Caching method for jointly storing push file and request file

Abstract

The invention discloses a caching method for jointly storing a push file and a request file, which relates to the technical field of D2D caching, and is characterized in that when a network load is in a non-peak state, hot contents are pushed to a user terminal with caching capability in a broadcast mode so as to meet subsequent frequent content access requirements; for a terminal direct transmission cache network in which all mobile users have cache capabilities, position distribution of the mobile users is modeled as a uniform Poisson point process, on the basis, randomness of content cache and content requests is combined, accurate analysis is performed on network interference, and approximate analysis in a specific scene is completed at the same time. And deducing a closed expression of the successful unloading probability of the D2D cache network by using a random geometry theory. Then, a Geo / G / 1 queuing model is established based on the access protocols, the model considers file transmission processes involved under different access protocols, and the queue state of a to-be-requested file in a user cache is analyzed;

Description

TECHNICAL FIELD

[0001] The present application relates to D2D caching technology, and in particular to a caching method for jointly storing and pushing files. BACKGROUND

[0002] To cope with the increasing demand for data services in cellular systems, Device-to-Device (D2D) caching technology has become a research hotspot in the industry. The basic idea is to use the wide range of idle cache space in the cellular system to place popular content in the user cache space during the idle period, so that in the busy period, the requesting user can obtain the required content directly from the adjacent cache user through D2D communication without going through the cellular backhaul link. This not only relieves the pressure on the cellular network infrastructure, but also provides faster and more efficient data service for users.

[0003] In recent years, a large number of documents have studied D2D caching network communication and designed different caching strategies to achieve different optimization goals. The D2D caching mechanism in the prior art is a request-driven passive cache, that is, the user caches the files that he has requested and stored by the base station to his cache.

[0004] Since most users only consume a small part (5%-10%) of "popular" content, and fewer users request less popular content. This principle predicts the files that may be requested in the future according to popularity, and stores them in the local cache before the actual request occurs. Specifically, popular files can be cached in user memory during off-peak hours (such as midnight), so that users can obtain request files from the device during peak hours (such as daytime), rather than from the base station. It is worth noting that a popular file can be requested by multiple users. Therefore, during the peak load period, wireless caching can meet the requests of more users. This way of storing popular files on mobile devices can significantly reduce base station traffic and improve system throughput.

[0005] Random geometry theory can be used to describe the spatial randomness of the network, derive network coverage range and other performance parameters, and then derive the influence of interference and channel fading on users in a complex network environment. Queuing theory focuses on the time dimension of the communication network, models the arrival and transmission process of user traffic groups, and through mathematical calculations, derives the system throughput, delay and other performance indicators that users are interested in after the system has been running for a long time to reach a stable state.

[0006] There are documents (Yang C, Chen Z, Yao Y, et al. Performance analysis of wireless heterogeneous networks with pushing and caching[C] / / ICC 2015 - 2015 IEEE International Conference on Communications. IEEE, 2015) assuming that all users with cache function store the same push file, that is, if there is no target file in the user's own cache, there is no user with the file in the surrounding, only users without cache function will establish D2D link to transmit popular files; this mechanism considers the importance of popular files, but only stores the highest degree of popular files, resulting in no differentiation between users with cache function. SUMMARY

[0007] In view of the above technical problems, the purpose of the present application is to combine the advantages of the above two theories, model the file transmission process with pre-storage mechanism in the cache cellular network, and propose a cache method for jointly storing push files and request files, which can maximize the self-offloading probability without increasing the network burden.

[0008] In order to achieve the above purpose, the technical scheme adopted by the present application is: in order to consider the difference of user demand, the user cache is divided into two parts: push cache queue and request cache queue. Among them, the operation mechanism of the push cache queue is: when the network load is in a non-peak running state, the most popular content can be pushed to the user with cache function through broadcast, and the user receives and caches it to meet the subsequent access demand; the request cache queue is for non-popular file content, if the user's own cache cannot meet the demand, the file can be transmitted by the surrounding user through device to device (D2D) communication. The user can obtain the required file through D2D transmission or base station transmission and save it, which improves the self-offloading probability of the file and reduces the network overhead.

[0009] A cache method for jointly storing push files and request files, comprising the following steps:

[0010] Step 1: for the D2D cache network in which all mobile users have cache capability, model the base station and user location as independent Poisson point processes;

[0011] Step 2: combine the randomness of content cache and content request to analyze the queue state of the files requested in the user cache;

[0012] Step 3: Using random geometry theory, derive a closed-form expression for the probability of successful offloading in the D2D cache network.

[0013] Step 4: Build a Geo / G / 1 queue model for the user request cache queue, where Geo represents geometric distribution, G stands for general distribution, and 1 represents the number of service desks. Analyze the arrival process, service process, and queue state transition patterns. Furthermore, construct a two-dimensional discrete-time Markov chain that includes queue length and transmission states.

[0014] Step 5: Use the quasi-birth-death process method to solve the steady-state probability distribution of the Markov chain and obtain the expression of the key performance parameters.

[0015] In step 1: the location distribution of the base station follows the density PPP process , the user's location distribution follows the density PPP process , are independent and identically distributed; assuming that some users have caching capabilities, Indicates the proportion of users with cache function, , these users have the same cache capacity ; If the distance between the requesting user and the user storing the target file is less than When a D2D communication link is established between users, the file can be transferred to the requesting user.

