this part, the buffer size is set to infinite to avoid introducing too many influential factors simultaneously. We shall take this factor into consideration in the following part. In the following experiments, we evaluate the different combinations of interest and content routing components as listed in Tab.2 extensively and present the temporal result analysis.

    3.2 Evaluation metrics

    The effectiveness and cost are two major aspects of evaluating the advantages and disadvantages of routing algorithms. Therefore, we use several metrics in each of them. The description of metrics is shown in Tab.3. The metrics “Efficiency of Content Dissemination”, “Fulfillment Ratio of Interests”, and “Average Content Holding” are defined in the following equations respectively.

    (8)

    (9)

    (10)

    In these equations,Nfis the number of successfully fulfilled interests,Niis the total number of interests,Ciis the number of content items the nodeiowns,Cis the total number of distinct content items,Nis the total number of nodes, andFiis the forwarded times of contenti(including all replicas).

    Tab.2 Routing methods for comparison

    Tab.3 Evaluation metrics

    3.3 Evaluation of the composite interest

    In this section, we use the flooding method both as the interest and content routing component to evaluate the performance of composite interest compared with the single interest. As shown in Fig.4, at the beginning, neither content nor interest appears in the network. This results in the ECD being 0 before 1 000 s. During 1 000 s and 1 800 s, no content successfully returns to the requester, so it still remains 0. With the progress of the simulation, this value begins to fluctuate severely at approximately 2 000 s. This can be attributed to the fact that the denominator, i.e., the forwarded times of contents, is too small to resist sudden change. After 3 000 s, this value becomes stable and indicates that the composite interest surpasses the single interest on the ECD for at least 10%. In other words, the redundant transmission of content is reduced, thus the efficiency and the precision of dissemination are enhanced. In the following experiments, composite interest is applied as the default type of interest.

    Fig.4 Comparison of single interest and composite interest on ECD

    3.4 Evaluation of IRINR

    In this experiment, in order to observe the performance of IRINR, we adopt the flooding method as the content routing component, although we adopt different interest routing methods as the interest routing component. The scheme settings are shown in Tab.4. As shown in Fig.5, with regards the sum of forwarded interests, IRINR is slightly more than the direct scheme and is far less than the flooding scheme. Associating with Fig.6, in which the FRI of IRINR can compete with that of the flooding scheme, we can say that it is fairly cost-effective. Another point in Fig.5 that needs explanation is that, a few interests are already forwarded before 1 000 s. These interests are empty beacons used for discovering neighbors. In another viewpoint, they are special empty interests announcing that their hosts do not desire anythingat that moment.

    Tab.4 Scheme settings for the comparison of interest routing

    Fig.5 Comparison of interest schemes on exchanged interests

    Fig.6 Comparison of interest schemes on FRI

    3.5 Evaluation of CRICO

    In this experiment, we use the prediction model in Prophet[21]as the core of CRICO. We adopt the flooding method as the interest routing component, although we use different content routing methods as the content routing component to clarify the performance of CRICO. The scheme settings are shown in Tab.5. The flooding scheme exchanges contents in an uncontrolled manner; hence, it can reach high FRI in a short time, as shown in Fig.7. Meanwhile, the buffer resource is exhausted rapidly as depicted in Fig.8. Comparatively, the increase speed of the FRI of CRICO is not so fast, but it saves approximately 30%-40% of buffer space, on average. In addition, most of the interests can be fulfilled in an acceptable delay. Take the 60% line in Fig.7 as example, the elapsed time before overall FRI reaches 60% of CRICO is approximately 10 000 s, which is only 30% more than that of the flooding scheme (approximately 6 600 s), and on the other hand is greatly less than that of the direct scheme (approximately 27 000 s). In Fig.9, we bring the results of CRICO on FRI and ACH together and discover that the increase speed on FRI is faster than that of ACH, indicating that a small addition in storage cost can pay more in terms of the successful content deliveries by CRICO.

