CN109194763A - Caching method based on small base station self-organizing cooperative in a kind of super-intensive network - Google Patents

Abstract

The invention provides a caching method based on self-organization and cooperation of small base stations in an ultra-dense network, and belongs to the technical field of wireless communication. Including: first, according to the load capacity and location of the small base stations in the ultra-dense network, the similarity matrix of the small base stations is obtained; then, according to the similarity matrix and the number k of the small base stations in each cluster, clustering is performed, and the most load capacity is selected. A good small base station acts as the cluster head; files are cached in the small base station and macro base station according to the file caching strategy; after the user accesses the base station, the user requests to obtain the file; finally, it is judged whether the value of k exceeds the maximum value K set in the cluster , if so, output the k value that minimizes the average download delay of the user; otherwise, continue to update the k value and continue to re-cluster. The invention allocates the resources in the cluster through the self-organization idea, utilizes the mutual cooperation and self-organization among the small base stations, effectively improves the cache hit rate and reduces the user download time delay, and meets various business requirements.

Description

A caching method based on self-organized cooperation of small base stations in ultra-dense networks

technical field

The invention belongs to the technical field of wireless communication, and in particular relates to a caching method based on self-organized cooperation of small base stations in an ultra-dense network.

Background technique

Ultra-dense networking, that is, ultra-dense network UDN technology, can achieve a huge improvement in frequency reuse efficiency by densely deploying small base station SBS, and is one of the key technologies to meet the thousand-fold increase in 5G capacity. Ultra-dense network UDN is considered to be one of the key technologies used to meet user needs and improve system capacity. UDN usually contains a large number of low-power small base station SBSs and the deployment density is much higher than the current mobile network scenario, which can provide users with extremely high data transmission rates. At the same time, caching technology has been introduced into ultra-dense networks, and network edge caching has attracted a lot of attention in recent years. Its main idea is to cache frequently requested content at the edge of the network, so that files in the network are closer to users, which can reduce redundancy. It can shorten the user's download delay, improve the user experience and improve the spectrum utilization. The combination of UDN scenarios and network edge caching technology can not only improve network throughput, reduce user download delay, but also reduce backhaul pressure, which will greatly improve network performance. However, considering the cost of ultra-dense deployment of SBSs in UDN, the cache capacity of a single SBS is very limited, so each SBS cannot cache all the files that users may need. Under this condition, various factors need to be considered. The caching method of SBSs, for example, should comprehensively consider the file popularity, hit rate and user experience, as well as mutual cooperation and self-organization between small base stations.

In the cache research for ultra-dense network scenarios, Reference 1 wants to maximize cache utilization considering the limitation of cache space. First, the optimization objective is transformed into a backhaul load minimization problem, and then machine learning related algorithms are used to predict and cache the corresponding content in the small base station, with low algorithm complexity and high accuracy. Reference 2 proposes a small base station social-aware caching strategy to improve network throughput and energy efficiency. First, the social network theory is used to study the social behavior of small base stations, and a few very important base stations are selected from base stations with high social bond factors, and resources are scheduled and allocated with other small cells based on these base stations.

In an ultra-dense scenario, the existing technology mainly focuses on a single base station to optimize the cache content, and does not fully consider the mutual cooperation and self-organization between small base stations. Provide services and reduce user download delay. Considering these problems can further improve the cache hit rate and user service experience.

references:

[1]G.Shen,L.Pei,P.Zhiwen,L.Nan and Y.Xiaohu,"Machine learning basedsmall cell cache strategy for ultra dense networks,"2017 9th International Conference on Wireless Communications and Signal Processing(WCSP),Nanjing , 2017, pp.1-6.

[2] Y.Li,X.Zhang,J.Zhang,S.Wang and D.Wang,"Base Station Social-AwareCaching Strategy for 5G Ultra Dense Networks,"2017IEEE Globecom Workshops(GCWkshps),Singapore,2017,pp. 1-6.

SUMMARY OF THE INVENTION

In view of the fact that the prior art does not comprehensively consider file popularity, hit rate and user experience, and does not fully consider the mutual cooperation and self-organization between small base stations, the present invention proposes a self-organized cooperation based on small base stations in an ultra-dense network. The caching method to achieve the purpose of reducing the user's download delay.