[0016] Multimedia file collection is ,in is the total number of files, For file collection dimensions, all multimedia files have the same size of , user request The probability of any file in follows the popularity distribution function of the file , assuming the function Obey the Zipf distribution, expressed as:

[0017]

[0018] in Reflects the skewness of the popularity function, high Indicates large file reusability; The physical meaning is The probability of The first few files with small values ​​have large popularity and correspond to most requests. Is a normalization factor used to ensure that the sum of the probabilities of all elements is 1, satisfying the basic properties of probability distribution. denotes a normalized parameter;

[0019] Definition There is a set of probabilities of each file for the user, where denotes the probability that the user has the file , there is a limited capacity of the D2D user buffer, so must satisfy:

[0020]

[0021] The user density that has the file is expressed as , and the set of users is expressed as .

[0022] In step 2: Push-based cache cellular network, the files cached by wireless communication are divided into pre-storage stage and delivery stage; the pre-storage stage is that the base station sends the files with high popularity to the user by using the broadcast channel before the actual request, and saves in the local cache in advance; the delivery stage is the unicast operation for the request when the user generates the request, the user with cache function generates the request, first checks whether the required file is in the local cache, if not, the user sends the request to the user with cache function around, and the user or the base station that receives the request sends the requested file to the user, that is, the transmission of the file only occurs when the request is responded.

[0023] In step 3, the cache hit probability includes the self-push-cache file hit probability and the D2D hit probability;

[0024] The self-push-cache file hit probability: the probability that the required file is satisfied by the popular file pushed by the base station without the need to request other devices; the dynamic process of the generation and disappearance of the file pushed by the base station is described by the birth-death process, for the service process, it refers to the existence time of the file in the user cache, which is subject to the exponential distribution with the parameter , and the limited space capacity of the push-cache queue of the user is assumed to be , so the number of pre-stored files in the queue after a period of time is , and the probability is expressed as:

[0025]

[0026] Where is the birth rate of the file, is the death rate, is the number of files in the user push-cache queue; the uninstallation probability is defined as the ratio of the uninstallation request to the total request, and is expressed as Indicates that for a system with a pre-store mechanism, when a user generates a request, the required file has been pushed to the user and saved by the broadcast system, and the cache queue is pushed. The file hit probability when there are 1 file can be expressed as:

[0027]

[0028] The average length of pre-stored files in the cache of a user is expressed as:

[0029]

[0030] The D2D hit probability: there is a cache and the file exists The user location distribution follows the density PPP process, then the file The probability of being hit by a D2D user is expressed as:

[0031] .

[0032] In step 3, for requests that cannot be satisfied by its own cache, the required files are transmitted by surrounding users or base stations, and the user requests the file. , when it is confirmed that there is no file in its own storage, the file will be The users in the network establish a D2D link, with the nearest user acting as the transmitter. As the receiving end, the file Send to user , assuming that multiple D2D links communicate on the same spectrum resources in a cell, the communication links from the base station to the user and between users experience large-scale fading models with different path loss factors, and the small-scale fading channels are modeled as mutually independent exponential distributions with a mean of 1, then the user The received signal-to-interference-and-noise ratio is expressed as:

[0033]

[0034] in is the user's transmit power, and are the small-scale fading and transmission distance of the D2D link, is the path loss factor of the link, is the noise power, is the set of interferences received, given by:

[0035]

[0036] The first term in the definition is , is from Other D2D transmitters The interference caused, and Requesting users To other users , Small-scale fading power gain, and user To other users The second term is , is from Base station that sends files to users The interference caused by, represents the transmission power of the base station, represents the path loss factor from the base station to the user;

[0037] If there is no file stored around the requesting user The user is the user Send files to the service base station To the user, the received signal-to-interference-and-noise ratio of the signal received from the base station is:

[0038]

[0039] in Indicates the transmit power of the base station, and Users To its service base station Small-scale fading and transmission distance of the transmission link, is the sum of the interfering signal powers and is given by:

[0040]

[0041] The first term in the definition is ,yes Medium The second item is the interference caused by base stations other than the one sending files to users. , from Other D2D transmitters The interference caused by this is described by the random geometry principle and the complementary cumulative distribution function (CCDF) of the cellular link SINR to describe the probability of successful transmission.

[0042] In step 4: Model the user with cache function, assuming that the user request arrives subject to the parameter Bernoulli distribution, the total cache capacity of the user is , is divided into two parts. One part is used to store popular files pushed by the base station, called push-cache queue, and the cache space is used The length of the queue is determined by the birth rate and death rate of the file itself, and the period is greater than the cache period; the other part of the cache stores the files requested by the user from the surrounding or base station and saved, which is called the request-cache queue, and the cache space is If or contains the file required by the user, the user does not need to request externally.

[0043] In step 5: the behavior of queuing is evaluated by using a quasi-birth-death process, and the state of the queue length is rearranged, which is regarded as a one-dimensional Markov process. . .

[0044] The state set is called the level , , and the phase is

[0045] A plurality of performance parameters can be obtained from the steady-state distribution, first of all, the self-unloading probability, which refers to the probability that the request of the user is satisfied by the file in the push-cache queue or the request-cache queue, the average queue length is considered by queuing theory, and the self-unloading probability expression is:

[0046]

[0047] The average queue length of the request-cache queue is expressed as:

[0048]

[0049] The packet loss rate is that when the request-cache queue is full, the request cannot be satisfied by the push-cache queue and the request-cache queue, and the user obtains the file from the D2D link or the cellular link, and the overflow packet loss occurs:

[0050] .

