A method for allocating network slice resources

Abstract

The application discloses a network slice resource allocation method, comprising the following steps: step 1, obtaining a plurality of user data, setting a corresponding threshold value for each characteristic in the user data, judging whether the characteristic value in the user data is greater than the corresponding threshold value, if yes, assigning 1 to the characteristic, if no, assigning 0 to the characteristic, and completing user data initialization; step 2, defining the characteristic frequent occurrence times as the total times of the characteristic value being 1 in single user data, scanning all the initialized user data to obtain the characteristic frequent occurrence times of each characteristic; step 3, clustering according to the characteristic frequent occurrence times, and making a hypergraph according to the clustering result; step 4, solving the normalized weight of the hyperedge in the hypergraph; and step 5, judging the strong correlation between characteristics by using the obtained hypergraph and formulating a network slice resource allocation scheme. The application can reasonably, efficiently and fairly allocate network slice resources.

Description

Technical Field

[0001] This invention relates to the field of 5G network technology, and in particular to a method for allocating network slice resources. Background Technology

[0002] With the rise of 5G, the high speed, low latency, and wide connectivity of 5G networks have made emerging communication scenarios such as ultra-reliable and low-latency communication, enhanced mobile broadband, and massive machine-type communication possible. However, traditional network architectures cannot support these emerging communication service requests. Network slicing is an effective solution to this problem. By separating logical and physical resources, it supports service customization and isolation on the infrastructure of a public physical network. Using network slicing, operators can establish several dedicated, virtualized, and isolated logical networks on a common underlying physical network system to adapt to various user needs in terms of latency, bandwidth, security, and reliability of the Internet. This not only simplifies the network architecture but also effectively reduces network operation and maintenance costs. However, to ensure and improve user experience and network resource utilization, it is necessary to establish novel and reasonable resource management mechanisms to meet the different service functions of multiple logical networks, i.e., network slices, on the same physical network. Therefore, researching resource management algorithms in a network slicing environment is of great significance.

[0003] Most existing dynamic allocation schemes for network slice resources lack consideration of the trade-off between the conflicting needs of resource utilization and tenant isolation. Improving operator resource utilization efficiency while flexibly ensuring network slice performance isolation is one of the challenges in dynamic allocation of network slice resources. The fundamental reason for this conflict is the random variation in slice load, leading to random changes in the actual resource requirements of slices. Rigid fixed resource reservation schemes cannot adapt to the time-varying actual resource demands of slices. How to dynamically allocate resources to slices based on load variation characteristics is another challenge that needs to be addressed. Furthermore, most existing network slice mapping problems are based on fixed resource reservation assumptions, lacking optimization for resource elasticity and utilization. To improve the resource utilization efficiency and elastic allocation of mapping problems, establishing a flexible resource sharing scheme is one of the challenges in resource-sharing-oriented slice mapping problems. Therefore, to meet users' needs for a comfortable experience and improve network resource utilization, a fair and easily implemented resource management mechanism is needed, enabling network slices to implement different service functions on the same physical network. Summary of the Invention

[0004] The purpose of this invention is to provide a method for allocating network slice resources. This invention can allocate network slice resources reasonably, efficiently, and fairly, so that the overall utility, operating cost benefits, and resource utilization of the network achieve ideal results, while also meeting the differentiated needs of users.

[0005] The technical solution of this invention: a method for allocating network slice resources, comprising the following steps:

[0006] Step 1: Obtain multiple user data sets, set a corresponding threshold for each characteristic in the user data, and determine whether the characteristic value in the user data is greater than the corresponding threshold. If it is greater, assign the value 1 to the characteristic; if it is less than the threshold, assign the value 0. This completes the user data initialization.

[0007] Step 2: Define the frequency of a feature as the total number of times the feature value is 1 in a single user's data. Scan all initialized user data to obtain the frequency of each feature.

[0008] Step 3: Cluster based on the frequency of occurrence of features, and create a hypergraph based on the clustering results;

[0009] Step 4: Solve for the normalized weights of the hyperedges in the hypergraph;

[0010] Step 5: Use the obtained hypergraph to determine the strong correlation between features and formulate a network slice resource allocation scheme. At the same time, use the normalized weights of the hyperedges to avoid the problem of difficulty in determining the priority between different features due to the different attributes of the features themselves, so as to implement a fair and reasonable allocation.