    Tab.5 Scheme settings for the comparison of content routing

    Fig.7 Comparison of the content schemes on FRI

    Fig.8 Comparison of the content schemes on ACH

    Fig.9 Comparison of the CRICO scheme on FRI and ACH

    3.6 Evaluation of INTR

    In this section, we integrate IRINR and CRICO and label it as INTR, to assess their performance as a whole. The scheme settings are shown in Tab.6. Fig.10 illustrates that, with respect to FRI, the flooding scheme still has superiority over other approaches. The INTR also behaves well and has a good trend: it ascends almost linearly under 80% and achieves final results that close to the flooding scheme at the end of the simulation. Figs.11-12 demonstrate that INTR has an effect on controlling overhead. First, the forwarded times of interests are less than a half of those in the flooding scheme, indicating only selected interests are forwarded towards promising directions. Correspondingly, the ACH of INTR is maintained at a very low level, suggesting that redundant content exchange is suppressed and a considerable amount of buffer space could be saved. To summarize, the integration of IRINR and CRICO is effective in achieving a good balance between performance and cost.

    Tab.6 Scheme settings for the evaluation of INTR

    Fig.10 Evaluation of the INTR scheme on FRI

    3.7 Evaluation under limited buffer size and shorter time

    In this section, we will evaluate the performance of INTR with the constraints of limited buffer size and shorter simulation time,in order to determine whether INTR performs well under these conditions. The scheme settings are the same as those shown in Tab.6 in Section 3.6. The parameters are unchanged as in Tab.1 in Section 3.1, with the exception of the simulation time being reduced from 36 000 s (10 h) to 21 600 s (6 h). The buffer size ranges from 100-200 MB, which we believe is acceptable for most user devices. The replacement policy is to remove the oldest content first. Each experiment is repeated for 10 times with different trajectories of nodes. The result presented is the mean number of all the derived values.

    Fig.11 Evaluation of the INTR scheme on exchanged interests

    Fig.12 Evaluation of the INTR on ACH

    From Fig.13, we can see that the buffer size has great impact on the delivery ratio of the flooding scheme. With the decreasing buffer size, the delivery ratio of flooding scheme declines evidently. The delivery ratio of the direct scheme is not influenced much by buffer size changing because it does not need much buffer resources to work and thus has a relatively low delivery ratio. Notably, the INTR surpasses the flooding scheme when the buffer resource is in a shortage. This indicates that the INTR can utilize limited buffer size with greater efficiency in order to deliver more contents. Meanwhile, augmenting the buffer size has a positive effect on the INTR as the delivery ratio increases gradually with the increase in buffer size. In addition, INTR shows better adaptability to the limitation of buffer size, because the difference of the delivery ratio between the maximum and the minimum buffer sizes is only approximately 7%.

    Fig.13 Delivery ratio with different buffer sizes

    In terms of average delay, as shown in Fig.14, the flooding scheme is the lowest among the three schemes. It can be observed that the average delay goes up first and then goes down with the increase in buffer size. This is mainly due to several reasons. ① When the buffer size is small, the stay duration of contents is very short as they are to be replaced soon. As such, those contents that are near the requesters have a higher possibility of successful delivery. Therefore, the average delay is low at the beginning. ② With the increasing buffer size, each node can possess more contents. This makes it possible for more deliveries to be completed at a later time, which contributes to the higher value of the average delay. ③ With the further increase in buffer size, due to the rapidly spreading nature of the flooding scheme, the possibility of demanded contents locating in the local buffer or nearby nodes also increases. Therefore, the average delay keeps decreasing when the buffer size is larger than 140 MB due to such immediate deliveries. The average delay of the direct scheme is also unaffected by variations in the buffer size. For the direct scheme, the buffer size is sufficient and the delay largely correlates with the routes and movements of nodes, for which we use the same 10 sets data across different experiment groups of buffer size. The average delay of the INTR is the highest of all schemes: it keeps increasing gradually as the buffer size increases. From this point, combined with the result of the delivery ratio, we can see that INTR can find and deliver contents efficiently in the long run. The longer delay reveals that, in the INTR, contents that are located far from the requesters or need multiple relays to be found and delivered are well-supported. This point is more clearly observable when the buffer size is limited, as we know that the flooding scheme provides poor support for such far-away contents. This finding validates the correctness of IRINR, as it can find demanded contents in potential directions. The increasing buffer size further strengthens such abilities.