[0051] Compared with the prior art, the beneficial effects of the present application are as follows:

[0052] (1) The present application first proposes a cache strategy for jointly storing push files and request files in a wireless cache network, and the user can select a D2D communication mode or a cellular communication mode under the strategy. Specifically, the user with cache function can store both popular push files and files requested from surrounding users or base stations; the self-unloading of the file occurs in the scene where the push-cache queue or the request-cache queue exists the target file, and the state of the request-cache queue depends on three modes: self-satisfaction, D2D transmission or base station transmission.

[0053] (2) The application combines random geometry and queuing theory to construct a space-time model, and analyzes the file transmission process with a pre-storage mechanism. First, the positions of base stations and users are modeled as independent Poisson point processes (PPP), and the successful transmission probability expressions of D2D communication and user-to-base station are derived; then, a Geo / G / 1 queuing model is established according to the access protocol, and the request buffer queue state in the user cache is analyzed.

[0054] (3) The application establishes a Geo / G / 1 queuing model for the user request buffer queue, analyzes the arrival process, service process and queue state transition law, further constructs a two-dimensional Markov chain containing the queue length state and transmission state, and finally obtains the expressions of key performance parameters by solving the steady-state probability distribution of the Markov chain using the quasi-birth-and-death process method on the basis of in-depth research on various transmission schemes. BRIEF DESCRIPTION OF DRAWINGS

[0055] Figure 1 Fig. 1 is a schematic diagram of the file receiving model in the D2D heterogeneous cellular network in the embodiment.

[0056] Figure 2 Fig. 2 is a schematic diagram of the user file acquisition process in the embodiment.

[0057] Figure 3 Fig. 3 is a schematic diagram of the state transition of the user request-buffer queue in the cache cellular network in the embodiment.

[0058] Figure 4 Fig. 4 is a schematic diagram of the queue length of the request-buffer queue changing with the request arrival rate in the embodiment.

[0059] Figure 5 Fig. 5 is a schematic diagram of the relationship between the user request arrival rate and the request-buffer queue packet loss rate in the embodiment.

[0060] Figure 6 Fig. 6 is a schematic diagram of the relationship between the self-unloading probability and in the embodiment. DETAILED DESCRIPTION

[0061] In order to better understand the technical solutions of the application, the following further explains and describes the embodiments in combination with the drawings.

[0062] I. System model and protocol assumptions

[0063] In the embodiment, the spatial position distribution of users in the D2D cache cellular network is first modeled by a Poisson point process, and then the file unloading protocol of users with a pre-storage mechanism is discussed.

[0064] In the embodiment, the D2D assisted cache cellular network is considered, and the position distribution of base stations obeys a Poisson point process with a density of PPP process , the user's location distribution follows the density PPP process , are independent and identically distributed. Assuming that some users have caching capabilities, Indicates the proportion of users with cache function, , these users have the same cache capacity In addition, if the distance between the requesting user and the user storing the target file is less than When a D2D communication link is established between users, the file can be transferred to the requesting user.

[0065] Multimedia file collection is ,in is the total number of files, For file collection dimensions, all multimedia files have the same size of , user request The probability of any file in follows the popularity distribution of the file , assuming Obey the Zipf distribution, expressed as

[0066] (1)

[0067] in Reflects the skewness of the popularity function, the higher Indicates greater file reusability. Indicates the file number, The physical meaning is that the requested file is The probability of The first few files with smaller values ​​have greater popularity and correspond to most requests. Is a normalization factor used to ensure that the sum of the probabilities of all elements is 1, satisfying the basic properties of probability distribution. represents the normalization parameter.

[0068] definition There is a probability set of each file stored for the user, where Indicates that the user has files The probability that , due to the limited capacity of D2D user buffer area, Must meet:

[0069] (2)

[0070] There are files The user density is expressed as , the user set is represented as .

[0071] File offloading protocol

[0072] To make full use of the broadcast channel resources to reduce the occupation of the cellular network channel for file transmission, the embodiment considers a push-based cache cellular network, and files cached by wireless communication are divided into a pre-storage stage and a delivery stage.

[0073] The pre-storage stage is that before the actual request, the base station transmits the most popular part of the file to the user by using the broadcast channel, and saves it in the local cache in advance. Since the user cache space is limited, it is assumed that the data pushed by the base station through the broadcast can be received by all users in the cell, and the user stores part of the popular file according to its own interest.

[0074] On the other hand, the delivery stage is a unicast operation for the request when the user generates the request, and the user with the cache function generates the request, first checks whether the required file is in the local cache, if not, the user sends a request to the surrounding user with the cache function, and the user or the base station receiving the request sends the requested file to the user, that is, the transmission of the file only occurs in response to the request.

[0075] Figure 1 The schematic diagram of the file receiving model. The time of the application is discretized into time slots, it is assumed that the file is requested and received in the same time slot, if the file is not successfully transmitted in the time slot, the transmission will continue in the subsequent time slot.

[0076] For a typical user located at the center of the cell, the user communicates with the nearest base station, and it is assumed that the user randomly requests any file in N according to the popularity function , and represents a two-dimensional real space, and there are three ways to obtain the required file of the user:

[0077] (1) Self-offloading mode

[0078] When the request is generated, the user first checks whether the file in the user's cache queue can meet the request, if so, the user immediately satisfies the request without delay.