The ultra-dense heterogeneous network scenario of the present invention includes: a core network, a macro base station, S small base stations densely distributed under the coverage of the macro base station, and U mobile users randomly distributed in the entire network scenario; The core network, the macro base station and the small base station are connected by wireless links. A caching method based on self-organized cooperation of small base stations in an ultra-dense network proposed by the present invention includes the following steps:

Step 1, according to the load capacity and location of the small base station in the ultra-dense network, obtain the similarity matrix of the small base station;

The similarity matrix records the similarity between any two different small base stations, the similarity between the small base stations S ₁ and S ₂ Expressed as: θ∈[0,1]; where θ is a set constant, is the location similarity _of base stations S1 and _S2 , is the load capacity difference _of base stations S1 and S2 _;

Step 2, according to the similarity matrix of small base stations and the number k of small and medium base stations in each cluster, all small base stations are clustered, and the small base station with the best load capacity in each cluster is selected as the cluster head; k is a positive integer;

Step 3, cache the file in the small base station and the macro base station according to the file caching strategy;

The file caching strategy is as follows: assuming that the cache capacities of the small base stations are all the same, and at most M complete files are cached; all the files that the user may request are sorted according to the file popularity; for a cluster, the kM files with the highest popularity are selected. file and divide each file into k segments and cache them in the corresponding small base stations; continue to cache the remaining files with the highest popularity in the macro base station according to the popularity of the files;

Step 4: After the user accesses the base station, the user starts to request to obtain files. At this time, three situations will occur:

First, judge whether the file required by the user is cached in the small base station, if so, the first situation occurs: the file requested by the user is cached in the small base station; if not, then judge whether the file required by the user is cached in the macro base station, if so, The second situation occurs: the file requested by the user is cached in the macro base station; if neither of the above two situations occurs, then the third situation occurs: the file requested by the user is not cached in the network, and the user obtains the requested file through the core network. document;

Calculate the user's transmission delay in the three cases, and then obtain the user's average download delay, and then perform step 5;

Step 5, set the number of small base stations in each cluster to be K at most, and K is a positive integer; determine whether the value of k exceeds K, if so, select the value of k that minimizes the average download delay of the user and output; otherwise, continue to update k value, make k increment by 1, and then go to step 2 to continue.

Compared with the prior art, the present invention has the following obvious advantages:

(1) This method proposes a cooperative caching strategy based on SBS clustering. According to the simulation results, it can be seen that this method can effectively improve the cache hit rate and reduce the user download delay. feasibility and applicability to the business needs.

(2) The method can utilize the clustering and mutual cooperation of the small base stations, and utilize the network self-optimization idea to improve the resource management efficiency.

Description of drawings

Fig. 1 is the flow chart of SBS self-organized cooperative cache method in the present invention;

Fig. 2 is the cooperative caching method system model diagram of SBS clustering in the present invention;

3 is a schematic diagram of the cooperative cache strategy of SBS clustering in the present invention;

Fig. 4 is the relation diagram of cache hit rate of the present invention and the number of small base stations in the cluster;

5 is a graph showing the relationship between the average download delay of users and the number of small base stations in the cluster in the present invention;

FIG. 6 is a graph showing the relationship between the average download delay of users and the total number of small and medium-sized base stations in the scene in the present invention;

FIG. 7 is a graph showing the relationship between the average download delay of the user and the Zipf decay constant α in the present invention.

Detailed ways

In order to facilitate the understanding and implementation of the present invention by those of ordinary skill in the art, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments.

The invention proposes a caching method based on SBS self-organization cooperation in an ultra-dense network, aiming at how to efficiently provide services to users and reduce user download delay under the condition that the cache capacity of a single SBS is very limited. The method of the invention divides all SBSs into clusters based on the load capacity and location information of the SBSs, and selects a cluster head in each cluster to allocate resources in the cluster by using the self-organization idea. Then the corresponding files are fragmented and different fragments are cached in different SBSs, each base station caches files of the same size and sorted by file popularity. Next, the optimization goal is set to minimize the average download delay of users, and finally change the number of small base stations in the cluster, and find the optimal solution through the traversal method.

