A collaborative caching approach in a hierarchical network architecture

Abstract

The present invention discloses a collaborative caching method in a hierarchical network architecture, including: constructing a network edge caching model between a core server, a macro base station and each vehicle; dividing the nodes in the edge layer into multiple clusters by region, with the number of nodes in each cluster not exceeding a threshold value β; modeling a set of popular file sets through zipper distribution to find the most popular files for different users in different vehicle nodes; using coding cache to divide the content into separate coding packages to construct a content placement matrix m; taking minimizing total energy consumption as the optimization goal, further optimizing the specific value of the content placement matrix m and formulating a content placement strategy. The present invention aims to select content caching nodes and collaboratively place content while minimizing energy consumption, so as to have higher content delivery efficiency.

Description

Technical Field

[0001] The present invention relates to the field of vehicle traffic technology, and in particular to a collaborative caching method in a layered network architecture in a vehicle networking environment. Background Art

[0002] Intelligent transportation systems play a vital role in today's society, and most users of autonomous vehicles are more interested in entertainment services. There has been an unprecedented increase in user requests on the Internet, which often include watching and sharing videos, audios, and images on different platforms. In order to obtain better quality of service, seamless content delivery is essential. However, it is difficult to achieve long-term vehicle-to-vehicle (V2V) and vehicle-to-infrastructure (V2I) connections in highly dynamic environments. In addition, the short transmission range of infrastructure components such as roadside units (RSUs) and base stations (BSs) will provide an inefficient solution for V2I communications. Due to the limited range and high mobility, frequent connection loss and connection reestablishment will bring a lot of network traffic overhead and content retrieval delays.

[0003] In IoT, devices will connect to the network and share the collected data to the cloud server. After the data is analyzed and processed on the cloud server, users often request useful information. Different users can request the same content in the same time period, resulting in repeated transmission of the same data content. This has prompted operators to rethink their current network architecture and seek more advanced technologies to bring content closer to end users in a low-latency and energy-efficient manner. Mobile edge computing / caching devices within the coverage of macro base stations (MBS) enable content caching to reduce energy consumption through a layered heterogeneous network architecture.

[0004] However, due to the limited memory of edge devices and the high mobility of vehicles, within the staggered coverage of multiple layers of edge cache servers (e.g., MBS and RSU) in heterogeneous vehicle networks, cross-layer collaboration between different servers can provide seamless connectivity to facilitate content downloading.

[0005] The invention with patent number CN112839082A mentions a heterogeneous edge cache allocation method, system, medium and application, which takes into account the power allocation problem while considering the content cache allocation, and allocates different transmission powers to different contents to optimize the user experience. However, this patent does not involve the problem of encoding cache.

[0006] The invention with patent number CN113543071A mentions a method for collaborative edge caching of heterogeneous nodes in a connected vehicle environment, which aims to minimize the delay and cost required for vehicles to obtain total content and place content collaboratively with a high hit rate. However, the invention uses fountain codes. Summary of the invention

[0007] In view of the deficiencies in the prior art, the present invention provides a collaborative caching method in a layered network architecture under an Internet of Vehicles environment, aiming to select content caching nodes and collaboratively place content while minimizing energy consumption, thereby achieving higher content delivery efficiency.

[0008] To achieve the above object, the present invention adopts the following technical solutions:

[0009] The embodiment of the present invention proposes a collaborative caching method in a layered network architecture, and the collaborative caching method includes the following steps:

[0010] S1, construct a network edge cache model between the core server, the macro base station and each vehicle, the network edge cache model is set within the coverage of a macro base station, and includes a typical layered network architecture of two layers, the core layer and the edge layer; wherein the macro base station and each vehicle contained therein are regarded as one of the cache nodes, and the nodes in the edge layer collaborate to perform content caching; the core layer including the macro base station supports large bandwidth and fast data packet transmission while caching the uncached part of the edge layer;

[0011] S2, divides the nodes in the edge layer into multiple clusters by region, and the number of nodes in each cluster does not exceed a threshold β; all nodes belonging to a specific cluster collaboratively perform caching by updating cache information to their cluster heads, and the node threshold β is related to the communication overhead within the cluster nodes;

[0012] S3, a set of popular files is modeled through zipper distribution to find the most popular files for different users on different vehicle nodes;

[0013] S4, by using the code cache, splits the content into separate code packets and divides the nth file into l n file segments, and each file is of equal size; at the same time, the files in the popular file set are cached in descending order of popularity, and a content placement matrix m is constructed; when user u1 requests the nth file, m kn The file segments are obtained from the kth vehicle node in the local associated cluster, and the remaining l n -m kn Each file segment will be obtained through coded multicast transmission of the macro base station;

[0014] S5, taking minimizing the total energy consumption as the optimization goal, further optimize the specific value of the content placement matrix m and formulate a content placement strategy.