[0079] (2) D2D offloading mode

[0080] If the user who sends the request does not have the cache function or does not have the required file, the user will turn to the nearby user to search for the required file, then the base station checks whether there is a user who caches the file in the area around the user , establishes a D2D link to send the required file from the surrounding user to the requesting user, and offloads the service request from the cellular link to the D2D transmission. The embodiment refers to the area as a cooperation area, and assumes that it is a circle with the user as the center, a circle with radius. If at least one user stores the desired content within the radius SINR, which stands for signal to interference plus noise ratio, is satisfied, the D2D communication link is established to help the node as a transmitter to send the file to the user.

[0081] (3) Base station response

[0082] If the request cannot be satisfied by local caching or D2D offloading, forwarding to the cellular base station and cellular network is needed, the requested content is responded and sent, then the base station takes over to send the file to the user in the form of unicast.

[0083] II. Offloading strategy

[0084] 1. Cache hit probability

[0085] (1) Self-push-cache file hit probability

[0086] The probability that the required file can be satisfied by the base station pushing the popular file without requesting other devices. The probability that a file can be self-satisfied is related to the characteristics of the file itself and the user's request. The dynamic process of the generation and extinction of the base station pushing file can be described by the birth-death process. For the service process, it refers to the existence time of the file in the user's cache, which obeys the exponential distribution with parameter . Assuming that the limited space capacity of the user's push-cache queue is , then the number of pre-stored files in the queue after a period of time is The probability that the number of files in the queue is

[0087] (3)

[0088] where is the birth rate of the file, is the death rate, is the number of files in the user's push-cache queue, m represents the number of files, e represents the natural constant, which is approximately equal to 2.71828. The offloading probability is defined as the ratio of the offloading request to the total request, which is represented by . For the system with pre-stored mechanism, when the user generates a request, the required file has been pushed to the user by the broadcast system and saved, then the file hit probability when there are files in the push-cache queue can be represented as:

[0089] (4)

[0090] The average queue length of the pre-stored file in the cache is represented by

[0091] (5)

[0092] (2) D2D hit probability

[0093] Has cache and stored files The user location distribution follows the density PPP process, then the file The probability of being hit by a D2D user is expressed as:

[0094] (6)

[0095] 2Probability of successful file transfer

[0096] For requests that cannot be satisfied by its own cache, the required files are transmitted by surrounding users or base stations. , when it is confirmed that there is no file in its own storage, the file will be The users in the network establish a D2D link, with the nearest user acting as the transmitter. As the receiving end, the file Send to user , assuming that multiple D2D links communicate on the same spectrum resources in a cell, the communication links from the base station to the user and between users experience large-scale fading models with different path loss factors, and the small-scale fading channels are modeled as independent exponential distributions with a mean of 1, then the user The received signal-to-interference-and-noise ratio can be expressed as:

[0097] (7)

[0098] in is the user's transmit power, and are the small-scale fading and transmission distance of the D2D link, is the path loss factor of the link, is the noise power, is the set of interferences received, given by

[0099] (8)

[0100] The first term in the definition (8) is , is from Other D2D transmitters The interference caused, and Requesting users To other users , Small-scale fading power gain, and user to other users . The second term is , the interference from base stations transmitting files to the user. denotes the transmit power of the base station, denotes the path loss factor from the base station to the user.

[0101] Similarly, if there is no user around the requesting user who has the file , the file is transmitted to the user by the serving base station of the user , the received signal-to-interference-plus-noise ratio (SINR) from the base station is:

[0102] (9)

[0103] where denotes the transmit power of the base station, and are the small-scale fading and the transmission distance of the user to its serving base station , respectively, is the sum of the interference signal powers, given by:

[0104] (10)

[0105] The first term in equation (10) is , the interference from the base stations other than to the user transmitting files, and the second term is , the interference from other D2D transmitters in . The successful transmission probability is characterized by Lemma 1, Lemma 2 using the stochastic geometry principle. Lemma 1 The complementary cumulative distribution function (CCDF) of the SINR of a D2D link in the described D2D heterogeneous cellular network, defined as

[0106] is the successful transmission probability of a D2D link, expressed as:

[0107] (11) where

[0108] and are the Laplace transforms of and , given by:

[0109] (12) ​​

[0110] (13)

[0111] Lemma 1 Proof:

[0112] (14)

[0113] in A typical user To the nearest existing file The probability density function of the user's distance is equal to . and They are and The Laplace transforms are expressed as

[0114] (15)

[0115] (16)

[0116] Lemma 2 is used to obtain the probability of successful transmission from the user to the base station.

[0117] The complementary cumulative distribution function (CCDF) of the cellular link SINR is defined as is the probability of successful transmission from user to base station:

[0118] (17)

[0119] in:

[0120] (18)

[0121] (19)

[0122] Similarly, from base station to user The probability of successful transmission is expressed as:

[0123] (20)

[0124] in A typical user The probability density function of the distance to its serving base station is equal to . and They are and The Laplace transforms of are:

[0125] (twenty one)

[0126] (twenty two)

[0127] 3. Analysis Model of File Receiving

[0128] This embodiment establishes a queuing model for files requested by users and stored in its own cache, analyzes their arrival process, service process, and queue state transitions, and constructs a two-dimensional Markov chain to analyze the relationship between the queue state of the requested files and the offloading strategy. Based on an in-depth study of each transmission scheme, the quasi-birth-death process method is used to solve the steady-state probability distribution of the queue length, deriving expressions for key performance parameters.