As shown in FIG. 1 , a cooperative caching method based on SBS self-organization is proposed for the present invention, which specifically includes steps 1 to 5, and each step is described in sequence below.

Step 1: Initialize parameters, and obtain a similarity matrix of small base stations according to the location information and load capacity information of all small base stations in the ultra-dense network. The initialized parameters include the location information and load capacity of the base station.

As shown in FIG. 2, it is the system model of the cooperative caching method for SBS clustering according to the present invention. The present invention considers the caching strategy in the ultra-dense heterogeneous network scenario. The network includes: a macro base station MBS and a core network, and a macro base station. It is connected with the core network, macro base station and small base station through wireless links. There are s small base stations SBS densely distributed under the coverage of the macro base station, and u mobile users are randomly distributed in the entire network scenario. This method assumes that users are in a static state when receiving files. A set of s small base stations in the network means that it has the form, with collection represents u mobile users in the network in the form,

The key to clustering small base stations is to select similarity features between base stations. Although there are many similarity features to choose from, two of the more important features are the location information and load capacity of small base stations [Reference 3: S . Samarakoon, M. Bennis, W. Saad, and M. Latva-aho, “Dynamic clustering and on/offstrategies for wireless small cell networks,” IEEE Transactions on Wireless Communications, vol. 15, no. 3, pp. 2164–2178 , March 2016.]. The location information reflects the ability of the base stations to cooperate with each other, and determines the operability of the cooperation between the base stations. The load capacity causes the problem of mutual interference between base stations and also shows the willingness of base stations to participate in cooperation, which determines whether it is necessary to cooperate. Next, the similarity of location information and load capacity between base stations will be calculated respectively.

On the one hand, the location similarity of any two different small base stations in the ultra-dense network is calculated from the location information of the small base stations, and for any two different small base stations, the Gaussian similarity corresponding to the two small base stations S ₁ and S ₂ is calculated, location similarity The formula is as follows:

in, respectively represent the coordinates of the two small base stations in the Euclidean space; σ _X is a set constant; r represents the maximum distance between the small base stations as neighbors. It can be seen that the smaller the distance between small base stations, the higher the location similarity, and the easier the base stations are to cooperate with each other. Furthermore, the location similarity between all s small base stations can be represented by an S×S matrix.

On the other hand, the similarity is calculated from the load capacity of the small base station. Since the larger the load capacity difference of the small base stations is, the more a cluster should be formed to perform inter-cluster task offloading, so the method of the present invention will calculate the load difference degree between the small base stations. For the convenience of calculation, the method of the present invention assumes that each user requests files at the same frequency. Therefore, the load capacity of the small base station can be simplified to the maximum number of connectable users. Finally, the load capacity difference between the small base stations S ₁ and S ₂ is obtained Calculated as follows:

Among them, σ _N is a set constant, used to adjust , in order to facilitate numerical calculation; σ _N determines the size of the numerical range of the degree of difference. respectively represent the maximum number of _connectable users _of the small base stations S1 and S2. with the above Same, It will also form an S×S matrix. The larger the value is, the easier it is for the two corresponding small base stations to form clusters.

Then, according to the location similarity and load capacity difference between the small base stations S ₁ and S ₂ , the similarity used to decide the clustering finally, the similarity between the small base stations S ₁ and S ₂ is obtained. Calculated as follows:

Among them, θ is a set constant, which is used to adjust the proportion of factors that determine clustering; the size of θ determines the degree of influence of the similarity of location information and the difference of load capacity on the overall similarity. Calculate the similarity of all any two different small base stations to form a similarity matrix of size S × S In this way, clustering is performed in the next step based on the similarity of the load capacity and the similarity of the location information of the SBS.

Step 2, according to the similarity matrix obtained by calculation, cluster the small base stations according to the small base station clustering strategy, and determine that the number of small base stations in each cluster is k. All small base stations in the network scenario can be divided into several clusters with the same number of small base stations in the cluster. In each cluster, the small base station with the best load capacity in each cluster is selected as the cluster head.

