Method and device for estimating resources of heterogeneous cache nodes of CDN

Abstract

The application relates to the technical field of communication, and provides a CDN heterogeneous cache node resource estimation method and device, which comprises the following steps: acquiring a data volume time sequence of requested cache content in each CDN heterogeneous cache node and an access record text sequence of each user terminal in a recent preset time period; constructing a CDN network topology graph with CDN content center nodes, CDN regional nodes, CDN edge service nodes and user terminals as nodes; inputting the data volume time sequence and the access record text sequence into a preset resource estimation model to obtain resource demand estimation values of each CDN heterogeneous cache node at a future preset time output by the resource estimation model. The application can more accurately and efficiently estimate the required resources of each CDN heterogeneous cache node according to user usage behavior preferences.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of communication technology, in particular to a CDN heterogeneous cache node resource estimation method and device. BACKGROUND

[0002] At present, in the content delivery network (CDN), the estimation of the resources (IT resources such as CPU, memory and bandwidth, etc.) required by each CDN heterogeneous cache node is mainly realized through the experience of the maintenance team. However, the CDN business faces customers widely, crosses operators and regions, and has the characteristics of demand diversity and demand burstiness, so it is difficult to capture the demand law through artificial experience, and there are problems of low accuracy of resource estimation of CDN heterogeneous cache nodes and time-consuming and laborious manual work. SUMMARY

[0003] The present application provides a CDN heterogeneous cache node resource estimation method and device to solve the problem of low accuracy of resource estimation of CDN heterogeneous cache nodes and time-consuming and laborious manual work in the prior art.

[0004] The present application provides a CDN heterogeneous cache node resource estimation method, comprising the following steps.

[0005] Obtain the data volume time sequence of the cached content requested in each CDN heterogeneous cache node and the access record text sequence of each user terminal in the recent preset time period.

[0006] Construct a CDN network topology graph with CDN content center nodes, CDN regional nodes, CDN edge service nodes and user terminals as nodes, the data transmission relationship between the four types of heterogeneous nodes as the edges of the CDN network topology graph, the data volume time sequence as the node features of the corresponding CDN heterogeneous cache nodes, and the access record text sequence as the node features of the corresponding user terminals.

[0007] Input the data volume time sequence and the access record text sequence into a preset resource estimation model to obtain the resource demand estimation value of each CDN heterogeneous cache node at a future preset time output by the resource estimation model.

[0008] The resource estimation model is trained based on the data volume time sequence sample, the access record text sequence sample and the resource demand amount label corresponding to the data volume time sequence sample at the future preset time.

[0009] According to the CDN heterogeneous cache node resource estimation method provided by the present application, the resource estimation model comprises a graph attention network and a long short-term memory neural network.

[0010] The graph attention network is used for fusing a target data volume time sequence and a neighbor reference sequence to obtain a target feature vector of the target node fused with neighbor node feature information, the target node being any one of a CDN content center node, a CDN region node and a CDN edge service node in a CDN network topology graph, the target data volume time sequence being a data volume time sequence of the target node, and the neighbor reference sequence being a data volume time sequence or an access record text sequence of a neighbor node directly connected with the target node.

[0011] The long short-term memory neural network is used for outputting a resource demand estimation value of the target node at a future preset time according to the target feature vector of the target node.

[0012] According to the CDN heterogeneous cache node resource estimation method provided by the application, the graph attention network comprises: a word embedding layer used for converting the target data volume time sequence into a fixed shape vector of a preset dimension to obtain a target data volume time sequence vector.

[0013] A first mapping layer is used for converting a first type of neighbor reference sequence dimension into a first type of neighbor reference sequence vector consistent with the dimension of the target data volume time sequence vector.

[0014] A second mapping layer is used for converting a second type of neighbor reference sequence dimension into a second type of neighbor reference sequence vector consistent with the dimension of the target data volume time sequence vector.

[0015] A first neighbor node level graph attention layer is used for fusing the target data volume time sequence vector and the first type of neighbor reference sequence vector and outputting a first type of intermediate feature vector of the target node fused with first type of neighbor node feature information.

[0016] A second neighbor node level graph attention layer is used for fusing the target data volume time sequence vector and the second type of neighbor reference sequence vector and outputting a second type of intermediate feature vector of the target node fused with second type of neighbor node feature information.

[0017] A target node level graph attention layer is used for fusing the first type of intermediate feature vector and the second type of intermediate feature vector to obtain the target feature vector.

[0018] According to the CDN heterogeneous cache node resource estimation method provided by the application, the long short-term memory neural network comprises: a long short-term memory layer used for extracting time features in the target feature vector and inputting the target feature vector with the extracted time features to a full connection layer.

[0019] The full connection layer is used for outputting a resource demand estimation value of the target node at a future preset time according to the target feature vector with the extracted time features.

[0020] According to the CDN heterogeneous cache node resource estimation method provided by the application, the loss function of the resource estimation model during training is: .

[0021] Wherein, represents the resource demand prediction value of the i th CDN heterogeneous cache node output by the resource estimation model at a future preset time during training, represents the resource demand label of the i th CDN heterogeneous cache node at a future preset time, and n represents the total number of CDN heterogeneous cache nodes.