[0129] 1 Queuing Model Analysis

[0130] Modeling users with caching capabilities, assuming that user requests arrive subject to parameters Bernoulli distribution, the total cache capacity of the user is , is divided into two parts. One part is used to store popular files pushed by the base station, called push-cache queue, and the cache space is used Indicates that its queue status is determined by the birth rate and death rate of the file itself, assuming that the period is greater than the cache period. The other part of the cache stores the files that users request and save from the surrounding or base station, which is called the request-cache queue. The cache space is .if or If the user's required file is already stored in the push-cache queue, the user does not need to request it externally. The state of the request-cache queue is related to the files already stored in the push-cache queue, the hit rate between the user and surrounding users, and the interference the D2D user experiences. To fully and comprehensively analyze the user's self-hit probability, the captain status of the user's request-cache queue is analyzed, and the state transition under different acquisition paths is discussed. Figure 2 Get a flow chart for user files.

[0131] Consider a time-slotted transmission system where the request and transmission of a file are completed in the same time slot. The queuing process is modeled as a two-dimensional Markov chain, and its state set is defined as ,in Indicates time slot The number of files requested by the user to be saved, defined There are three working modes in which the request file is satisfied, represented as

[0132] (twenty three)

[0133] So the state space of S , The state transition probability can be obtained by and The transfer is expressed as:

[0134] (24)

[0135] Figure 3 is the state transition diagram of user request-cache queue in D2D caching cellular network. The interrelation of state transition of queue length and working mode is summarized as the following four cases.

[0136] Case 1: When the working mode remains unchanged, the change of queue length is determined by the success transmission probability.

[0137] According to the different working mode, there are several cases, when the working mode is 0, i.e. the user request can be satisfied by its own cache, the change of queue length is determined by whether the request is generated and whether there is a file in push-cache queue to satisfy the request. When the working mode is 1, the state of queue length unchanged is caused by two cases, the first case is that no request is generated, then the number of files in queue is unchanged; the second case is that the request is generated and found in the surrounding users, transmitted in D2D mode but not successfully transmitted. When the working mode is 2, i.e. the file cannot be obtained from its own or other users' cache, but help from BS, and in the state of transmission from BS to user, the queue state unchanged is due to no request or request but the channel state is too bad to make the transmission unsuccessful. It can be expressed by formula as

[0138] (25)

[0139] Case 2: The working mode increases by 1, when the transmission mode is changed from self-satisfaction to sending request to surrounding users, it is divided into and two cases.

[0140] When , i.e. the current request-cache queue is empty, only when the pre-stored file cannot satisfy the request, it needs to request help from surrounding users. When , i.e. the request-cache queue is not empty, at this time, the reason why it needs to request file from surrounding users is that the files in the user push-cache queue cannot satisfy the requirement.

[0141] When the working mode is changed from D2D transmission to BS transmission, BS represents base station, it will only be because range does not contain the user with the file.

[0142] (26)

[0143] Case 3: Mode From D2D transmission or BS transmission mode to 0 state that satisfies itself, the case that the length of the queue is increased by 1 is that the request can be satisfied by D2D link or BS transmission and the sending is successful. The special case is that when the request-cache queue is full, the requested file cannot be stored in the queue, at this time the length of the queue does not change.

[0144] (27)

[0145] Case 4: Especially, when the working mode is unchanged, but the length of the queue is decreased by 1, which is due to the fact that after the request is generated, the file pushed by itself-cache cannot satisfy the request, and there is a file containing the required file in the request-cache file, according to the protocol, the request-cache file after being used again will be deleted, which is manifested as the length of the queue decreased by 1.

[0146] (28)

[0147] 2. Analysis of quasi-birth-and-death process transition probability matrix

[0148] The quasi-birth-and-death process QBD is used to evaluate the behavior of the queue, and the state of the length of the queue is rearranged to be regarded as a one-dimensional Markov process. The state set is called level , and the phase. The transition probability matrix of the QBD method is defined as:

[0149] (29)

[0150] where represents the transition probability sub-matrix of the state subset to .

[0151] (30)

[0152] For ,

[0153] (31)

[0154] The boundary conditions of the matrix , are:

[0155] (32)

[0156] ​ denotes the subset of states transitions to the subset of states The transition sub-matrix of the transition from the subset of states to the subset of states. According to case 4 of the previous section, denotes that the request of the user is satisfied by a file saved through the request, according to the protocol, only in this case the queue length of the requested file queue is reduced by 1:

[0157] (33)

[0158] denotes the subset of states transitions to the subset of states The transition sub-matrix of the transition from the subset of states to the subset of states. denotes the transition from the D2D state back to the self-unloading state, or from the base station satisfying state to the D2D transmission state:

[0159] (34)

[0160] 3 Performance parameters

[0161] The steady-state distribution of the quasi-birth-and-death process can be obtained from the linear balance equations where is the column matrix of the steady-state probabilities, in order to exploit the regular structure of the block matrix , we partition according to the levels of the quasi-birth-and-death process as follows:

[0162] (35)

[0163] where .