[0011] In the above-mentioned method for allocating network slice resources, step 3 involves a clustering process where a minimum support k is set based on actual needs. All features in the user data are used as candidate sets to be scanned. After scanning all features, candidate sets whose frequency of occurrence of the same feature meets the minimum support are retained to obtain a first-level candidate set. Then, based on the first-level candidate set, candidates are selected pairwise to form a new set to be scanned, forming a second-level candidate set. The user data is then scanned again, and the frequency of occurrence of the second-level candidate set is counted. Candidates with a support less than the minimum support k are removed. Then, based on the second-level candidate set, three features are selected to form a new set to be scanned, forming a third-level candidate set. The user data is scanned again, and the frequency of occurrence of the third-level candidate set is counted. This process is repeated until the end.

[0012] In the aforementioned method for allocating network slice resources, step 3 involves creating the hypergraph by setting the characteristics of each set in the obtained clustering results to different vertices within the same hyperedge, resulting in a weightless hypergraph G; then, within the hypergraph G, for each hyperedge e... i Assign a positive weight w(e)i Let w(e) i The value of w(e) is numerically equal to the frequency of occurrence of the characteristic of the candidate option. i Let V be the weight of the hyperedge, and we get the weighted hypergraph G(V,E,W), where V represents the set of vertices, E represents the set of hyperedges, and W represents the set of positive weights.

[0013] In the aforementioned method for allocating network slice resources, step 4 involves calculating the normalized weights as follows:

[0014] For a weighted hypergraph G(V,E,W), when vertex v i ∈e i It is called superedge e i With vertex v i Related; for any v i ∈V, vertex v i degree d(v) i Let ) denote the number of hyperedges associated with a vertex; then for any hyperedge e i ∈E, its degree is defined as:

[0015] δ(e i )=|e i |;

[0016] The vertex degree matrix is ​​given by D e Represent the sum of the vertex degrees of all vertices, denoted as D. k The diagonal matrix of the hyperedge is W H To represent;

[0017] In a hyperedge, every node has δ(e)-1 edges connecting it to another hyperedge node. Therefore, the normalized matrix D can be obtained through the formula. f :

[0018]

[0019] From the normalized matrix D f The normalized weights of the superedges are obtained.

[0020] Compared with existing technologies, this invention provides a network slice resource allocation method by applying hypergraph clustering, thereby achieving a reasonable, efficient, and fair allocation. Hypergraph clustering is a clustering method for high-dimensional data, which can effectively reduce the impact of the curse of dimensionality. Hypergraph-based clustering maps the relationships of high-dimensional data onto a hypergraph, where each edge represents the relationship between these data, and the weight of the edge indicates the closeness of the relationship. Therefore, this invention uses the obtained hypergraph to determine the strong correlation between characteristics and formulate a network slice resource allocation scheme. Simultaneously, it utilizes the normalized weights of hyperedges to avoid the problem of difficulty in determining the priority between different characteristics due to differences in their inherent attributes, thus implementing a fair and reasonable allocation. From the perspective of network operators, the overall system utility, network operating cost benefits, and resource utilization achieve ideal results. From the user's perspective, on the one hand, user needs are met and service quality is improved; on the other hand, personalized services are realized, providing customized and fair services based on the differentiated needs of different users. Attached Figure Description

[0021] Figure 1 This is a flowchart illustrating the present invention;

[0022] Figure 2 This is an example diagram of a simulated sample of the present invention;

[0023] Figure 3 This is a schematic diagram of the clustering process;

[0024] Figure 4 Representative according to Figure 3 Hypergraph generated from clustering results;

[0025] Figure 5 The matrix is ​​obtained from the hypergraph and the normalized matrix is ​​derived from each matrix. Detailed Implementation

[0026] The present invention will be further described below with reference to the accompanying drawings and embodiments, but this should not be construed as limiting the present invention.