[0022] According to the CDN heterogeneous cache node resource estimation method provided by the application, before the data volume time sequence and the access record text sequence are input into the preset resource estimation model, the data volume time sequence is standardized, and the access record text sequence of each user terminal is encoded into a sequence with the same length.

[0023] The application also provides a CDN heterogeneous cache node resource estimation device, comprising the following modules.

[0024] The data acquisition module is used to acquire the data volume time sequence of the requested cache content in each CDN heterogeneous cache node and the access record text sequence of each user terminal in a recent preset time period.

[0025] The topology construction module is used to construct a CDN network topology graph taking the CDN content center node, the CDN regional node, the CDN edge service node and the user terminal as nodes, taking the data transmission relationship among the four types of heterogeneous nodes as the edge of the CDN network topology graph, taking the data volume time sequence as the node feature of the corresponding CDN heterogeneous cache node, and taking the access record text sequence as the node feature of the corresponding user terminal.

[0026] The model execution module is used to input the data volume time sequence and the access record text sequence into the preset resource estimation model to obtain the resource demand estimation value of each CDN heterogeneous cache node at a future preset time output by the resource estimation model.

[0027] Wherein, the resource estimation model is trained based on the data volume time sequence sample, the access record text sequence sample and the resource demand label of the future preset time corresponding to the data volume time sequence sample.

[0028] The application also provides an electronic device comprising a memory, a processor and a computer program stored in the memory and executable on the processor, wherein the processor implements the CDN heterogeneous cache node resource estimation method according to any one of the above when executing the computer program.

[0029] The application further provides a non-transitory computer-readable storage medium, which has stored thereon a computer program, and the computer program, when executed by a processor, implements the CDN heterogeneous cache node resource estimation method according to any one of the above.

[0030] The application further provides a computer program product, which comprises a computer program, and the computer program, when executed by a processor, implements the CDN heterogeneous cache node resource estimation method according to any one of the above.

[0031] The CDN heterogeneous cache node resource estimation method and device provided by the application, by obtaining the data volume time sequence of the cache content requested in each CDN heterogeneous cache node in the recent preset time period and the access record text sequence of each user terminal, constructing a CDN network topology graph taking the CDN content center node, the CDN regional node, the CDN edge service node and the user terminal as nodes, taking the data transmission relationship among the four types of heterogeneous nodes as the edge of the CDN network topology graph, taking the data volume time sequence as the node feature of the corresponding CDN heterogeneous cache node, and taking the access record text sequence as the node feature of the corresponding user terminal, inputting the data volume time sequence and the access record text sequence into a preset resource estimation model to obtain the resource demand estimation value of each CDN heterogeneous cache node at a future preset time output by the resource estimation model, wherein the resource estimation model is trained based on the data volume time sequence sample, the access record text sequence sample and the resource demand amount label of the future preset time corresponding to the data volume time sequence sample. Since the resource estimation model is trained based on the data volume time sequence sample, the access record text sequence sample and the resource demand amount label of the future preset time corresponding to the data volume time sequence sample, the model can reflect the relationship between the demand amount of each CDN heterogeneous cache node and the access behavior of the user terminal, so that the required resources of each CDN heterogeneous cache node can be more accurately and efficiently estimated according to the user behavior preference, and the CDN service affected by resource shortage can be avoided to the greatest extent. BRIEF DESCRIPTION OF DRAWINGS

[0032] In order to more clearly illustrate the technical solutions of the present application or the prior art, the following will briefly introduce the drawings needed to be used in the embodiments or prior art description. Obviously, the drawings in the following description are some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0033] Figure 1 is a flowchart of the CDN heterogeneous cache node resource estimation method provided by the application.

[0034] Figure 2It is a CDN network topology schematic diagram constructed in the CDN heterogeneous cache node resource estimation method provided by the application.

[0035] Figure 3 It is a resource estimation model structure schematic diagram in the CDN heterogeneous cache node resource estimation method provided by the application.

[0036] Figure 4 It is a structure schematic diagram of the CDN heterogeneous cache node resource estimation device provided by the application.

[0037] Figure 5 It is a structure schematic diagram of the electronic device provided by the application. DETAILED DESCRIPTION

[0038] In order to make the purpose, technical scheme and advantages of the application clearer, the technical scheme in the application will be described clearly and completely below in combination with the drawings in the application. Obviously, the described embodiments are part of the embodiments of the application, rather than all the embodiments. Based on the embodiments in the application, all other embodiments obtained by those skilled in the art without creative labor belong to the protection scope of the application.

[0039] In order to better understand the content of the embodiments of the application, the network structure of the CDN network and the functions of the main service nodes are introduced as follows:

[0040] The main service nodes of the CDN network include: CDN content center nodes, CDN regional nodes and CDN edge service nodes, and the CDN edge service nodes are connected with user terminals. The CDN content center nodes, CDN regional nodes and CDN edge service nodes form a three-level network structure from top to bottom, and jointly constitute the service system of the CDN to realize the data service function. The data service function in the CDN network includes: access, storage / cache, processing (data slicing, transcoding, etc.) and distribution of user access content, etc.