In this step, after determining the number k of small base stations in each cluster, randomly select a small base station in the scene as the center point of the cluster, according to the similarity matrix Obtain k-1 small base stations with the highest similarity to form a cluster. Select the unclustered small base station closest to the center point of the current cluster as the center point of the next group of clusters, and then select k-1 small base stations with the highest similarity to the cluster center point according to the similarity matrix of the small base station to form a cluster. cluster. The clustering operation is thus repeated. With the progress of clustering, if there are less than k-1 small base stations that meet the conditions, at this time, several small base stations with high similarity in the cluster are selected to fill up the base stations in the new cluster to k. And so on, until all the small base stations complete the clustering, the cluster head in each cluster is served by the base station with the best load capacity, and the cluster head uses the self-optimization idea to allocate the resources in the cluster and coordinate the assignment of tasks in the cluster.

Step 3: Cache the file to the corresponding small base station according to the file cache policy. Since the cache capacity of the small base station is very limited, in order to improve the cache hit rate, this method enables different small base stations in the same cluster to cache different fragments of files with high popularity, and the macro base station caches the remaining uncached complete files in the network.

The file caching strategy is that the small base station counts the number of file types I that users may request, and sorts all files according to the file popularity. The smaller the serial number, the higher the popularity of the file. Then, according to the file caching strategy, the files with the highest file popularity will be fragmented and the corresponding fragments will be cached in the small base station; some complete files with the highest popularity will be cached in the macro base station, and the files will still be cached according to the file popularity. , cache the most popular files first, until the cache space is exhausted.

The method of the present invention assumes that all small base stations have the same cache capacity, at most M complete files are cached, and the macro base station can cache N complete files. Sort all the files in the network according to the popularity of the files, and obtain the popularity file library The smaller the file subscript value, the more popular the corresponding file is. Since there are k small base stations in a cluster, each cluster can cache up to kM complete files. As shown in FIG. 3 , the cooperative caching strategy of SBS clustering in the present invention selects k×M files with the highest popularity, and divides each file into k segments evenly, and then caches these file segments in different in small base stations. The files whose popularity is between k×M and k×M+N will be completely cached in the macro base station.

According to the above caching strategy, the files cached by k small base stations in each cluster can be mathematically modeled. Let f _1,k denote the k th file fragment of the file with the first popularity, then for the k th small base station, all files C _k cached by it are expressed as:

C _k ={f _1,k ,f _2,k ,...,f _k(M-1)+1,k ...,f _kM,k }

Among them, f _kM,k represents the kth file fragment of the file with the kMth popularity.

Step 4: After the user accesses the base station, it starts to request to obtain files. The transmission strategy and the caching strategy rely on each other, and are divided into three cases. The transmission delay in each case is calculated, and the average download delay of the user is obtained, and then step 5 is executed. .

When a user wants to obtain a certain file from the network, when the file transfer problem is involved, three situations will occur at this time, and the user will obtain the desired file through three forms.

The first situation is that the file requested by the user is cached in the small base station, and the file requested by the user is transmitted through the small base station.

The user first sends a request for the file to the small base station to which it is connected, and the base station feeds back the request information to the cluster head. The cluster head judges whether there is a file required by the user in the cluster according to the popularity of the requested file. If so, it checks the current load of each small base station in the current cluster to determine which small base stations jointly serve the user, and allocates corresponding resources at the same time. If some small base stations that cache file fragments required by users cannot communicate directly with users, the cluster head will also confirm which base station to relay transmission through. This transmission process realizes the file transmission between the small base station and the user through the self-organization idea.

In the second case, the file requested by the user is cached in the macro base station, and the download delay is calculated according to the MBS transmission. If the cluster head finds that there is no file required by the user in the cluster, the cluster head will request the file from the macro base station. If the corresponding file is cached in the macro base station, the file is first transmitted to the cluster head via the macro base station, and then transmitted to the user by the cluster head.