[0027] Example: A method for allocating network slice resources, such as... Figure 1 As shown, it includes the following steps:

[0028] Step 1: Obtain g user data points. Set a corresponding threshold for each characteristic in the user data. The total number of thresholds is g, denoted as a. n(n = 1, 2, 3...g) represents the threshold; scan all user data samples for characteristic values, and determine whether the characteristic value in the user data is greater than or equal to the corresponding threshold. If it is greater, assign a value of 1 to the characteristic; if it is less than, assign a value of 0. By default, a characteristic with a value of 1 indicates that the user has a more prominent need, and a characteristic with a value of 0 indicates that the user has a less prominent need. This completes the user data initialization. In this embodiment, as... Figure 2 As shown, five features are presented to represent the different needs of five different users, with a threshold 'a' manually set for each feature. n Each feature is compared with its corresponding threshold. If it is higher than the threshold, 1 is entered in the initialized table, indicating that the user has a high demand for the feature. Otherwise, 0 is entered, indicating that the user does not have a high demand for the feature.

[0029] Step 2: Define the frequency of a feature as the total number of times the feature value is 1 in a single user's data. Scan all initialized user data to obtain the frequency of each feature.

[0030] Step 3: Cluster based on the frequency of occurrence of features, and create a hypergraph based on the clustering results. The clustering process involves setting a minimum support k according to actual needs, using all features in the user data as candidate sets to be scanned, and after scanning all features, retaining the candidate sets whose frequency of occurrence of the same feature satisfies (i.e., is greater than or equal to) the minimum support to obtain the first-level candidate set. Then, based on the first-level candidate set, select candidate features pairwise to form a new set to be scanned to form a second-level candidate set. The user data is then scanned again, and the frequency of occurrence in the second-level candidate set is counted. Candidates with a support less than the minimum support k are removed. Then, based on the second-level candidate set, three features are selected to form a new set to be scanned to form a third-level candidate set. The user data is scanned again, and the frequency of occurrence in the third-level candidate set is counted. The above operations are repeated until the final step is completed. Figure 3 As shown, Figure 3 This is a schematic diagram of the clustering process. First, the initialized data is scanned, recording the frequency of each feature being 1. A minimum support value k is artificially assigned, and candidate items with support values ​​less than k are removed. Figure 3 The first step removes feature three, forming a new table. The features remaining in the candidate sets in the new table are then paired. The initialized samples are scanned based on the newly formed feature combinations. When all feature combinations for the same user are 1, the frequency of occurrence of that feature combination is incremented by one, and the data is recorded to form the second-level candidate set. Feature groups with less than the minimum support k are removed, forming a new table. Feature combinations with the same items in the new table are then paired, and the scan is performed again according to the above rules to obtain the third-level candidate set.

[0031] The hypergraph is constructed by setting the characteristics of each set in the obtained clustering results to different vertices in the same hyperedge, resulting in a weightless hypergraph G; then, in hypergraph G, for each hyperedge e i Assign a positive weight w(e) i Let w(e) i The value of w(e) is numerically equal to the frequency of occurrence of the characteristic of the candidate option. i Let V be the weight of the hyperedges, and we obtain the weighted hypergraph G(V,E,W), where V represents the set of vertices, E represents the set of hyperedges, and W represents the set of positive weights. For example... Figure 4 As shown, Figure 4 Representative according to Figure 3 The hypergraph generated from the clustering results. (Through...) Figure 3 Multiple data tables can be obtained. Select one suitable candidate set that meets the minimum support and create a hypergraph from it. In this example, the second-level candidate set that meets the minimum support k is selected. Each feature combination is made into a different node in the same hyperedge. The hyperedge is assigned a value that is numerically equal to the number of times the feature of the combination occurs frequently.

[0032] Step 4: Solve for the normalized weights of the hyperedges in the hypergraph; the normalized weights are solved as follows:

[0033] For a weighted hypergraph G(V,E,W), when vertex v i ∈e i It is called superedge e i With vertex v i Related; for any v i ∈V, vertex v i degree d(v) i Let ) denote the number of hyperedges associated with a vertex; then for any hyperedge e i ∈E, its degree is defined as:

[0034] δ(e i )=|e i |;

[0035] The vertex degree matrix is ​​given by D e Represent the sum of the vertex degrees of all vertices, denoted as D. k The diagonal matrix of the hyperedge is W H To represent;

[0036] In a hyperedge, every node has δ(e)-1 edges connecting it to another hyperedge node. Therefore, the normalized matrix D can be obtained through the formula. f :

[0037]

[0038] From the normalized matrix Df The normalized weights of the superedges are obtained.