[0041] The CDN content center node: mainly responsible for and connected with the business system, realizes CDN content access, management, storage and active distribution to the CDN regional nodes, and specifically includes the following functions.

[0042] Content access: according to the content injection instruction of the content management system (CMS), the specified content is obtained and injected into the content storage, and at the same time, the registration is carried out in the content management system; or the content is not injected through the CMS, but the content source is accessed by using the back-to-source mode.

[0043] Content distribution: according to the scheduling strategy of the content management system, the content is distributed and transmitted.

[0044] Local load balancing: receiving content positioning and requests of lower nodes (CDN regional nodes), conducting content addressing, and providing services according to load balancing strategies, and load balancing of devices in the nodes.

[0045] Content storage: storing content according to strategies in the content management system, and storing and updating content according to cache strategies.

[0046] Content processing: pre-processing of injected content, such as slicing, transcoding, and transencapsulation.

[0047] CDN regional node: mainly responsible for aggregating traffic of lower nodes (CDN edge service nodes), reducing back-to-source traffic, and specifically including the following functions.

[0048] Content distribution: distributing and transmitting content according to scheduling strategies of the content management system.

[0049] Local load balancing: receiving content positioning and requests of lower nodes, conducting content addressing, and providing services according to load balancing strategies, and load balancing of devices in the nodes.

[0050] Content cache: storing and updating content according to cache strategies.

[0051] CDN edge service node: as a main entity of CDN service, mainly responsible for receiving user terminal requests, verifying, and providing local cached content services to users, if the content is not hit, obtaining and caching content from the upper node, or providing services after redirection to the upper node, and specifically including the following functions.

[0052] Content service control: searching for content service nodes according to content IDs.

[0053] Global load balancing: receiving terminal service requests, and controlling edge service nodes to provide edge services to terminals according to content routing results, node states, load balancing strategies, and the like, load balancing of devices in the nodes, and providing node fault switching.

[0054] Content service: providing various content services, such as application services, streaming media services, and download services, and the like, for various business terminals.

[0055] Content processing: processing of content, such as slicing, transcoding, and transencapsulation.

[0056] Content cache: storing and updating content according to cache strategies.

[0057] Based on the CDN network, the CDN heterogeneous cache node resource estimation method of the embodiment of the application is as shown in the figure, and includes the following steps S110 to S130. Figure 1 ​

[0058] Step S110: Obtain the data volume time series of the requested contents in each CDN heterogeneous cache node and the access record text sequence of each user terminal in the recent preset time period. In the CDN network, the CDN content center node, the CDN regional node, the CDN edge service node, and the user terminal are four types of nodes that are heterogeneous (node type heterogeneous).

[0059] For the user terminal, when requesting resources, it will access the CDN edge service node, and the access record text sequence is the access record text recorded by time, for example: in the recent preset T period, each access is recorded in the order of access time, and the recorded content includes: access content and access duration, wherein the access content reflects the user's behavior preference.

[0060] For the CDN content center node, the CDN regional node and the CDN edge service node, the CDN edge service node will receive the content request of the user terminal, if the content is not hit, the CDN edge service node will request the content from the CDN regional node, if the content is not hit in the CDN regional node, the CDN regional node will request the content from the CDN content center node. Therefore, for the three types of nodes, the data volume time series of the requested contents in each CDN node in the recent preset time period can be obtained, and the elements in the sequence are the data volume of the requested contents in the CDN node in the preset time period recorded in time order.

[0061] Step S120: Construct a CDN network topology graph with CDN content center nodes, CDN regional nodes, CDN edge service nodes and user terminals as nodes, the data transmission relationship between the four types of heterogeneous nodes as the edges of the CDN network topology graph, the data volume time series as the node features of the corresponding CDN heterogeneous cache node, and the access record text sequence as the node features of the corresponding user terminal. As shown in Figure 2 , the CDN network topology graph constructed in this step is constructed according to the actual CDN network connection structure.

[0062] Specifically, for the CDN network topology graph of Figure 2 , the feature of the CDN content center node is the data volume of the requested contents in the CDN content center node in the preset time period, denoted as A={a1, a2, a3, …, an}, where a u represents the data volume time series of the requested contents in the u-th CDN content center node. u

[0063] ​The feature of the CDN regional node is the data volume of the cached content requested in a preset time period, denoted as B={b1, b2, b3, …, b v} where b v represents the time series of the data volume of the cached content requested in the vth CDN regional node.

[0064] The feature of the CDN edge service node is the data volume of the cached content requested in a preset time period, denoted as C={c1, c2, c3, …, c s} where c s represents the time series of the data volume of the cached content requested in the s th CDN edge service node.

[0065] The feature of the user terminal is the access record text sequence of the user terminal in a preset time period, denoted as D={d1, d2, d3, …, d z} where d z represents the access record text sequence of the z th user terminal.