[0164] From the steady-state distribution, several important performance parameters can be obtained, the first one is the self-unloading probability, which refers to the probability that the request of the user is satisfied by a file in the push-cache queue or the request-cache queue, while through the previous analysis, the files in the request-cache queue are the personalized files saved by D2D transmission and BS transmission, the strategy proposed in this embodiment is to consider the possibility of reusing the files obtained through the request under the premise of considering the pre-storage mechanism, that is, to consider the two queues comprehensively, that is, to meet the demand for popular files, and to consider the personalized needs of users, to consider the average queue length according to the queuing theory, and to obtain the expression of the self-unloading probability:

[0165] (36)

[0166] The average queue length of the request-cache queue is expressed as:

[0167] (37)

[0168] ​The packet loss rate is that when the request-cache queue is full, the request cannot be satisfied by the push-cache queue and the request-cache queue. When the user obtains the file through the D2D link or cellular link, overflow packet loss occurs:

[0169] (38)

[0170] 4. Numerical and Simulation Results

[0171] This embodiment uses MATLAB simulation to verify the correctness of the results obtained under the above cache strategy. The simulation results are The plane area is repeated 10,000 times. The positions of users and base stations are modeled as independent Poisson point processes with densities of and . Set the path loss factor , , the user's transmission power is 20 , the Gaussian white noise power is -104 , Content Library , cache space , the distance for establishing D2D link is 25m, , , , .

[0172] Figure 4 Given The curves of the request-cache queue length changing with the request arrival rate under three parameters show that the theoretical value and the simulation result are consistent, proving the correctness of the theoretical deduction. As can be seen from the figure, The larger the value, the more popular the pushed files are, the greater the probability that these files will be requested, the less the system needs to request files from surrounding users, and the less likely the request-cache queue will be saturated. When the request arrival rate continues to increase, the push-cache queue cannot meet the requests, and the number of requested files increases, leading to queue saturation and overflow packet loss. Figure 5 This can be confirmed in.

[0173] exist Figure 5 Given in different The curve of the packet loss rate of the request-cache queue changes with the arrival rate of user requests. The theoretical value in the figure is consistent with the simulation. As the value of , the packet loss rate curve shifts to the right. This is because the Zipf parameter The larger the value, the more popular the files pushed by the base station tend to be. The requested files are more likely to be satisfied by the push-cache queue. If the request-cache queue is not saturated, no packet loss will occur. =0.5 and =0.9 two curves, the smaller The request-cache queue is saturated at a smaller request arrival rate, at which time the requests that the cache cannot satisfy ask for help from surrounding users, and the requested files cannot be saved and overflow packets are generated.

[0174] Figure 6 The influence of the joint cache strategy proposed in the embodiment on the self-unloading probability is given compared with the traditional request-cache strategy. The traditional request-cache strategy refers to that popular files are not pushed by the base station, and the files requested by the user are obtained from surrounding users or the base station and saved. The joint cache strategy proposed in the embodiment comprehensively considers that popular files are pushed and saved by the base station in a non-peak period, and the remaining files that cannot be satisfied by the push queue are provided by the D2D link. As can be seen from the figure, with the increase of the Zipf parameter, the joint cache strategy can effectively improve the self-unloading probability.

[0175] The present application is dedicated to analyzing the push and cache based content delivery in a wireless cache network. When the network load is in a non-peak state, the most popular content can be pushed to the users with cache through broadcasting, and then saved in the cache area for frequent access.

[0176] The base station and the user position are modeled as independent Poisson point processes, and then the queue state of the requested files in the user cache is analyzed. Considering the file transmission process of different access protocols, for the self-unloading state, the pre-stored files of the user satisfy the unloading probability, and for the file transmission through the D2D link, the radius between the users and the instantaneous SINR change divided by time slots determine the success transmission probability.

[0177] The length state of the requested file queue is further analyzed by a two-dimensional discrete time Markov chain, on the basis of in-depth research on each transmission scheme, the steady-state probability distribution is solved by using the quasi-birth-and-death process method, and the expression of the key performance parameter is obtained.

[0178] The simulation results show that the theoretical value and the simulation results are consistent, which proves the correctness of the theoretical derivation. The larger the Zipf parameter is, the more concentrated the popularity of the pushed files is, the greater the probability of the files being requested is, the less the system needs to request files from surrounding users, and the less likely the request-cache queue is saturated.

[0179] The joint cache strategy proposed in the embodiment comprehensively considers that popular files are pushed and saved by the base station in a non-peak period, and the remaining files that cannot be satisfied by the push queue are provided by the D2D link, and with the increase of the Zipf parameter, the joint cache strategy can effectively improve the self-unloading probability.

[0180] Although the present application has been disclosed with reference to the preferred embodiments, the embodiments and drawings are not intended to limit the present application, and any person skilled in the art, without departing from the spirit and scope of the present application, can make various changes or modifications, but the same is within the scope of protection of the present application. Therefore, the scope of protection of the present application should be defined by the scope of protection of the claims of the present application.

Claims

1. A caching method for jointly storing push files and request files, characterized in that: The following steps are involved: Step 1: For a D2D caching network where all mobile users have caching capabilities, base station and user locations are modeled as independent Poisson point processes. Step 2: Combining the randomness of content caching and content requests, analyze the queue status of the requested files in the user cache; Step 3: Using random geometry theory, derive a closed-form expression for the probability of successful offloading in the D2D cache network. Step 4: Build a Geo / G / 1 queue model for the user request cache queue, analyze its arrival process, service process, and queue state transition patterns, and then construct a two-dimensional discrete-time Markov chain that includes queue length state and transmission state. Step 5: Use the quasi-birth-death process method to solve the steady-state probability distribution of the Markov chain and obtain the expression of the key performance parameters.