[0039] Figure 5 This represents the matrix obtained from the hypergraph and the normalized matrix derived from each matrix. Vertex degree matrix D e The hyperedge diagonal matrix W represents the number of times each characteristic appears in different hyperedges. H The normalized matrix D represents the assignment of each hyperedge itself. f This is achieved through the hyperdiagonal matrix W. H Divide by the vertex degree matrix D e We obtain the sum of the diagonals.

[0040] Step 5: From the weighted hypergraph G(V,E,W), we can obtain the strong correlation between features and the normalized weights of different feature categories. Obtaining the strong correlation between features is helpful in formulating a reasonable network slicing resource allocation scheme, while the obtained normalized weights of the hyperedges can avoid the problem of difficulty in determining which feature to prioritize due to the different attributes of the features themselves. Thus, we can allocate network slicing schemes to users fairly and reasonably, and ultimately provide users with the most comfortable network experience.

[0041] In summary, this invention provides a network slice resource allocation method by applying hypergraph clustering, thereby achieving a reasonable, efficient, and fair allocation. From the network operator's perspective, the overall system utility, network operating cost benefits, and resource utilization rate achieve ideal results. From the user's perspective, on the one hand, user needs are met and service quality is improved; on the other hand, personalized services are realized, providing customized and fair services based on the differentiated needs of different users.

Claims

1. A method for allocating network slice resources, characterized in that: Includes the following steps: Step 1: Obtain multiple user data sets, set a corresponding threshold for each characteristic in the user data, and determine whether the characteristic value in the user data is greater than the corresponding threshold. If it is greater, assign the value 1 to the characteristic; if it is less than the threshold, assign the value 0. This completes the user data initialization. Step 2: Define the frequency of a feature as the total number of times the feature value is 1 in a single user's data. Scan all initialized user data to obtain the frequency of each feature. Step 3: Cluster based on the frequency of occurrence of features, and create a hypergraph based on the clustering results; Step 4: Solve for the normalized weights of the hyperedges in the hypergraph; Step 5: Determine the strong correlation between features based on the obtained hypergraph and allocate network slice resources; In step 3, the clustering process involves setting a minimum support k based on actual needs, using all features in the user data as candidate sets to be scanned, and after scanning all features, retaining the candidate sets whose frequency of occurrence of the same feature meets the minimum support to obtain the first-level candidate sets. Based on the first-level candidate sets, candidate sets are combined in pairs to form new sets to be scanned to form second-level candidate sets. The user data is then scanned again, and the frequency of occurrence of the second-level candidate sets is counted. Candidates with a support less than the minimum support k are removed. Then, based on the secondary candidate set, three characteristics are selected to form a new set to be scanned, forming a tertiary candidate set. The user data is scanned again, the number of times the tertiary candidate set appears is counted, and the clustering process is repeated until the end. In step 3, the hypergraph is created by setting the characteristics of each set in the obtained clustering results as different vertices in the same hyperedge, thus obtaining a weightless hypergraph. Then in the supergraph In the middle, for each hyperedge Assign a positive weight ,make The value that is numerically equal to the frequency of occurrence of the characteristic of the candidate is called... We obtain the weighted hypergraph by assigning weights to the hyperedges. ,in Represents the set of vertices. Denotes the set of superedges. Represents the set of positive weights; In step 4, the normalized weights are calculated as follows: For weighted hypergraphs When the vertex It is called a superedge With vertex Related; for any ,vertex degree Let represent the number of hyperedges associated with a vertex; then for any hyperedge... Its degree is defined as: ； The vertex degree matrix is ​​composed of Represent the sum of the degree of all vertices, and record it as . The diagonal matrix of the hyperedge is composed of To represent; In a hyperedge, any node has An edge connecting to another superedge node can be used to obtain the normalized matrix using the formula. : From the normalized matrix The normalized weights of the superedges are obtained.

Images