[0066] Step S130: inputting the data volume time series and the access record text sequence into a preset resource estimation model to obtain the resource demand prediction value of each CDN heterogeneous cache node at a future preset time point output by the resource estimation model. The resource estimation model is trained based on the data volume time series sample, the access record text sequence sample, and the resource demand volume label of the future preset time point corresponding to the data volume time series sample. Specifically, the data volume time series sample and the access record text sequence sample are historical data, i.e., data in a historical T’ time interval, and the resource demand volume label can be a label matrix Y of the resource demand volume of each CDN heterogeneous cache node at L future time points to be predicted (relative to the future of the historical T’ time interval). The predicted resource demand volume of the i th CDN heterogeneous cache node can be represented as {y i t+1 , y i t+2 ,y i t+3 ,…, y i t+L-1 , y i t+L}, the shape (i.e., matrix dimension) is N*L, t represents the current time point, and N represents the number of resource types (including: CPU, storage, memory, and bandwidth, etc.).

[0067] The CDN heterogeneous cache node resource estimation method provided in the embodiment can reflect the relationship between the resource demand of each CDN heterogeneous cache node and the access behavior of the user terminal, so that the required resources of each CDN heterogeneous cache node can be more accurately and efficiently estimated according to the user usage behavior preference, and the CDN service affected by resource shortage can be avoided to the greatest extent.

[0068] In some embodiments, the resource estimation model comprises a graph attention network (GAT) and a long short-term memory (LSTM).

[0069] The graph attention network is used to fuse the target data volume time sequence and the neighbor reference sequence to obtain a target feature vector of the target node, which is any one of the CDN content center node, the CDN regional node and the CDN edge service node in the CDN network topology graph. The target data volume time sequence is the data volume time sequence of the target node, and the neighbor reference sequence is the data volume time sequence or the access record text sequence of the neighbor node directly connected to the target node. That is, the spatial dependency relationship between the three types of CDN heterogeneous cache nodes and the user terminal is captured by the graph attention network.

[0070] The long short-term memory neural network is used to output the resource demand estimation value of the target node at the future preset time according to the target feature vector of the target node. Specifically, the long short-term memory neural network is used to learn the dynamic changes of the data volume time sequence of the requested cache content and the user terminal behavior in the three types of CDN heterogeneous cache nodes to capture the time dependency relationship.

[0071] In the embodiment, the spatial dependency relationship between the three types of CDN heterogeneous cache nodes and the user terminal is captured by the graph attention network, and the time dependency relationship is captured by learning the dynamic changes of the data volume time sequence of the requested cache content and the user terminal access behavior in the three types of CDN heterogeneous environment nodes by the long short-term memory neural network, so that the required resources of the CDN content cache node can be more accurately and efficiently estimated according to the user usage behavior preference.

[0072] In some embodiments, as shown in Figure 3 The graph attention network comprises the following hierarchical structure.

[0073] The word embedding layer is used to convert the target data volume time sequence into a fixed shape vector of a preset dimension to obtain a target data volume time sequence vector.

[0074] The first mapping layer is configured to convert the first type of neighbor reference sequence dimension into a first type of neighbor reference sequence vector consistent with the dimension of the target data volume time sequence vector.

[0075] The second mapping layer is configured to convert the second type of neighbor reference sequence dimension into a second type of neighbor reference sequence vector consistent with the dimension of the target data volume time sequence vector.

[0076] The first neighbor node level graph attention layer is configured to fuse the target data volume time sequence vector and the first type of neighbor reference sequence vector and output a first type of intermediate feature vector of the target node fused with first type of neighbor node feature information.

[0077] The second neighbor node level graph attention layer is configured to fuse the target data volume time sequence vector and the second type of neighbor reference sequence vector and output a second type of intermediate feature vector of the target node fused with second type of neighbor node feature information.

[0078] The target node level graph attention layer is configured to fuse the first type of intermediate feature vector and the second type of intermediate feature vector to obtain the target feature vector.

[0079] By Figure 2 It can be known that the upper layer and the lower layer of the CDN region node and the CDN edge service node have other heterogeneous nodes. Therefore, there are two types of neighbor nodes for the CDN region node and the CDN edge service node, and the upper layer node can be the first type of neighbor node and the lower layer node can be the second type of neighbor node.

[0080] Like all attention mechanisms, the calculation of GAT is also divided into two steps: calculating the attention coefficient and weighted summation (aggregate). The GAT model is implemented by stacking graph attention layers. The input of each graph attention layer is the text and index feature set generated by the node.

[0081] .

[0082] The output is a new text and index feature set.

[0083] .

[0084] In order to calculate the weight of each neighbor node, an FxF' shared weight matrix W is applied to each node, and then the attention coefficient can be calculated, which can represent the importance of the node j relative to the node i . e ij.

[0085] .

[0086] In order to make the attention coefficient easier to calculate and facilitate comparison, a softmax is introduced to all i Adjacent nodes of j are regularized.

[0087] .

[0088] Therefore, the feature representation of the node i after the graph attention layer is: That is, the new feature (fusing neighborhood information) of each vertex i output by GAT, σ is an activation function.

[0089] .