    Fig.14 Average delay with different buffer sizes

    Next, we examine the results from the aspect of the amount of content. Tabs.7-8 provide the comparison of the transmitted and removed content amounts, respectively. As can be seen, the flooding scheme has an overwhelming amount of transmitted data, but most of them are removed. The severe turbulence greatly affects the performance of the flooding scheme and leads to wasted link resources when the buffer size is small. In comparison, when the buffer size becomes larger, as stated previously, local or nearby content replicas relieve such phenomena. In the direct scheme, both the transmitted and removed content amounts are extremely low as contents are exchanged in an overly conservative way. Therefore, the direct scheme does not fully utilize the buffer resource. As for the INTR, we can see that the transmitted content amount is only approximately 10% of the flooding scheme, and the removed content amount is only approximately 5% of the flooding scheme. This demonstrates the coaction effects of both the IRINR and CRICO. To avoid a great number of content replicas being generated, IRINR only selectively disseminates interests based on the matching results. Meanwhile, CRICO only forwards contents with a greater delivery possibility to the next hop. The results given in Fig.15 indicate that the average content number possessed by each node in INTR is only a half that of the flooding scheme. This is a reasonable buffer use level for each node, because each node has a number of contents and also needs to simultaneously keep a portion of available space for incoming valuable contents. These results demonstrate the effectiveness of the INTR in controlling the amount of content in the network and its efficiency in maximizing the utilization of limited link and buffer resources.

    Tab.7 Average transferred content amount with different buffer sizes

    Tab.8 Average removed content amount with different buffer sizes

    Fig.15 Average number of contents in each node with different buffer sizes

    To conclude this section, we briefly review the key findings below.

    ①The INTR can achieve a better delivery ratio when the buffer size is limited. The overall performance is not influenced greatly by the buffer variance.

    ② The longer average delay of the INTR reveals that INTR supports multi-hop content finding and delivery processes more efficiently; it does so by sending selected interests towards the potential locations of demanded contents and by exchanging more valuable contents.

    ③The INTR is more storage-efficient compared with other schemes and is capable of efficiently utilizing the link and buffer resources with less content transfer and dropping.

    4 Conclusions

    In this paper, based on DMCSN, the framework we have presented for improving content sharing in wireless network, we expanded on our past research by completing several tasks. First, we use the attribute-value pairs to describe the names of interests and contents. This method is flexible and allows us to quantify and exploit the relevance of multiple names. Then, we propose the composite interest structure that is used to express the interests of users. This structure facilitates both transmission and management, while simultaneously serving as an implicit acknowledgement to suppress the redundant transmission of successfully delivered contents. Next, we design the IRINR for the routing of interest. The main feature of the IRINR is that it can lead interests towards potential nodes, which either possess demanded contents or have highly relevant interests or contents, based on the matching results. Finally, we use the classical routing method prophet as the core of CRICO to function as the content routing and combine it with the IRINR to form the complete routing solution, which we call INTR.

    We obtained several findings from the experiments. First, composite interest can reduce the redundant transmission of contents. Second, the IRINR can effectively guide interests to existing or potential content locations. Finally, the INTR maintains a good balance between cost and performance. It also shows better tolerance on limited buffer size, can provide better support for matching across multi-hop, and is capable of utilizing link and storage resources more efficiently.