The third situation is that the file requested by the user is not cached in the network, the file needs to be transmitted through the core network, and the download delay is calculated according to the core network transmission. If there is still no such file in the macro base station, the macro base station will continue to send requests to the core network. The core network transmits the file required by the user to the macro base station through the wireless link, and the macro base station transmits the file to the cluster head through the wireless link. It is then transmitted to the user by the cluster head. In this case, the file download delay will increase due to the longer transmission path.

The popularity of each file is different, and the probability of the user requesting the corresponding file from the small base station is different. This method assumes that the probability of users requesting different files obeys Zipf's law [Reference 4: L. Breslau, P. Cao, L. Fan, G. Phillips, and S. Shenker, "Web caching and zipf-like distributions: evidence and Implications,” in INFOCOM’99. Eighteenth Annual Joint Conference of the IEEE Computer and Communications Societies. Proceedings. IEEE, vol.1, Mar 1999, pp.126–134vol.1.]. P _i is used to represent the probability that a user requests a file whose popularity is ranked i from the small base station. According to Zipf's law, P _i is obtained as:

Among them, α represents the decay constant; j represents the file type; i represents the popularity ranking; after calculating the probability of each file being requested, the theoretical cache hit rate of the small base station for the user can be calculated

In order to facilitate the calculation, when the small base station sends the file to the user, the method of the present invention assumes that the small base station does not share spectrum resources with the macro base station. In addition, different small base stations in the same cluster will be allocated mutually orthogonal frequency bands for transmitting files, avoiding mutual interference between small base stations. Let the available bandwidths of the small base station and the macro base station be ω _S and ω _M respectively; h _s,u is the channel gain between the sth small base station and the uth user, and h _M,s is the macro base station and the sth small base station. Channel gain between base stations. use separately and Calculate the corresponding channel gain, where L ₀ represents the gain coefficient, which is a constant, β is the path loss indicator factor, d _{s, u} represents the distance between the s-th small base station and the u-th user, and d _M,s represents the The distance between the macro base station and the sth small base station.

The signal-to-noise ratio SNR _s,u between the s-th small base station and the u-th user and the signal-to-noise ratio SNR _M,s between the macro base station and the s-th small base station are calculated by the following formulas.

SNR _s,u =P _S ·h _s,u /σ ²

SNR _M,s =P _M ·h _M,s /σ ²

Among them, σ ² represents the noise power in the network scenario; _{PM is set as the transmit power when the macro base station communicates with a single small base station; P S is} set as the transmit power when the small base station communicates with a single user.

For the first case, the cluster head uses the self-optimization idea to allocate resources in the cluster, and controls all the small base stations in the cluster directly connected to the user to cooperate to transmit the file, thereby shortening the download delay. However, there may also be cases where the small base stations in the cluster need to transmit files for users through relays, so the download delay is also related to the number of base stations in the cluster that are directly connected to the user.

In the first case, based on Shannon's formula, the transmission delay of the u-th user when the file is obtained through the small base station is obtained for:

Among them, Num _s represents the number of file fragments that the sth small base station needs to transmit for the user; F is the size of the required file; t represents the number of small base stations that communicate directly with the user.

In the second case, the transmission delay when the uth user obtains the file through the macro base station It includes two parts: the macro base station transmits files to the cluster head and the cluster head transmits files to the user:

In the third case, the uth user obtains the file through the core network. Transmission delay when the uth user obtains files through the core network In addition to containing In addition, the communication time between the core network and the macro base station is also included. In order to simplify the calculation, the method of the present invention uses the constant Const to represent this part of the time delay. therefore is calculated as follows:

Let the probability of the u-th user obtaining files from the small base station, the macro base station and the core network be expressed as as follows:

Therefore, the calculated average download delay Delay ^u of the u-th user is as follows:

Step 5, given that the maximum number of small base stations that can exist in each cluster is K, the optimization goal of the method of the present invention is to minimize the average download delay of users under the condition of satisfying a certain cache hit rate. The calculation formula of the optimization goal is as follows:

subject to: k∈{1,2,…,K}

Changing the size of k will adjust the number of small and medium-sized base stations in each cluster after clustering. After each change of k, you need to perform step 2 according to the value of k, re-cluster, and then continue to perform steps 3 and 4 to obtain the corresponding Small base station cache hit rate and user average download delay. By traversing the value of k, the value of k that minimizes the average download delay of the user is obtained.