[0090] Wherein N i represents the neighbor node set of node i , the coefficient α That is, the coefficient used for weighted summation at each convolution.

[0091] In some embodiments, the long short-term memory neural network comprises the following hierarchical structure.

[0092] The long short-term memory layer is used to extract the time feature in the target feature vector, and the target feature vector with the extracted time feature is input to the fully connected layer.

[0093] It should be noted that the target feature vector after the graph attention network is a target feature vector with spatial features, and after the long short-term memory layer, the target feature vector contains the spatial features and time features of the CDN network topology.

[0094] The fully connected layer is used to output the resource demand prediction value of the target node at a future preset time according to the target feature vector with the extracted time feature.

[0095] As shown in Figure 3 , the resource estimation model of the above embodiment is described in detail as follows taking the prediction of CDN edge service node resource demand as an example.

[0096] The graph attention network has two branches, branch 1 and branch 2, the structure of branch 1 includes the above-mentioned first word embedding layer, first mapping layer and first neighbor node level graph attention layer, and the structure of branch 2 includes: second word embedding layer, second mapping layer and second neighbor node level graph attention layer.

[0097] Branch 1: learn the weight of the first type of neighbor node (i.e. CDN regional node) of CDN edge service node i to evaluate its importance. Input the requested data volume time series h ci and the requested data volume time series h bi of the CDN regional node, which are converted into vectors after passing through the word embedding layer and the first mapping layer respectively, and then input into the neighbor node level graph attention layer, and output the new CDN edge service node features h ci 1 .

[0098] Branch 2: learn the weight of the second type of neighbor node (i.e. user terminal) of CDN edge service node i to evaluate its importance. Input the requested data volume time series h ci and the APP access record text sequence h di of the terminal user, which are converted into vectors after passing through the word embedding layer and the second mapping layer respectively, and then input into the neighbor node level graph attention layer, and output the new CDN edge service node features h ci 2 .

[0099] Where the word embedding layer (embedding): the target data volume time series is converted into a vector by using word embedding, the input data dimension is z, the output is set to convert the word into a 64-dimensional space vector, the input sequence length is F, so the shape of the output data of this layer is (None, F, 64). The role of this layer is to map the input word into a vector, and convert the index of each word into a 64-dimensional fixed shape vector.

[0100] Where the mapping layer: the conversion matrix of the mapping layer is used to convert the dimension of the running index features of the neighbor node (the first type of neighbor reference sequence or the second type of neighbor reference sequence) to the dimension consistent with the vector of the target data volume time series.

[0101] Where, the neighbor node level graph attention layer: the number of convolution kernels is 256, and the activation function is set to “relu”.

[0102] Then input the CDN edge service node new features h ci 1 fused with the CDN regional node feature information h ci 2 and the CDN edge service node new features h ci.

[0103] In the center node level graph attention layer, the number of convolution kernels is 128, and the activation function is set to "relu".

[0104] In some embodiments, the loss function of the resource estimation model during training is: .

[0105] wherein, represents the predicted value of the resource demand of the i-th CDN heterogeneous cache node at the future preset time output by the resource estimation model during training, represents the resource demand label of the i-th CDN heterogeneous cache node at the future preset time, and n represents the total number of CDN heterogeneous cache nodes.

[0106] Specifically, the model can be trained for 1000 iterations (epochs = 1000), the batch size is set to 32 (batch_size = 32), and the mean squared error (MSE) is selected as the loss function. The total sample data set can be divided into training data and test data, and 80% of the entire sample data set is taken as training data and the remaining 20% is taken as test data. The training set is used for training, and the test set is used to evaluate and verify the model. During training, only a group of data volume time series samples, access record text sequence samples, and resource demand labels corresponding to the future preset time of the data volume time series samples are input into the resource estimation model. The predicted value of the model is substituted into the loss function, and the model weight value is updated according to the value of the loss function. Thus, one iteration is completed. The next group of data volume time series samples, access record text sequence samples, and resource demand labels corresponding to the future preset time of the data volume time series samples are input into the resource estimation model until 1000 iterations and the loss function converges, and the model training is completed.

[0107] During the training process, the gradient descent optimization algorithm selects the adam optimizer to improve the learning speed of the traditional gradient descent (optimizer = 'adam'). The neural network can find the optimal weight value that minimizes the objective function through gradient descent. The neural network learns the weight value autonomously through training, and the weight of the model is derived after the model converges.

[0108] In some embodiments, before inputting the data volume time series and access record text sequence into the preset resource estimation model, the data volume time series is standardized, and the access record text sequence of each user terminal is encoded into a sequence of the same length.

[0109] Specifically, the data volume time series of the requested content in the three types of CDN heterogeneous cache nodes are standardized: (X-mean) / std. When calculating, each dimension is calculated separately, and the data is subtracted from the mean (mean) and divided by the variance std according to the attribute (according to the column). For data volume time series samples, the standardized data will improve the convergence speed of the model and improve the accuracy of the model during model training. During model application, the data volume time series also need to be standardized.