    久久精品夜色国产| 成人欧美大片| 黄色配什么色好看| 亚洲欧美精品专区久久| 亚洲av福利一区| 好男人视频免费观看在线| 成年免费大片在线观看| 麻豆成人av视频| 亚洲精品乱码久久久久久按摩| 我的女老师完整版在线观看| 国产av国产精品国产| 青春草国产在线视频| 男女无遮挡免费网站观看| 搞女人的毛片| 精品人妻视频免费看| 啦啦啦在线观看免费高清www| 97热精品久久久久久| 中文字幕亚洲精品专区| 日韩国内少妇激情av| 午夜福利在线在线| 舔av片在线| 亚洲四区av| 极品少妇高潮喷水抽搐| 精品国产露脸久久av麻豆| 最新中文字幕久久久久| 一二三四中文在线观看免费高清| 欧美高清性xxxxhd video| 内射极品少妇av片p| 国产精品国产三级专区第一集| 国产黄a三级三级三级人| 99久久精品国产国产毛片| 新久久久久国产一级毛片| 看十八女毛片水多多多| 两个人的视频大全免费| 亚洲欧美一区二区三区黑人 | 丰满少妇做爰视频| 日本猛色少妇xxxxx猛交久久| 久久精品久久久久久噜噜老黄| 国产真实伦视频高清在线观看| 亚洲真实伦在线观看| 黄片无遮挡物在线观看| eeuss影院久久| 日韩制服骚丝袜av| 国产精品久久久久久精品电影| 一区二区三区免费毛片| 亚洲婷婷狠狠爱综合网| 国产亚洲精品久久久com| 国产片特级美女逼逼视频| 国产成人aa在线观看| 国产精品国产三级国产专区5o| 国产成人aa在线观看| 欧美3d第一页| 亚洲最大成人手机在线| 一区二区三区四区激情视频| 欧美国产精品一级二级三级 | 2021少妇久久久久久久久久久| 偷拍熟女少妇极品色| 久久久久久久午夜电影| 综合色av麻豆| 嫩草影院新地址| 嫩草影院新地址| 插阴视频在线观看视频| 九九久久精品国产亚洲av麻豆| 国产精品国产三级专区第一集| 精华霜和精华液先用哪个| 又大又黄又爽视频免费| 亚洲怡红院男人天堂| 日韩不卡一区二区三区视频在线| 亚洲av.av天堂| 男女国产视频网站| 99九九线精品视频在线观看视频| 一个人观看的视频www高清免费观看| 91在线精品国自产拍蜜月| 午夜亚洲福利在线播放| 日韩亚洲欧美综合| 国产老妇伦熟女老妇高清| 男人爽女人下面视频在线观看| 亚洲精品一区蜜桃| av福利片在线观看| 自拍欧美九色日韩亚洲蝌蚪91 | 亚洲av二区三区四区| 最新中文字幕久久久久| 亚洲四区av| 精品视频人人做人人爽| 久久久亚洲精品成人影院| 国产亚洲午夜精品一区二区久久 | 成人黄色视频免费在线看| 亚洲av一区综合| 啦啦啦啦在线视频资源| 看免费成人av毛片| 免费电影在线观看免费观看| 毛片女人毛片| 亚洲人成网站在线观看播放| 深夜a级毛片| 日韩成人av中文字幕在线观看| 一级av片app| 内射极品少妇av片p| 久久精品国产自在天天线| 国产成人a∨麻豆精品| 女人被狂操c到高潮| 久久亚洲国产成人精品v| 午夜福利在线在线| 校园人妻丝袜中文字幕| 国产v大片淫在线免费观看| 国产一区二区三区综合在线观看 | 精品国产露脸久久av麻豆| 一级毛片aaaaaa免费看小| 王馨瑶露胸无遮挡在线观看| 日韩免费高清中文字幕av| 久久久久性生活片| 欧美性猛交╳xxx乱大交人| 久久久久久国产a免费观看| 久久久色成人| 在线亚洲精品国产二区图片欧美 | 国产大屁股一区二区在线视频| 亚洲av中文字字幕乱码综合| 尾随美女入室| 亚洲av欧美aⅴ国产| 97人妻精品一区二区三区麻豆| 国产精品国产三级国产av玫瑰| 听说在线观看完整版免费高清| 