In step 5, first determine whether the value of k exceeds K, if so, output the value of k that minimizes the average download delay of the user, and end the method of the present invention; if not, continue to change the value of k, and execute step 2. The present invention traverses the value of k from small to large until it reaches K.

The method of the present invention is simulated and verified as follows. The comparison algorithm of this simulation is the "most popular" file caching scheme [Reference 5: T.Wu, X.Li, H.Ji, and H.Zhang, "An energy-efficient sleep management algorithm for udn with edge caching,” in 2017 IEEE Globecom Workshops (GCWkshps), Dec 2017, pp.1–5.]. In the "most popular" file caching scheme, all small cells cache the same M complete files with the highest popularity.

The simulation verification is as follows:

If no special instructions are given, the total number of file types the user may request is I=200, the size of the requested file is F=10Mbits, the attenuation constant α in Zipf’s law is set to α=0.8, the storage capacity of the small base station is For M=10 files, the storage capacity of the macro base station is N=100 files. The size of the simulation scene is 200m*200m and the macro base station is placed in the center. The number of small base stations in the network is 50 and the coverage radius is r=30m, the sub-channel available bandwidth of macro base station and small base station is ω _M =ω _S =1MHz, the transmit power is P _S =100mW, P _M =20W, noise power σ ² =-100dBm, gain coefficient L ₀ =-30dB. In order to simplify the calculation, in this method, the transmission delay from the core network to the macro base station is set as a fixed value C=1s.

As shown in Fig. 4, it is a graph showing the relationship between the cache hit rate and the number of small base stations in the cluster according to the method of the present invention, which shows the relationship between the average cache hit rate and the number k of small base stations in the cluster. The horizontal axis represents the number of base stations, and the vertical axis represents the cache. The average hit rate, different curves represent different file types, and any curve is selected from Figure 4, it can be seen that the small base station cache hit rate increases with the increase of the number of small base stations in the cluster, and the increasing trend is getting slower and slower. Because the more small base stations in the cluster, the more different types of files are cached in the cluster, and the higher the probability that the files requested by the user can be obtained directly from the small base stations, that is to say, the higher the cache hit rate is. According to Zipf's law, the probability of a file with a lower popularity being requested by a user is smaller, and the caching strategy of this method is to cache files in sequence according to their popularity. The contribution is not large, so there will be a situation where the increasing trend is getting slower and slower. At the same time, changing the number of file types I requested by users in the network will also affect the cache hit rate of the small base station. The larger the number of files in the network, the smaller the probability of a user requesting a single file, which leads to a smaller probability that the file cached in the cluster is requested by the user. Therefore, it can be seen from FIG. 4 that when the number of file types I=200, I=240, I=280, and I=320, as the number of file types I is smaller, the cache hit rate of the small base station increases instead.

From the cache hit rate alone, it seems that the more small base stations in the cluster, the better. However, with the increase of the number of small base stations in the cluster, the problem of resource allocation in the cluster and the computing power of the cluster head will be tested, and thus the average download delay of users may increase. Therefore, the next step will be a trade-off between the cache hit rate and the average download delay of users.

As shown in FIG. 5 , it is a graph of the relationship between the average download delay of users and the number of small base stations in the cluster in the method of the present invention, which shows the relationship between the average download delay of users and the number k of small base stations in the cluster. It should be pointed out here that when the number of small base stations in the cluster is 1, all files do not need to be fragmented, and all files are cached in the base station, which is consistent with the "most popular" caching scheme. It can be seen from Figure 5 that in the process of increasing the number of small base stations in the cluster from 1 to 6, the average download delay of users first decreases and then increases. The reason for the reduction in the average download delay of users is that users can obtain files faster from small base stations than from macro base stations or the core network. After that, the delay began to increase again because with the increase of the number of base stations in the cluster, the cooperation between the small base stations in the cluster became more complicated, more relays may be required, and the spectrum resources in the cluster were also fixed, so This results in slower user access to complete files from small cells. In addition, when the total number of small base stations is 50, the "most popular" caching scheme can be compared from Figure 5, the two points when k=1 represent the "most popular" caching scheme and the proposed method based on The cooperative caching mechanism of SBS clustering (k=4), the average download delay of the two users is basically the same, but the cache hit rate has increased from 0.35 to 0.6, as can be seen from the curve corresponding to I=200 in Figure 4 , the parameter settings are consistent with Figure 5. This proves that the cache mechanism proposed by the method of the present invention can indeed improve the cache hit rate in the UDN scenario, and at the same time will greatly reduce the pressure on the backhaul network.