[0110] In this embodiment, the access record text sequence of each user terminal is encoded into a sequence representation, the encoding sequence length of each node feature is defined as F, the longest length F in the data set is taken as the encoding sequence length, and the length of each data is filled to F. The size of the dictionary is z. The same length of the encoding sequence length helps the model training, improves the convergence speed and accuracy of the model.

[0111] The CDN heterogeneous cache node resource estimation device provided by the application is described below. The CDN heterogeneous cache node resource estimation device described below can be correspondingly referred to the CDN heterogeneous cache node resource estimation method described above.

[0112] As shown in Figure 4 The CDN heterogeneous cache node resource estimation device of the application includes the following modules.

[0113] The data acquisition module 410 is configured to acquire the data volume time series of the requested content in each CDN heterogeneous cache node and the access record text sequence of each user terminal in a recent preset time period.

[0114] The topology construction module 420 is configured to construct a CDN network topology graph with CDN content center nodes, CDN regional nodes, CDN edge service nodes and user terminals as nodes, data transmission relationships between the four types of heterogeneous nodes as edges of the CDN network topology graph, data volume time series as node features of the corresponding CDN heterogeneous cache nodes, and access record text sequences as node features of the corresponding user terminals.

[0115] The model execution module 430 is configured to input the data volume time series and the access record text sequence into a preset resource estimation model to obtain resource demand estimation values of each CDN heterogeneous cache node at a future preset time output by the resource estimation model.

[0116] The resource estimation model is trained based on the data volume time series samples, the access record text sequence samples and the resource demand amount labels of the future preset time corresponding to the data volume time series samples.

[0117] The CDN heterogeneous cache node resource estimation device of the application, since the resource estimation model is trained based on the data volume time sequence sample, the access record text sequence sample and the resource demand amount label of the future preset time corresponding to the data volume time sequence sample, the model can reflect the relationship between the demand amount of each CDN heterogeneous cache node and the user terminal access behavior, so that the required resources of each CDN heterogeneous cache node can be more accurately and efficiently estimated according to the user usage behavior preference, and the CDN service affected by resource shortage can be avoided to the greatest extent.

[0118] In some embodiments, the resource estimation model comprises a graph attention network and a long short-term memory neural network.

[0119] The graph attention network is used to fuse the target data volume time sequence and the neighbor reference sequence to obtain a target feature vector of the target node, which is any one of the CDN content center node, the CDN regional node and the CDN edge service node in the CDN network topology graph, the target data volume time sequence is the data volume time sequence of the target node, and the neighbor reference sequence is the data volume time sequence or the access record text sequence of the neighbor node directly connected with the target node.

[0120] The long short-term memory neural network is used to output the resource demand estimation value of the target node at the future preset time according to the target feature vector of the target node.

[0121] In some embodiments, the graph attention network comprises the following hierarchical structure.

[0122] The word embedding layer is used to convert the target data volume time sequence into a fixed shape vector of a preset dimension to obtain a target data volume time sequence vector.

[0123] The first mapping layer is used to convert the first type of neighbor reference sequence dimension into a first type of neighbor reference sequence vector consistent with the dimension of the target data volume time sequence vector.

[0124] The second mapping layer is used to convert the second type of neighbor reference sequence dimension into a second type of neighbor reference sequence vector consistent with the dimension of the target data volume time sequence vector.

[0125] The first neighbor node level graph attention layer is used to fuse the target data volume time sequence vector and the first type of neighbor reference sequence vector and output a first type of intermediate feature vector of the target node fused with the first type of neighbor node feature information.

[0126] a second neighbor node level graph attention layer, configured to fuse the target data volume time sequence vector and the second type of neighbor reference sequence vector and output a second type of intermediate feature vector of the target node fused with second type of neighbor node feature information.

[0127] a target node level graph attention layer, configured to fuse the first type of intermediate feature vector and the second type of intermediate feature vector to obtain the target feature vector.

[0128] In some embodiments, the long short-term memory neural network comprises the following hierarchical structure.

[0129] a long short-term memory layer, configured to extract time features in the target feature vector, and input the target feature vector with the extracted time features to a fully connected layer.

[0130] a fully connected layer, configured to output a resource demand prediction value of the target node at a future preset time according to the target feature vector with the extracted time features.

[0131] In some embodiments, the loss function of the resource estimation model during training is: .

[0132] wherein, represents a resource demand prediction value of an i th CDN heterogeneous cache node output by the resource estimation model at a future preset time during training, represents a resource demand volume label of the i th CDN heterogeneous cache node at the future preset time, and n represents a total number of CDN heterogeneous cache nodes.

[0133] In some embodiments, the CDN heterogeneous cache node resource estimation apparatus further comprises a normalization processing module, configured to normalize the data volume time sequence before inputting the data volume time sequence and the access record text sequence into the preset resource estimation model, and encode the access record text sequence of each user terminal into a sequence with the same length.

[0134] Figure 5 An example of an entity structure diagram of an electronic device is shown in Figure 5 As shown in the figure, the electronic device can include a processor 510, a communications interface 520, a memory 530, and a communications bus 540, wherein the processor 510, the communications interface 520, and the memory 530 complete mutual communication through the communications bus 540. The processor 510 can invoke a logical instruction in the memory 530 to execute a CDN heterogeneous cache node resource estimation method, which comprises the following steps.