熟女人妻精品中文字幕| av在线老鸭窝| 亚洲成色77777| 亚洲自拍偷在线| 少妇人妻一区二区三区视频| 在线观看三级黄色| 哪个播放器可以免费观看大片| 精华霜和精华液先用哪个| 国产精品人妻久久久影院| 亚洲欧洲日产国产| 麻豆国产97在线/欧美| 一级毛片电影观看| 91久久精品国产一区二区成人| 麻豆国产97在线/欧美| 久久综合国产亚洲精品| 超碰av人人做人人爽久久| 王馨瑶露胸无遮挡在线观看| 国产精品一二三区在线看| 晚上一个人看的免费电影| 另类亚洲欧美激情| 成年女人在线观看亚洲视频 | 国产片特级美女逼逼视频| 免费看不卡的av| av网站免费在线观看视频| 亚洲精品国产色婷婷电影| 免费看av在线观看网站| 国产成人精品久久久久久| 午夜激情久久久久久久| 又爽又黄无遮挡网站| 女人久久www免费人成看片| 各种免费的搞黄视频| 亚洲欧美中文字幕日韩二区| av线在线观看网站| 亚洲成人久久爱视频| 亚洲人成网站高清观看| 亚洲av在线观看美女高潮| 中文字幕免费在线视频6| 韩国av在线不卡| 中文资源天堂在线| 亚洲国产日韩一区二区| 亚洲精品亚洲一区二区| 成人毛片60女人毛片免费| 欧美日韩精品成人综合77777| 一级二级三级毛片免费看| 男人和女人高潮做爰伦理| 精品人妻一区二区三区麻豆| 在线a可以看的网站| 国产真实伦视频高清在线观看| 国产成人精品婷婷| 欧美性猛交╳xxx乱大交人| 丰满少妇做爰视频| 简卡轻食公司| 99热这里只有精品一区| 国产白丝娇喘喷水9色精品| av黄色大香蕉| 日本三级黄在线观看| 搞女人的毛片| 国产精品偷伦视频观看了| 国产男女内射视频| 国产亚洲午夜精品一区二区久久 | 国精品久久久久久国模美| 69人妻影院| 色视频在线一区二区三区| 亚洲欧美清纯卡通| 蜜桃久久精品国产亚洲av| a级一级毛片免费在线观看| 尾随美女入室| 午夜日本视频在线| 男女国产视频网站| 热99国产精品久久久久久7| 免费在线观看成人毛片| 欧美 日韩 精品 国产| 国产 一区 欧美 日韩| 国产伦精品一区二区三区四那| 嫩草影院入口| 日本与韩国留学比较| 又大又黄又爽视频免费| 亚洲婷婷狠狠爱综合网| 熟女人妻精品中文字幕| 欧美3d第一页| 久久久久久久国产电影| 欧美97在线视频| 777米奇影视久久| 欧美激情在线99| 午夜福利视频1000在线观看| 国产精品一二三区在线看| 在线观看人妻少妇| 中国美白少妇内射xxxbb| 久久久久久久久久成人| 久久99精品国语久久久| 日韩成人av中文字幕在线观看| 女的被弄到高潮叫床怎么办| 国产伦在线观看视频一区| 精品人妻偷拍中文字幕| 搞女人的毛片| 亚洲欧美精品专区久久| 欧美zozozo另类| 亚洲图色成人| 制服丝袜香蕉在线| 久久ye,这里只有精品| 免费看日本二区| .国产精品久久| 久久人人爽人人爽人人片va| 久久精品国产亚洲网站| 99re6热这里在线精品视频| 亚洲av.av天堂| 亚洲成人久久爱视频| 国产一区二区三区综合在线观看 | 免费少妇av软件| 亚洲精品亚洲一区二区| 午夜福利视频精品| 免费看日本二区| 91精品一卡2卡3卡4卡| 一级毛片黄色毛片免费观看视频| 欧美xxⅹ黑人| 国产一区二区亚洲精品在线观看| 免费看av在线观看网站| 最近中文字幕2019免费版| 七月丁香在线播放| 欧美zozozo另类| 亚洲精品456在线播放app| 国产免费一级a男人的天堂| 亚洲色图综合在线观看| 日韩一区二区视频免费看| 亚洲精品成人久久久久久| 久久久久网色| 