[0015] Furthermore, in step S1, the nodes in the edge layer cooperate to perform content caching, which means that only the selected vehicles are allowed to cache the information of other vehicles' caching events.

[0016] Furthermore, in step S2, the process of dividing the nodes in the edge layer into multiple clusters by region includes the following steps:

[0017] The entire edge network is divided into two groups, where the vehicle nodes in the same area are grouped together, and the number of nodes in each cluster does not exceed a threshold value β. Each group corresponds to an edge cluster, and each cluster is further divided into various node types: cluster head nodes, gateway nodes, and cluster members. All nodes belonging to a specific cluster collaborate to perform caching by updating cache information to their cluster head nodes, and there is no direct communication between cluster members.

[0018] Furthermore, the cluster head node selection process includes the following steps:

[0019] The following formula is used to calculate the priority evaluation value of each node in the cluster:

[0020]

[0021] Where w is the weight factor corresponding to different parameters, w1+w2+w3+w4+w5=1; d k is the number of nodes within the range of node k, is the average inverse of the node degree of all network nodes; T k is the transmission time between node k and the nodes within its range, is the average transmission time of all network nodes, H k is the weighted sum of the distances between node k and all nodes within its range, is the average weighted sum of all system width hops, p k is the transmission power of node k, is the average power utilization of all network nodes, c k is the cache utilization of node k, is the average cache space utilization of all network nodes.

[0022] Furthermore, in step S3, a set of popular files is modeled by zipper distribution, and the process of finding the most popular files for different users in different vehicle nodes includes the following steps:

[0023] S31, calculating the probability that a user requests the nth file in the kth node according to the following formula:

[0024]

[0025] In the formula, q kn represents the probability that the user of the kth SBS requests the nth file; α k is the compression index used to describe the skewness in the request pattern, α kA larger value means that content requests are more concentrated on high-ranking content files; kn is the ranking order of the nth file in the kth node;

[0026] S32, in each transmission period t, the kth vehicle node will receive an average of U from all relevant users in its coverage area k Requests; within the coverage of the kth node, the average number of requests for the nth file is defined as ω kn , using ω kn =U k q kn express;

[0027] S33, within the multicast period t, the probability that all users within the coverage of the kth node request a file j times is:

[0028]

[0029] Furthermore, in step S4, the content placement matrix m is:

[0030]

[0031] In the formula, m kn represents the number of file segments of the nth file cached by the kth vehicle node in each cluster; the value of m satisfies the following constraints:

[0032]

[0033] Where M k It represents the maximum capacity that each vehicle node can cache, and M represents the total file size.

[0034] Furthermore, in step S5, the optimization content placement matrix is ​​modeled as:

[0035]

[0036]

[0037]

[0038] Where, the constraints C1 and C2 are limits on the cache capacity, s0 is the size of each file segment, and M is the total file size; E C is the energy consumption of content caching, is the energy consumption of each vehicle node transmission, represents the probability that the user requests a file n times zero within the kth node range; is the transmission energy consumption of MBS, represents the energy consumption per bit of data transmitted by MBS, B represents the total number of bits transmitted by MBS through coded multicast; N is the total number of files in the popular file set.

[0039] The beneficial effects of the present invention are:

[0040] The collaborative caching method in the hierarchical network architecture proposed by the present invention, for the coded cache, in the placement phase, the content is not encoded and stored in the edge server, but in the transmission phase, different user requests can be satisfied simultaneously through the encoded multicast packets. The present invention can achieve the best memory rate trade-off. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] Figure 1 It is a schematic diagram of the structure of a network edge cache model according to an embodiment of the present invention.

[0042] Figure 2 It is a flow chart of a collaborative caching method in a layered network architecture according to an embodiment of the present invention. DETAILED DESCRIPTION

[0043] The present invention will now be described in further detail with reference to the accompanying drawings.

[0044] It should be noted that the terms such as "upper", "lower", "left", "right", "front", "back", etc. cited in the invention are only for the convenience of description and are not used to limit the scope of implementation of the present invention. Changes or adjustments in their relative relationships should be regarded as the scope of implementation of the present invention without substantially changing the technical content.