[0135] Obtain the data volume time sequence of the cached content requested in each CDN heterogeneous cache node and the access record text sequence of each user terminal in the recent preset time period.

[0136] Construct a CDN network topology graph with CDN content center nodes, CDN regional nodes, CDN edge service nodes and user terminals as nodes, the data transmission relationship between the four types of heterogeneous nodes as the edges of the CDN network topology graph, the data volume time sequence as the node features of the corresponding CDN heterogeneous cache nodes, and the access record text sequence as the node features of the corresponding user terminals.

[0137] Input the data volume time sequence and the access record text sequence into a preset resource estimation model to obtain the resource demand estimation value of each CDN heterogeneous cache node at a future preset time output by the resource estimation model.

[0138] The resource estimation model is trained based on the data volume time sequence sample, the access record text sequence sample, and the resource demand amount label corresponding to the future preset time of the data volume time sequence sample.

[0139] In addition, the logical instructions in the memory 530 described above can be implemented in the form of a software function unit and sold or used as an independent product, which can be stored in a computer readable storage medium. Based on this understanding, the technical solutions of the present application essentially or the parts that contribute to the prior art or parts of the technical solutions can be embodied in the form of a software product, which is stored in a storage medium, includes a number of instructions to make a computer device (which can be a personal computer, a server, or a network device, etc.) execute all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: a U disk, a mobile hard disk, a read-only memory (ROM, Read-Only Memory), a random access memory (RAM, Random Access Memory), a magnetic disk or an optical disk, and various media that can store program codes.

[0140] On the other hand, the present application also provides a computer program product, which comprises a computer program, the computer program can be stored on a non-transitory computer readable storage medium, and the computer program is executed by a processor, so that the computer can execute the CDN heterogeneous cache node resource estimation method provided by the above-mentioned methods, which comprises the following steps.

[0141] Obtain the data volume time sequence of the cached content requested in each CDN heterogeneous cache node and the access record text sequence of each user terminal in the recent preset time period.

[0142] A CDN network topology graph is constructed with CDN content center nodes, CDN regional nodes, CDN edge service nodes and user terminals as nodes, data transmission relationships among the four types of heterogeneous nodes as edges of the CDN network topology graph, data volume time series as node features of corresponding CDN heterogeneous cache nodes, and access record text sequences as node features of corresponding user terminals.

[0143] The data volume time series and the access record text sequences are input into a preset resource estimation model to obtain resource demand estimation values of each CDN heterogeneous cache node at a future preset time point output by the resource estimation model.

[0144] The resource estimation model is trained based on data volume time series samples, access record text sequence samples and resource demand labels of the future preset time point corresponding to the data volume time series samples.

[0145] In another aspect, the present application also provides a non-transitory computer readable storage medium having a computer program stored thereon, the computer program being executed by a processor to implement the CDN heterogeneous cache node resource estimation method provided by the above method, and the method comprises the following steps.

[0146] Data volume time series of cache content requested in each CDN heterogeneous cache node and access record text sequences of each user terminal in a recent preset time period are obtained.

[0147] A CDN network topology graph is constructed with CDN content center nodes, CDN regional nodes, CDN edge service nodes and user terminals as nodes, data transmission relationships among the four types of heterogeneous nodes as edges of the CDN network topology graph, data volume time series as node features of corresponding CDN heterogeneous cache nodes, and access record text sequences as node features of corresponding user terminals.

[0148] The data volume time series and the access record text sequences are input into a preset resource estimation model to obtain resource demand estimation values of each CDN heterogeneous cache node at a future preset time point output by the resource estimation model.

[0149] The resource estimation model is trained based on data volume time series samples, access record text sequence samples and resource demand labels of the future preset time point corresponding to the data volume time series samples.

[0150] The device embodiments described above are merely illustrative, wherein the units described as separate components can or can not be physically separate, and the components displayed as units can or can not be physical units, i.e., can be located in one place, or can be distributed to multiple network units. Part or all of the modules can be selected to achieve the purposes of the embodiments according to actual needs. Those skilled in the art can understand and implement without creative labor.

[0151] Through the description of the above embodiments, those skilled in the art can clearly understand that the embodiments can be realized by means of software and the necessary general hardware platform, and of course can also be realized by hardware. Based on such understanding, the above technical solutions can be embodied in the form of a software product, which can be stored in a computer readable storage medium, such as a ROM / RAM, a magnetic disk, an optical disk, etc., and includes a number of instructions to make a computer device (which can be a personal computer, a server, or a network device, etc.) execute the methods described in each embodiment or some parts of the embodiments.