As shown in FIG. 6 , it is the relationship between the average download delay of users and the total number of small base stations in the scene in the present invention, which shows the relationship between the average download delay of users and the total number of small base stations in the scene. In the simulation process, the number of small base stations in the cluster is 4. The total number of small base stations in the network starts from 50, increases by 10 each time, and finally increases to 100. It can be seen from Fig. 6 that under the cooperative caching mechanism based on SBS clustering proposed in the present invention, the average download delay of users decreases with the increase of the total number of small base stations in the scene, while the "most popular" file caches There is basically no change in the average download delay of users under the scheme. This is because in order to simplify the calculation during the simulation, the method of the present invention assumes that the distance between the user and the cluster head base station is fixed, so the change of the total number of small base stations does not affect the distance between the user and the base station under the "most popular" cache scheme , so the average download delay of users does not change much. However, under the cooperative caching mechanism based on SBS clustering, the distance between users and other base stations except the cluster head decreases with the increase of the total number of small base stations in the scene, so the average download delay of users also decreases. .

As shown in FIG. 7 , it is the relationship between the user average download delay and the Zipf decay constant α in the present invention, which shows the relationship between the user average download delay and the Zipf decay constant α. In the simulation process of this figure, the number of small base stations in the cluster is 4, the number of small base stations in the network is 50, and the Zipf attenuation constant α is increased from 0.6 to 2.0. In Figure 7, with the increase of the Zipf decay constant α, the average download delay of users decreases continuously, and the decreasing trend is also getting smaller and smaller. This is because a high α means that the most popular files are more likely to be requested, and these files are often cached in small cells, and the latency of obtaining files from small cells is generally lower than The delay of obtaining files from the macro base station or from the core network, so the average download delay of users will be reduced accordingly. In addition, considering that all the small base stations in the cluster jointly allocate the spectrum resources in the cluster, and each small base station in the "most popular" caching scheme has its own corresponding spectrum resources, when the "most popular" caching scheme is used, the small base station The delay required for transmission is smaller. However, the average user download delay is also related to the cache hit rate. When α < 0.9, using the cooperative caching strategy based on SBS clustering makes the user more likely to obtain files from the small base station, and the average user download delay is better at this time. For the "most popular" caching scheme. With the further increase of α, the difference between the two schemes in the cache hit rate becomes smaller and smaller, resulting in a relatively large variation of the delay of the "most popular" cache scheme, which is better than the strategy proposed by the method of the present invention. In general, the value of α takes 0.6 to 0.8.

Claims (5)

1. A caching method based on small base station self-organizing cooperation in an ultra-dense network is characterized by comprising the following steps:

step 1, obtaining a similarity matrix of small base stations according to the load capacity and the positions of the small base stations in the ultra-dense network;

the similarity matrix records the similarity of any two different small base stations, namely a small base station S₁And S₂Degree of similarity ofExpressed as:θ∈[0,1](ii) a Where, theta is a constant that is set,is a base station S₁And S₂The degree of similarity of the positions of (a),is a base station S₁And S₂The degree of load capacity variation of (a);

step 2, clustering all small base stations according to the similarity matrix of the small base stations and the number k of the small base stations in each cluster, and selecting the small base station with the best load capacity in each cluster as a cluster head; k is a positive integer;

step 3, caching files into the small base station and the macro base station according to a file caching strategy;