最近中文字幕2019免费版| 视频中文字幕在线观看| 赤兔流量卡办理| 国产精品国产三级国产av玫瑰| 国产一级毛片在线| 2022亚洲国产成人精品| 春色校园在线视频观看| 看十八女毛片水多多多| 亚洲色图综合在线观看| 高清视频免费观看一区二区| 一级片'在线观看视频| 高清毛片免费看| 国产精品偷伦视频观看了| 中文乱码字字幕精品一区二区三区| 国产黄片美女视频| 久久99热这里只频精品6学生| 国产探花在线观看一区二区| 亚洲av免费在线观看| 特大巨黑吊av在线直播| 中文欧美无线码| 老司机影院成人| 波野结衣二区三区在线| av国产精品久久久久影院| 另类亚洲欧美激情| 97在线视频观看| 内地一区二区视频在线| 免费观看a级毛片全部| 丝袜脚勾引网站| 一个人看的www免费观看视频| 一级毛片久久久久久久久女| 欧美日本视频| 一级毛片电影观看| 秋霞伦理黄片| 国产成人福利小说| 欧美97在线视频| 精品一区二区免费观看| 丝瓜视频免费看黄片| 国产极品天堂在线| 国产美女午夜福利| 国产免费福利视频在线观看| 大香蕉97超碰在线| 精品国产露脸久久av麻豆| 国精品久久久久久国模美| 欧美老熟妇乱子伦牲交| 激情 狠狠 欧美| 国产免费又黄又爽又色| 成人国产av品久久久| 男的添女的下面高潮视频| 一级毛片我不卡| 日本熟妇午夜| 成人美女网站在线观看视频| 国产成人91sexporn| 特大巨黑吊av在线直播| 亚洲精品成人久久久久久| 免费在线观看成人毛片| 午夜福利在线在线| 国产精品国产三级国产专区5o| 91久久精品电影网| 久久久色成人| 两个人的视频大全免费| av女优亚洲男人天堂| 精品国产一区二区三区久久久樱花 | 91午夜精品亚洲一区二区三区| 亚洲高清免费不卡视频| 国产日韩欧美在线精品| 两个人的视频大全免费| 午夜福利网站1000一区二区三区| 噜噜噜噜噜久久久久久91| 欧美高清性xxxxhd video| 美女被艹到高潮喷水动态| 久久精品国产自在天天线| 99久久精品国产国产毛片| 可以在线观看毛片的网站| 亚洲av二区三区四区| 啦啦啦在线观看免费高清www| 啦啦啦啦在线视频资源| 国产精品.久久久| 自拍欧美九色日韩亚洲蝌蚪91 | 亚洲精品乱码久久久久久按摩| 91久久精品国产一区二区成人| 精品视频人人做人人爽| 欧美日韩精品成人综合77777| 老司机影院成人| 大香蕉久久网| av又黄又爽大尺度在线免费看| 水蜜桃什么品种好| 国产精品一区二区三区四区免费观看| 日韩av免费高清视频| a级毛色黄片| 在线天堂最新版资源| 99九九线精品视频在线观看视频| 欧美成人午夜免费资源| 成人毛片60女人毛片免费| 免费观看无遮挡的男女| 欧美97在线视频| 日韩伦理黄色片| 国产老妇伦熟女老妇高清| 久久久久久久午夜电影| av女优亚洲男人天堂| 国产精品人妻久久久久久| 美女主播在线视频| 欧美成人一区二区免费高清观看| 男人添女人高潮全过程视频| 国产成人精品久久久久久| 国产一区亚洲一区在线观看| 51国产日韩欧美| 亚洲av二区三区四区| 成人特级av手机在线观看| 别揉我奶头 嗯啊视频| 亚洲精品视频女| 国产综合精华液| 久久久久久久亚洲中文字幕| 欧美日韩视频精品一区| 天天躁日日操中文字幕| 成人美女网站在线观看视频| 成人黄色视频免费在线看| 男人和女人高潮做爰伦理| av专区在线播放| freevideosex欧美| 精品久久久久久久末码| 国产色爽女视频免费观看| 免费高清在线观看视频在线观看| 久久97久久精品| 