[0045] Figure 2 FIG. 1 is a flow chart of a collaborative caching method in a layered network architecture according to an embodiment of the present invention. Figure 2 , the caching method comprises the following steps:

[0046] S1, builds a network edge cache model between the core server, macro base station and each vehicle, which is set within the coverage of a macro base station.

[0047] Figure 1 It is a two-layered vehicle network architecture with caching capability. Figure 1 The model construction principle of this embodiment includes: designing a two-layer typical hierarchical network architecture consisting of a core layer and an edge layer, the core layer includes a macro base station and a remote server, and the edge layer is composed of vehicle users; the macro base station and each vehicle it contains are regarded as one of the cache nodes, and the nodes in the edge layer collaborate to perform content caching. The core layer supports large bandwidth and fast data packet transmission, and caches the uncached part of the edge layer.

[0048] All network users are represented by the set U = {u1,u2,...,u M}, all nodes of the network are represented by K, where k∈K specifies any network node. In order to effectively utilize resources, only selected vehicles are allowed to cache the information of other vehicles’ cache events; the popular file set is represented by N={1,2,...,N}; the corresponding file size is represented by s={s1,s2,...,s N}express.

[0049] Due to the limited cache space of edge layer nodes and the high speed of network vehicles, users can only receive a portion of the requested multimedia files and cache the remaining portion in the core layer through encoded transmission.

[0050] S2, divide all nodes in the edge layer into multiple clusters.

[0051] Each cluster is further divided into various node types, cluster head nodes (CH), gateway nodes (GN), and cluster members (CM). All nodes belonging to a particular cluster collaboratively perform caching by updating cache information to their cluster head (CH), where, unlike traditional collaborative caching, direct communication between members of a single cluster is not possible.

[0052] The CH is mainly responsible for caching the updated cache event information of all its CMs, and provides communication links between nodes in its own cluster and nodes in neighboring clusters through the GN. The main responsibility of the GN is the transmission of packets between clusters, and it participates in the cluster head selection process. The remaining nodes are all considered cluster members, and the CM is the node that actually caches the content based on its popularity. All CMs participate in the cluster head selection process.

[0053] Clustering head CH selection method: The traditionally designed WCA algorithm considers four factors to calculate the weight function between the node degree of a particular node and the node threshold β. For WCA, a weight function is calculated using all the factors together, which in turn produce a value to score each network node. The node with the lowest score among all nodes is selected as the cluster head. This patent also follows the WCA method for clustering, while considering different factors as shown in the following formula:

[0054]

[0055] In the formula, w1, w2, w3, w4, and w5 are weight factors corresponding to different parameters, expressed as:

[0056] w1+w2+w3+w4+w5=1

[0057] The first factor corresponds to the d of node k kThe degree (i.e., the nodes within the transmission range of node k, also called the neighbors of node k) is calculated as:

[0058]

[0059] Where dist(k,q)=1 corresponds to q being within the transmission range of k, and all network nodes The average inverse of the node degree can be used express.

[0060] The second factor is the average transmission time T from node k to all nodes within k. k In this way, T k It is calculated as:

[0061]

[0062] In the formula, S k For k and t q,k The set of nodes in the range, k represents the transmission time from node k to any node q, and the average transmission time of all network nodes is express.

[0063] The third factor is the weighted sum of the distances between node k and its neighbor hop points, H k To reduce the network load of sending and receiving buffered information, the hop count from CH to CM should be kept to a minimum. k The value is Indicates that count m is the count of the m-th hop neighbor of node k, γ m is the weight factor. Similarly, the average weighted sum of all system wide hops is calculated using express.

[0064] The fourth factor is to designate node k as the cumulative power P during CH k , the average power utilization of all network nodes can be calculated as

[0065] The fifth factor is the cache utilization C of node n k , mainly in CH, the average cache space utilization of all network nodes can be calculated as

[0066] Cluster member CM selection method: In order to reduce the communication overhead within the cluster nodes, a node threshold β is set for each cluster, that is, the upper limit of the maximum number of nodes that can be accommodated in a cluster excluding CH. This limits the maximum number of nodes that the cluster can manage.

[0067] S3, models a set of popular files through zipper distribution and finds different users’ preferences for files in different vehicle nodes.

[0068] Different users have different preferences for files in different vehicle nodes, so a popular file in one vehicle node may not be popular in other nodes. Assuming that the popularity distribution of files follows a zipper distribution, the probability of a user requesting the nth file in the kth node is calculated according to the following formula:

[0069]

[0070] In the formula, q kn represents the probability that the user of the kth SBS requests the nth file, α k is the compression index used to describe the skewness in the request pattern, α k A larger value means that content requests are more concentrated on high-ranking content files; kn is the ranking order of the nth file in the kth node.