the file caching strategy is as follows: setting the same cache capacity of the small base station, and caching M complete files at most; sorting all files possibly requested by a user according to the popularity of the files; for one cluster, selecting kM files with the highest popularity and dividing each file into k segments to be respectively cached in the corresponding small base stations; continuously caching the remaining files with former popularity according to the file popularity in the macro base station;

step 4, after the user accesses the base station, the user starts to request to acquire the file, and at this time, three situations occur:

firstly, judging whether a file required by a user is cached in a small-sized base station, if so, a first condition occurs: the small base station caches files requested by a user; if not, judging whether the file required by the user is cached in the macro base station, if so, the second condition occurs: caching a file requested by a user in a macro base station; if neither of the above two cases is true, then a third case occurs: the file requested by the user is not cached in the network, and the user acquires the requested file through the core network;

calculating the transmission time delay of the user under the three conditions so as to obtain the average downloading time delay of the user, and then executing the step 5;

step 5, setting the number of the small base stations in each cluster to be at most K, wherein K is a positive integer; judging whether the value of K exceeds K, if so, selecting the K value which enables the average download time delay of the user to be minimum, and outputting the K value; otherwise, continuing to update the value of k, increasing k by 1, and then continuing to execute the step 2.

2. The caching method based on small cell self-organizing coordination in the ultra-dense network as claimed in claim 1, wherein in the step 1, the small cell S₁And S₂Position similarity ofThe calculation formula of (a) is as follows:

wherein,respectively representing the coordinates of the two small base stations in the Euclidean space; sigma_XIs a set constant; r represents the maximum distance between the small base stations which are neighbors;

small base station S₁And S₂Degree of load capacity differenceThe calculation formula of (a) is as follows:

wherein σ_NIs a constant that is set, and the setting value,respectively represent smallType base station S₁And S₂The maximum number of connectable users.

3. The caching method based on small base station self-organizing cooperation in the ultra-dense network according to claim 1, wherein in the step 2, after the number k of small base stations in each cluster is determined, one small base station in a scene is randomly selected as a cluster center point, and then according to a similarity matrix of the small base stations, k-1 small base stations with the highest similarity with the cluster center point are selected to form a cluster; then selecting a small base station closest to the current cluster center point from the small base stations which are not clustered as the center point of the next group of clusters, and continuously selecting the small base stations from the new cluster center point according to the similarity matrix of the small base stations to form a cluster; repeating clustering operation until all the small base stations finish clustering, and enabling the cluster head in each cluster to be served by the small base station with the best load capacity; when small base stations in a cluster are selected according to the similarity matrix of the small base stations, if less than k-1 base stations meeting the conditions are found from the small base stations which are not clustered, the base stations with higher similarity are selected from the small base stations which are clustered, and the base stations in the new cluster are supplemented to k.

4. The caching method based on small-sized base station self-organizing cooperation in the ultra-dense network according to claim 1, wherein in the step 4, the transmission delays of the users under three conditions are respectively as follows:

in the first case, the transmission delay of the u-th user when obtaining the file through the small-sized base stationComprises the following steps:

wherein, ω is_SRepresenting the available bandwidth of the small base station; num_sThe number of file fragments which need to be transmitted for the user by the s small base station is represented;f is the size of the file required by the user; t represents the number of small base stations communicating directly with the user; SNR_s,uRepresenting the signal-to-noise ratio between the s small cell base station and the u user; u is a positive integer;

in the second case, the transmission delay of the u-th user when obtaining the file through the macro base stationComprises the following steps:

wherein, ω is_MRepresents the available bandwidth of the macro base station; SNR_M,sRepresenting the signal-to-noise ratio between the macro base station and the s-th small base station;

in the third case, the transmission delay of the u-th user when obtaining the file through the core network

Where Const denotes a set constant.

5. The caching method based on small-sized base station self-organizing coordination in the ultra-dense network according to claim 1 or 4, wherein in the step 4, the average download delay of the user is obtained as follows:

for the u-th user, the transmission delay of the u-th user under the first condition, the second condition and the third condition is obtained through calculationAndthen, the u-th usage is calculatedAverage download Delay of user^uThe following were used:

wherein, P₁ ^u,And respectively representing the probability of acquiring files from the small base station, the macro base station and the core network by u users.