欧美潮喷喷水| 亚洲欧洲国产日韩| 少妇 在线观看| av福利片在线观看| 日韩强制内射视频| 日日啪夜夜爽| 亚洲av国产av综合av卡| 国产免费视频播放在线视频| 亚洲在线观看片| 亚洲欧美日韩无卡精品| 色播亚洲综合网| 国产又色又爽无遮挡免| 日本wwww免费看| 亚洲欧美日韩另类电影网站 | 亚洲天堂国产精品一区在线| 国内少妇人妻偷人精品xxx网站| 视频中文字幕在线观看| 国产日韩欧美在线精品| 哪个播放器可以免费观看大片| 成年人午夜在线观看视频| 天天一区二区日本电影三级| 国产黄a三级三级三级人| 99热这里只有是精品在线观看| 久久女婷五月综合色啪小说 | 欧美 日韩 精品 国产| 国产高潮美女av| 国产av国产精品国产| 另类亚洲欧美激情| 自拍欧美九色日韩亚洲蝌蚪91 | 精品一区在线观看国产| 日本一本二区三区精品| 精华霜和精华液先用哪个| 国产黄色视频一区二区在线观看| av福利片在线观看| 嫩草影院新地址| 日韩一区二区视频免费看| 日日撸夜夜添| 国产v大片淫在线免费观看| 日韩人妻高清精品专区| 最后的刺客免费高清国语| 欧美+日韩+精品| 国产一区二区亚洲精品在线观看| 成年免费大片在线观看| 成年女人看的毛片在线观看| 亚洲精品国产色婷婷电影| 中文字幕久久专区| 99久久精品热视频| 美女高潮的动态| 亚洲国产高清在线一区二区三| 免费观看无遮挡的男女| 亚洲丝袜综合中文字幕| 精华霜和精华液先用哪个| 丝袜脚勾引网站| 插逼视频在线观看| 国产av国产精品国产| 欧美老熟妇乱子伦牲交| 亚洲自偷自拍三级| 免费黄网站久久成人精品| 亚洲精品久久久久久婷婷小说| 男插女下体视频免费在线播放| 91狼人影院| 亚洲内射少妇av| 最近最新中文字幕大全电影3| 3wmmmm亚洲av在线观看| 国产精品99久久99久久久不卡 | 小蜜桃在线观看免费完整版高清| 草草在线视频免费看| 麻豆乱淫一区二区| 麻豆久久精品国产亚洲av| 99热这里只有是精品50| 91午夜精品亚洲一区二区三区| 欧美日韩国产mv在线观看视频 | 肉色欧美久久久久久久蜜桃 | 夫妻性生交免费视频一级片| 久久99热这里只频精品6学生| 日韩在线高清观看一区二区三区| 高清av免费在线| 国产综合懂色| 欧美日韩视频精品一区| 狂野欧美激情性bbbbbb| 亚洲国产高清在线一区二区三| 久久韩国三级中文字幕| 伦精品一区二区三区| 久久久久久国产a免费观看| 国产大屁股一区二区在线视频| 超碰av人人做人人爽久久| 亚洲欧美清纯卡通| 久久鲁丝午夜福利片| 少妇的逼好多水| 新久久久久国产一级毛片| 美女主播在线视频| 三级国产精品片| 中文天堂在线官网| 日韩欧美精品免费久久| 内射极品少妇av片p| 91久久精品国产一区二区三区| 国产一区有黄有色的免费视频| 99久久精品热视频| 91精品国产九色| 免费看a级黄色片| 免费黄色在线免费观看| 国产免费福利视频在线观看| 69av精品久久久久久| 亚洲天堂国产精品一区在线| 久久99蜜桃精品久久| 国产中年淑女户外野战色| 欧美日韩一区二区视频在线观看视频在线 | 欧美少妇被猛烈插入视频| 丝袜喷水一区| 久久精品久久久久久噜噜老黄| 又爽又黄无遮挡网站| 欧美亚洲 丝袜 人妻 在线| 下体分泌物呈黄色| 午夜福利在线观看免费完整高清在| 亚洲欧美日韩卡通动漫| 久久精品国产亚洲av涩爱| 美女xxoo啪啪120秒动态图| 免费av观看视频| 午夜激情久久久久久久| videos熟女内射| 精品人妻视频免费看| 国产精品一二三区在线看| 