[0071] In each transmission period t, the kth vehicle node will receive an average of U from all relevant users in its coverage area. k Requests. Within the coverage of the kth node, the average number of requests for the nth file is defined as ω kn , using ω kn =U k q kn express.

[0072] The number of requests for the nth file by users in the coverage of the kth vehicle node follows a Poisson process. Therefore, within the multicast period t, the probability that all users in the coverage of the kth node request a file j times can be written as:

[0073]

[0074] S4, by using codecache, splits the content into separate code packets.

[0075] Due to the limitation of cache size of each cluster node and the heterogeneity of files, each node in the edge layer can only cache part of the files. By using code cache, the content is divided into separate code packets, and the nth file is divided into l n file segments, and the size of each file segment is equal to s0, m kn represents the number of file segments of the nth file cached by the kth edge device in each cluster. Therefore, the content placement matrix m can be expressed as:

[0076]

[0077] User u1 requests the nth file and needs to get its l n File segments, m kn The file segments are obtained from the kth vehicle node in the local associated cluster, and the remaining l n -m kn The file segments will be obtained through coded multicast transmission of the macro base station. Through optimization, the specific value of the content placement matrix m can be obtained to form a content placement strategy. Due to the cache size of the vehicle node, the value of m should meet the following constraints:

[0078]

[0079] Where M k It represents the maximum capacity that each vehicle node can cache, and M represents the total file size.

[0080] S5, with minimizing the energy consumption of users obtaining content as the optimization goal, optimizes the network edge cache model, optimizes the specific value of the content placement matrix m, and places the vehicle transmission content; among them, the popular files in the popular file set are cached in order from high to low popularity.

[0081] First, we analyze the energy consumption of content caching, which is defined as E c ,During the placement phase, each file is partially cached in the vehicle node, resulting in energy consumption. In general, the cache energy consumption is proportional to the number of file bits stored on each node. Then, as long as the content placement matrix m is known, the cache energy consumption can be obtained by calculating the number of data bits cached by each node. The calculation is expressed as:

[0082]

[0083] In the formula, e ca is the energy coefficient of the vehicle node cache in joules / bit, and s0 is the size of each file segment.

[0084] Secondly, the energy consumption of vehicle node transmission is analyzed, namely In the transmission phase, when a user initiates a request, the associated node will first transmit its cached file segments to the user. The remaining file segments are transmitted by the MBS to the relevant vehicle nodes, and then the associated nodes transmit them to the user. Therefore, the energy consumption of transmitting files from the vehicle node includes two parts: one is the energy consumption of transmitting its own cached file segments to the user, and the other is the energy consumption of the uncached file segments obtained from the MBS. The total energy consumption is calculated as:

[0085]

[0086] In the formula, is the energy consumption per bit of data transmitted by the vehicle node, It indicates the probability that the user requests the file n times zero within the range of the kth node.

[0087] Due to the limitation of the cache size of the vehicle node, some file segments cannot be obtained from the cluster. These file segments will be transmitted from the MBS to the vehicle node through the backhaul link. Therefore, the energy consumption of transmitting files on the MBS is mainly proportional to the number of bits of the file transmitted on the MBS. The transmission energy consumption of the MBS can be expressed as:

[0088]

[0089] In the formula, represents the energy consumption per bit of data transmitted by MBS, and B represents the total number of bits transmitted by MBS through coded multicast, which can be expressed as:

[0090]

[0091] In the formula, K j represents a vehicle node set consisting of j, 1≤j≤K,

[0092] Assuming that the cache size of each vehicle node is known, the k-th vehicle node content placement optimization problem is modeled as:

[0093]

[0094]

[0095]

[0096] Wherein, the constraints C1 and C2 are the limits on the cache capacity.

[0097] By finding the content placement of the kth vehicle node, the content placement of all vehicle nodes in each cluster can be deduced in this way to obtain the specific value of the content placement matrix m.

[0098] The above are only preferred embodiments of the present invention, and the protection scope of the present invention is not limited to the above embodiments. All technical solutions under the concept of the present invention belong to the protection scope of the present invention. It should be pointed out that for ordinary technicians in this technical field, some improvements and modifications without departing from the principle of the present invention should be regarded as the protection scope of the present invention.