2. A caching method for jointly storing push files and request files according to claim 1, characterized in that: In step 1: The location distribution of base stations follows the density PPP process , the user's location distribution follows the density PPP process , are independent and identically distributed; assuming that some users have caching capabilities, Indicates the proportion of users with cache function, , these users have the same cache capacity ; If the distance between the requesting user and the user storing the target file is less than When a D2D communication link is established between users, the file can be transferred to the requesting user.

3. The caching method for jointly storing push files and request files according to claim 2, characterized in that: Multimedia file collection is ,in is the total number of files, For file collection dimensions, all multimedia files have the same size of , user request The probability of any file in follows the popularity distribution function of the file , assuming the function Obey the Zipf distribution, expressed as: , in Reflects the skewness of the popularity function, high Indicates large file reusability; Indicates the file number, The physical meaning is The probability of The first few files with small values ​​have large popularity and correspond to most requests. Is a normalization factor used to ensure that the sum of the probabilities of all elements is 1, satisfying the basic properties of probability distribution. represents the normalization parameter; definition There is a probability set of each file stored for the user, where Indicates that the user has files The probability that , due to the limited capacity of D2D user buffer area, Must meet: , There are files The user density is expressed as , the user set is represented as .

4. A caching method for jointly storing push files and request files according to claim 3, characterized in that: In step 2: In a push-based cache cellular network, files cached via wireless communication are divided into a pre-storage phase and a delivery phase. In the pre-storage phase, the base station uses a broadcast channel to send popular files to users before the actual request, and stores them in the local cache in advance. The delivery phase is a unicast operation for requests that occurs when a user generates a request. When a user with a cache function generates a request, it first checks whether the required file is in its own cache. If not, the user sends a request to nearby users with a cache function. The user or base station that receives the request sends the requested file to the user. That is, file transmission only occurs in response to a request.

5. A caching method for jointly storing push files and request files according to claim 4, characterized in that: In step 2: For the center users, Represents a two-dimensional real space, the user communicates with the nearest base station, assuming that it follows the popularity function , randomly request any file in N. There are three ways to obtain the file the user needs: Self-unloading mode: When a request is generated, the user first checks whether the files in its own cache queue can satisfy the request. If so, it is immediately self-satisfied without delay; D2D offloading method: If the user making the request does not have a cache function or does not store the required file, the user will turn to nearby users to search for the required file, and the base station will check whether there is a user in the area around the user with the file cache. , establish a D2D link to send the required files from the surrounding users to the requesting user, and divert the service requests from the cellular link to the D2D transmission; this area is called the collaborative area, and it is assumed to be a circle with the user as the center. A circle with radius If the desired content is stored in the network, then under the condition of meeting SINR (Signal to Interference and Noise Ratio), a D2D communication link is established to help the node as the sender to send the file required by the user to the user; Base station response method: If the request cannot be satisfied through local cache or D2D offloading, it needs to be forwarded to the cellular base station and cellular network to respond and send the requested content. The base station then takes over and sends the file to the user via unicast.

6. A caching method for jointly storing push files and request files according to claim 1, characterized in that: In step 3, the cache hit probability includes the self-pushed-cache file hit probability and the D2D hit probability; The self-pushed-cache file hit probability is the probability that the required file is satisfied by the popular file pushed by the base station without requesting other devices; the dynamic process of the generation and disappearance of the base station pushed file is described by the birth and death process. For the service process, it refers to the existence time of the file in the user cache, and the obedient parameter is Exponential distribution, assuming that the user's push cache queue has a limited space capacity of , then after a period of time, the number of files stored in the queue is The probability of is expressed as: , in The rate at which files are generated. is the extinction rate, Push-cache queue for users, where m represents the number of files. represents a natural constant, and the offloading probability is defined as the ratio of offloading requests to total requests, and is expressed as Indicates that for a system with a pre-store mechanism, when a user generates a request, the required file has been pushed to the user and saved by the broadcast system, and the cache queue is pushed. The file hit probability when there are files is expressed as: , The average length of pre-stored files in the cache among users Expressed as: , The D2D hit probability: there is a cache and the file exists The user location distribution follows the density PPP process, then the file The probability of being hit by a D2D user is expressed as: 。 7. A caching method for jointly storing push files and request files according to claim 1, characterized in that: In step 3, the probability of successful file transmission is: For requests that cannot be satisfied by its own cache, the required files are transmitted by surrounding users or base stations. , when it is confirmed that there is no file in its own storage, the file will be The users in the network establish a D2D link, with the nearest user acting as the transmitter. As the receiving end, the file Send to user , assuming that multiple D2D links communicate on the same spectrum resources in a cell, the communication links from the base station to the user and between users experience large-scale fading models with different path loss factors, and the small-scale fading channels are modeled as mutually independent exponential distributions with a mean of 1, then the user The received signal-to-interference-and-noise ratio is expressed as: , in is the user's transmit power, and are the small-scale fading and transmission distance of the D2D link, is the path loss factor of the link, is the noise power, is the set of interferences received, given by: , The first term in the definition is , is from Other D2D transmitters The interference caused, and Requesting users To other users , Small-scale fading power gain, and user To other users The second term is , is from Base station that sends files to users The interference caused, Indicates the transmit power of the base station, represents the path loss factor from the base station to the user; If there is no file stored around the requesting user The user is the user Send files to the service base station To the user, the received signal-to-interference-and-noise ratio of the signal received from the base station is: , in Indicates the transmit power of the base station, and Users To its service base station Small-scale fading and transmission distance of the transmission link, is the sum of the interfering signal powers and is given by: , The first term in the definition is ,yes Medium The second item is the interference caused by base stations other than the one sending files to users. , from Other D2D transmitters The interference caused by the random geometry principle is used to describe the probability of successful transmission through the complementary cumulative distribution function CCDF of the cellular link SINR.