在线观看av片永久免费下载| 高清欧美精品videossex| av福利片在线观看| 久久韩国三级中文字幕| 亚洲国产日韩一区二区| 欧美一区二区亚洲| 国产成人午夜福利电影在线观看| 国产男女内射视频| 99热这里只有是精品50| 我的女老师完整版在线观看| 高清毛片免费看| 搡女人真爽免费视频火全软件| 美女视频免费永久观看网站| 寂寞人妻少妇视频99o| 一区二区三区四区激情视频| 亚洲国产精品999| 久久久久久国产a免费观看| 国产一区有黄有色的免费视频| 我的女老师完整版在线观看| 国产乱人偷精品视频| 99热网站在线观看| 精品人妻一区二区三区麻豆| 亚洲怡红院男人天堂| 特大巨黑吊av在线直播| 亚洲精品日韩av片在线观看| 青春草国产在线视频| 亚洲自拍偷在线| 免费看日本二区| 国产一区二区三区av在线| 免费看日本二区| 97人妻精品一区二区三区麻豆| 欧美 日韩 精品 国产| 99热国产这里只有精品6| 亚洲美女搞黄在线观看| 久久国产乱子免费精品| 免费看av在线观看网站| 亚洲欧美日韩无卡精品| 如何舔出高潮| 免费播放大片免费观看视频在线观看| 久久久久久久大尺度免费视频| 3wmmmm亚洲av在线观看| 中文字幕亚洲精品专区| 日韩中字成人| 亚洲人成网站在线播| 久久久成人免费电影| 精品亚洲乱码少妇综合久久| 久久久成人免费电影| av黄色大香蕉| 七月丁香在线播放| 国产色爽女视频免费观看| 亚洲国产欧美在线一区| 男女无遮挡免费网站观看| 国产精品三级大全| 国产精品人妻久久久久久| 深爱激情五月婷婷| 国产成人一区二区在线| 深爱激情五月婷婷| 看黄色毛片网站| 国产精品爽爽va在线观看网站| 熟女av电影| 一级毛片久久久久久久久女| 久久精品国产亚洲av涩爱| 欧美激情在线99| av天堂中文字幕网| 欧美xxxx黑人xx丫x性爽| 免费av不卡在线播放| 国产成人a∨麻豆精品| 国产精品.久久久| 亚洲精品影视一区二区三区av| 99精国产麻豆久久婷婷| 美女cb高潮喷水在线观看| 在线a可以看的网站| 毛片女人毛片| 成人毛片60女人毛片免费| 午夜激情福利司机影院| 男人舔奶头视频| 亚洲国产精品成人综合色| 亚洲三级黄色毛片| 午夜日本视频在线| 午夜福利视频精品| 黄色欧美视频在线观看| 国产午夜精品久久久久久一区二区三区| 一区二区三区乱码不卡18| 国产综合懂色| 免费观看a级毛片全部| 午夜福利视频1000在线观看| 国产精品99久久久久久久久| 九九久久精品国产亚洲av麻豆| 熟女人妻精品中文字幕| 不卡视频在线观看欧美| 少妇人妻久久综合中文| 精品人妻熟女av久视频| 人人妻人人爽人人添夜夜欢视频 | 美女视频免费永久观看网站| 日韩伦理黄色片| 国产 一区 欧美 日韩| 国产精品三级大全| 久久久久久久久久成人| 91久久精品国产一区二区成人| 美女脱内裤让男人舔精品视频| 国产老妇伦熟女老妇高清| 日产精品乱码卡一卡2卡三| 性色av一级| 搡女人真爽免费视频火全软件| 国产一区有黄有色的免费视频| 男女边吃奶边做爰视频| 80岁老熟妇乱子伦牲交| 国产综合精华液| 国产高清不卡午夜福利| 好男人在线观看高清免费视频| 免费av毛片视频| 国产v大片淫在线免费观看| 99久久精品热视频| 男人添女人高潮全过程视频| av在线天堂中文字幕| 网址你懂的国产日韩在线| 99久国产av精品国产电影| 国内少妇人妻偷人精品xxx网站| 美女高潮的动态| 精品久久久久久电影网| 久久这里有精品视频免费| 午夜免费鲁丝| 波野结衣二区三区在线|