Claims

1. A collaborative caching method in a hierarchical network architecture, characterized in that: The collaborative caching method comprises the following steps: S1, construct a network edge cache model between the core server, the macro base station and each vehicle, the network edge cache model is set within the coverage of a macro base station, and includes a typical layered network architecture of two layers, the core layer and the edge layer; wherein the macro base station and each vehicle contained therein are regarded as one of the cache nodes, and the nodes in the edge layer collaborate to perform content caching; the core layer including the macro base station supports large bandwidth and fast data packet transmission while caching the uncached part of the edge layer; S2, divides the nodes in the edge layer into multiple clusters by region, and the number of nodes in each cluster does not exceed a threshold β; all nodes belonging to a specific cluster collaboratively perform caching by updating cache information to their cluster heads, and the node threshold β is related to the communication overhead within the cluster nodes; S3, a set of popular files is modeled through zipper distribution to find the most popular files for different users on different vehicle nodes; S4, by using the code cache, splits the content into separate code packets and divides the nth file into l n file segments, and each file is of equal size; at the same time, the files in the popular file set are cached in descending order of popularity, and a content placement matrix m is constructed; when user u1 requests the nth file, m kn The file segments are obtained from the kth vehicle node in the local associated cluster, and the remaining l n -m kn Each file segment will be obtained through coded multicast transmission of the macro base station; S5, taking minimizing total energy consumption as the optimization goal, further optimizing the specific value of the content placement matrix m and formulating a content placement strategy; In step S3, a set of popular files is modeled by zipper distribution, and the process of finding the most popular files for different users in different vehicle nodes includes the following steps: S31, calculating the probability that a user requests the nth file in the kth node according to the following formula: In the formula, q kn represents the probability that the user of the kth SBS requests the nth file; α k is the compression index used to describe the skewness in the request pattern, α k A larger value means that content requests are more concentrated on high-ranking content files; kn is the ranking order of the nth file in the kth node; S32, in each transmission period t, the kth vehicle node will receive an average of U from all relevant users in its coverage area k Requests; within the coverage of the kth node, the average number of requests for the nth file is defined as ω kn , using ω kn =U k q kn express; S33, within the multicast period t, the probability that all users within the coverage of the kth node request a file j times is: In step S4, the content placement matrix m is: In the formula, m kn represents the number of file segments of the nth file cached by the kth vehicle node in each cluster; the value of m satisfies the following constraints: Where M k represents the maximum capacity that each vehicle node can cache, and M represents the size of the total file; In step S5, the optimized content placement matrix is ​​modeled as: Where, the constraints C1 and C2 are the limits on the cache capacity, s0 is the size of each file segment, and M is the total file size; E C is the energy consumption of content caching, e ca Energy coefficient of vehicle node cache in joules / bit; is the energy consumption of each vehicle node transmission, Energy consumption per bit of data transmitted for the vehicle node; represents the probability that the user requests a file n times zero within the kth node range; is the transmission energy consumption of MBS, represents the energy consumption per bit of data transmitted by MBS, B represents the total number of bits transmitted by MBS through coded multicast; N is the total number of files in the popular file set.

2. The collaborative caching method in a hierarchical network architecture according to claim 1, characterized in that: In step S1, the nodes in the edge layer cooperate to perform content caching, which means that only selected vehicles are allowed to cache information of other vehicles' cache events.

3. The collaborative caching method in a hierarchical network architecture according to claim 1, characterized in that: In step S2, the process of dividing the nodes in the edge layer into multiple clusters by region includes the following steps: The entire edge network is divided into groups, and the vehicle nodes in the same area are divided into a group. The number of nodes in each cluster does not exceed the threshold β. Each group corresponds to an edge cluster, and each cluster is further divided into various node types: cluster head nodes, gateway nodes, and cluster members. All nodes belonging to a specific cluster collaborate to perform caching by updating the cache information to its cluster head node, and there is no direct communication between cluster members.

4. The collaborative caching method in a layered network architecture according to claim 3, characterized in that: The cluster head node selection process includes the following steps: The following formula is used to calculate the priority evaluation value of each node in the cluster: Where w is the weight factor corresponding to different parameters, w1+w2+w3+w4+w5=1; d k is the number of nodes within the range of node k, is the average inverse of the node degree of all network nodes; T k is the transmission time between node k and the nodes within its range, is the average transmission time of all network nodes, H k is the weighted sum of the distances between node k and all nodes within its range, is the average weighted sum of all system width hops, p k is the transmission power of node k, is the average power utilization of all network nodes, c k is the cache utilization of node k, is the average cache space utilization of all network nodes.

Images