[0152] Finally, it should be noted that: the above embodiments are only used to illustrate the technical solutions of the present application, and not to limit them; although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that: it can still modify the technical solutions recorded in the foregoing embodiments, or make equivalent replacement to part of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A method for estimating resources of heterogeneous CDN cache nodes, characterized in that, include: Get the time series of data volume requests for cached content in each CDN heterogeneous cache node within the most recent preset time period, as well as the text sequence of access records for each user terminal. A CDN network topology is constructed with CDN content center nodes, CDN regional nodes, CDN edge service nodes and user terminals as nodes. The data transmission relationship between the four types of heterogeneous nodes is used as the edge of the CDN network topology. The data volume time series is used as the node feature of the corresponding CDN heterogeneous cache node, and the access record text sequence is used as the node feature of the corresponding user terminal. Input the data volume time series and access record text series into a preset resource estimation model to obtain the estimated resource demand of each CDN heterogeneous cache node at a preset time in the future, as output by the resource estimation model. The resource estimation model is trained based on time series data samples, access record text sequence samples, and resource demand labels for future preset times corresponding to the time series data samples.

2. The CDN heterogeneous cache node resource estimation method according to claim 1, characterized in that, Resource estimation models include: graph attention networks and long short-term memory neural networks; The graph attention network is used to fuse the target data volume time series and the neighbor reference sequence to obtain a target feature vector of the target node that incorporates the feature information of the neighbor nodes. The target node is any node in the CDN network topology graph, including the CDN content center node, CDN regional node, and CDN edge service node. The target data volume time series is the data volume time series of the target node, and the neighbor reference sequence is the data volume time series or access record text sequence of the neighbor nodes directly connected to the target node. Long Short-Term Memory (LSTM) neural networks are used to output the estimated resource requirements of a target node at a predetermined future time based on the target feature vector of the target node.

3. The CDN heterogeneous cache node resource estimation method according to claim 2, characterized in that, The graph attention network includes: The word embedding layer is used to convert the target data volume time series into a fixed-shape vector of a preset dimension to obtain the target data volume time series vector; The first mapping layer is used to convert the dimension of the first type of neighbor reference sequence into a first type of neighbor reference sequence vector that is consistent with the dimension of the target data volume time series vector; The second mapping layer is used to convert the dimension of the second type of neighbor reference sequence into a second type of neighbor reference sequence vector that is consistent with the dimension of the target data volume time series vector; The first neighbor node level graph attention layer is used to fuse the target data volume time series vector and the first type of neighbor reference sequence vector and output the first type of intermediate feature vector of the target node which is fused with the feature information of the first type of neighbor nodes. The second neighbor node level graph attention layer is used to fuse the target data volume time series vector and the second type of neighbor reference sequence vector and output the second type of intermediate feature vector of the target node which is fused with the feature information of the second type of neighbor nodes. The target node-level graph attention layer is used to fuse the first type of intermediate feature vector and the second type of intermediate feature vector to obtain the target feature vector.

4. The CDN heterogeneous cache node resource estimation method according to claim 2, characterized in that, The long short-term memory neural network includes: A long short-term memory layer is used to extract the temporal features from the target feature vector, and the target feature vector with extracted temporal features is input to a fully connected layer. The fully connected layer is used to output the estimated resource demand of the target node at a future preset time based on the target feature vector from which time features are extracted.

5. The CDN heterogeneous cache node resource estimation method according to claim 2, characterized in that, The loss function of the resource estimation model during training is: ； in, This represents the predicted resource demand of the i-th heterogeneous CDN cache node at a preset time in the future, as output by the resource estimation model during training. This represents the resource demand label of the i-th CDN heterogeneous cache node at a future preset time, where n represents the total number of CDN heterogeneous cache nodes.

6. The CDN heterogeneous cache node resource estimation method according to any one of claims 1 to 5, characterized in that, Before inputting the data volume time series and access record text series into the preset resource estimation model, the following steps are also included: The data volume is standardized over time, and the access record text sequence of each user terminal is encoded into a sequence of the same length.

7. A CDN heterogeneous cache node resource estimation device, characterized in that, include: The data acquisition module is used to acquire the time series of data volume requested by cached content in each CDN heterogeneous cache node within the most recent preset time period, as well as the text sequence of access records of each user terminal. The topology construction module is used to construct a CDN network topology graph with CDN content center nodes, CDN regional nodes, CDN edge service nodes and user terminals as nodes. The data transmission relationship between the four types of heterogeneous nodes is used as the edge of the CDN network topology graph. The data volume time series is used as the node feature of the corresponding CDN heterogeneous cache node, and the access record text sequence is used as the node feature of the corresponding user terminal. The model execution module is used to input the data volume time series and access record text series into a preset resource estimation model to obtain the resource demand estimate of each CDN heterogeneous cache node at a preset time in the future, output by the resource estimation model. The resource estimation model is trained based on time series data samples, access record text sequence samples, and resource demand labels for future preset times corresponding to the time series data samples.

8. An electronic device, comprising: A memory, a processor, and a computer program stored on the memory and executable on the processor, characterized in that, when the processor executes the computer program, it implements the CDN heterogeneous cache node resource estimation method as described in any one of claims 1 to 6.

9. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the CDN heterogeneous cache node resource estimation method as described in any one of claims 1 to 6.

10. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by the processor, it implements the CDN heterogeneous cache node resource estimation method as described in any one of claims 1 to 6.

Images