8. The caching method for jointly storing push files and request files according to claim 7, characterized in that: In the described D2D heterogeneous cellular network, the complementary cumulative distribution function (CCDF) of the SINR of the D2D link is defined as is the successful transmission probability of the D2D link, which is expressed as: , in and yes and The Laplace transform of is given by: , , Derivation process: , in is a user To the nearest existing file The probability density function of the user's distance is equal to ; and They are and The Laplace transforms are expressed as follows: ； ； The complementary cumulative distribution function (CCDF) of the cellular link SINR is defined as is the probability of successful transmission from user to base station: , in , , Base station to user The probability of successful transmission is expressed as: , in is a user The probability density function of the distance to its serving base station is equal to ; and They are and The Laplace transforms of are: , 。 9. A caching method for jointly storing push files and request files according to claim 1, characterized in that: In step 4: Modeling users with caching capabilities, assuming that user requests arrive subject to parameters Bernoulli distribution, the total cache capacity of the user is , is divided into two parts. One part is used to store popular files pushed by the base station, called push-cache queue, and the cache space is used Indicates that its queue status is determined by the birth rate and death rate of the file itself, assuming that the period is greater than the cache period; the other part of the cache stores the files that users request and save from the surrounding or base station, which is called the request-cache queue, and the cache space is ;like or If the file required by the user is stored in the server, the user does not need to request it from outside.

10. The caching method for jointly storing push files and request files according to claim 9, characterized in that: Consider a time-slotted transmission system where the request and transmission of a file are completed in the same time slot. The queuing process is modeled as a two-dimensional Markov chain, and its state set is defined as ,in Indicates time slot The number of files requested by the user to be saved, defined There are three working modes in which the request file is satisfied, represented as , So the state space of S , The state transition probability is given by and The transfer is expressed as: , The request-cache queue leader and working mode The mutual relationship of state transitions can be summarized into the following four cases: Case 1: When working mode When the same is kept, the change of captain is determined by the probability of successful sending; According to the different working modes, there are several cases. When the working mode is 0, that is, the user request can be satisfied by its own cache, then the captain The change is determined by whether a request is generated and whether there are files that meet the request in the push-cache queue; When the working mode is 1, the captain The unchanged state is caused by two situations. The first is that no request is generated, and the number of files in the queue remains unchanged. The second is that a request is generated and found among nearby users, and transmitted in D2D mode but is not sent successfully. When the working mode is 2, that is, the file cannot be obtained from the cache of itself or other users, it turns to the BS for help. In the transmission state from BS to user, the queue state The invariance is due to the fact that no request is generated or there is a request but the channel state is too bad to make the transmission successful; it can be expressed as: , Case 2: Working mode Add 1, when the transmission mode changes from self-satisfaction to surrounding users sending requests, it is divided into and Two situations; when When the current request-cache queue is empty, only when the pre-stored files cannot satisfy the request, it is necessary to ask for help from surrounding users; when When the request-cache queue is not empty, the reason for requesting files from surrounding users is that none of the files in the user's push-cache queue can meet the requirements; When working mode The state is sent by D2D transmission to BS, BS means base station, only because There is no user in the scope that contains the file; , Case 3: Working mode From D2D transmission or BS transmission mode to the self-satisfied state 0, the captain status increases by 1 when the request can be satisfied by the D2D link or BS transmission and sent successfully. In the special case, when the request-cache queue is full, the requested file cannot be stored in the queue, and the captain status does not change at this time: , Case 4: When working mode Unchanged, but captain status When a minus 1 change occurs, it is because after the request is generated, the self-pushed-cached file cannot satisfy it, and the request-cached file has Contains the required files. According to the agreement, the cached files will be deleted after being reused, which is shown as the captain Subtract 1: 。 11. A caching method for jointly storing push files and request files according to claim 1, characterized in that: In step 5: Use the quasi-birth-death process QBD to evaluate the queue behavior and rearrange the queue leader state, and regard it as a one-dimensional Markov process ;State set called horizontal , is the phase; the transition probability matrix of the QBD method is defined as: , in Represents a state subset Transfer to The transition probability submatrix of , for , , matrix , The boundary conditions are: , Represents a state subset Transfer to state subset The transfer matrix of Indicates that the user's request is satisfied by the file saved after the request. According to the protocol, only in this case, the length of the requested file queue is reduced by 1: , Represents a state subset Transfer to state subset The transition sub-matrix represents the transition from the D2D state back to the self-unloading state, or from the base station satisfied state to the D2D transmission state: , The steady-state distribution of the quasi-birth-death process is given by the linear equilibrium equations Find, where The column matrix of steady-state probabilities, in order to use the block matrix The rule structure of According to the level of pseudo-birth and death process, it can be divided into: , in: , Several performance parameters can be derived from the steady-state distribution. The first is the self-unloading probability, which refers to the probability that a user's request is satisfied by a file in the push-cache queue or the request-cache queue. Based on queueing theory and considering the average queue length, the self-unloading probability expression is: , The average length of the request-cache queue is expressed as: , The packet loss rate is that when the request-cache queue is full, the request cannot be satisfied by the push-cache queue and the request-cache queue. When the user obtains the file through the D2D link or cellular link, overflow packet loss occurs: 。

Images