CN115209431B - A triggering method, device, equipment and computer storage medium - Google Patents

Abstract

The invention discloses a triggering method, which comprises the following steps: and acquiring a slice instance of the network slice, inputting the slice instance into a trained Node2Vec model, outputting a vector of a Node in the acquired slice instance, and determining whether to trigger reconstruction of the slice instance according to the vector of the Node. The embodiment of the invention also discloses a triggering device, equipment and a computer storage medium, which improve the decision efficiency of whether the network slice needs to be reconstructed or not, and further improve the reconstruction efficiency of the network slice.

Description

A triggering method, device, equipment and computer storage medium

Technical field

The present invention relates to network slicing reconstruction triggering technology, and in particular, to a triggering method, device, equipment and computer storage medium.

Background technique

Currently, network slicing is an important part of 5G construction. This technology meets the differentiated needs of vertical industries for network services by building multiple logically independent proprietary networks on the same physical basic platform. Different from the traditional single network management model, network slicing technology provides a greater choice for customization of individual needs, and provides operators with more convenient, efficient, secure and low-cost operation and maintenance for carrying diverse network services. plan.

The 3rd Generation Partnership Project (3GPP) stipulates four necessary stages in the slice life cycle, namely preparation stage, instantiation, configuration, activation stage, runtime stage and offline stage. Among them, in the preparation stage, it is necessary to clarify the requirements of all types of services that the bearer network will carry, and customize different slicing templates according to different types of requirements. In the instantiation, configuration, and activation phases, business requirements need to be converted into network performance requirements, the corresponding slice templates are selected, the virtual topology is associated with the physical bearer topology, the corresponding network resources are configured, and the slices are instantiated based on the template. In the runtime phase, the business needs to be carried on the instantiated slice. During the offline phase, it is necessary to delete the instantiated slices that have been served and recycle related underlying resources.

Among them, during the runtime phase, intelligent network operation and maintenance methods can be used to monitor the slicing status and reconstruct the instantiated slices according to changes in business requirements to realize the adaptation of the slicing structure and resources to changes in demand, thus increasing the number of slicing flexibility and improve network service quality.

In related technologies, network slices are usually identified and calculated based on a matrix to trigger the reconstruction of network slices. However, this method requires a large amount of calculation, making the triggering method highly complex and causing the network slice to restructure. The reconstruction triggering efficiency of network slicing is poor; it can be seen that the existing reconstruction triggering method of network slicing has technical problems of low efficiency.

Contents of the invention

In view of this, the present invention provides a triggering method, device, equipment and computer storage medium to solve the technical problem of low efficiency in the network slice reconstruction triggering method existing in the prior art.

The technical solution of the present invention is implemented as follows:

In a first aspect, an embodiment of the present invention provides a triggering method, which includes:

Get the slice instance of the network slice;

Input the slice instance into the trained Node2Vec model, and output the vector of the nodes in the slice instance; wherein the vector elements of the node include: an amount used to represent the node structure and an amount used to represent the node load ;

Based on the node's vector, it is determined whether to trigger reconstruction of the slice instance.

In the above method, the trained Node2Vec model is obtained in the following way:

Obtain the starting node and the next hop node of the starting node from the slice instance to be trained, determine the next hop node as the current node, and set the initial value of the sampling number i to 1;

Get the adjacent nodes of the current node;

When i is smaller than the sampling step, the bias walk probability of the current node and the adjacent node is calculated according to the preset bias walk probability algorithm; wherein the preset bias walk probability is the same as the current bias walk probability. The network performance parameters of the current business when slicing instances are used at all times are positively correlated;

Sort the calculated bias wander probabilities from large to small, select the top M adjacent nodes, determine the selected adjacent nodes as the current node, update i to i+1, and return to execution Obtaining the adjacent nodes of the current node; wherein, M is a positive integer smaller than the number of adjacent nodes;

When i is greater than or equal to the sampling step, at least two sets of node sequences obtained by sampling are input into the preset Node2Vec model for training, and the trained Node2Vec model is obtained.

In the above method, the network performance parameters of the current service include one or more of the following:

The average occupied bandwidth of the current business, the average message delay of the current business, and the average packet loss rate of the current business. .

In the above method, the following formula is used to calculate the bias wander probability π _vx from the current node to the adjacent node:

π _vx =α _pq (t,x)·w _vx

Among them, v represents the current node, t represents the adjacent node of the current node, W ₀ represents the average occupied bandwidth of the current business at the current moment, T ₀ represents the average message delay of the current business at the current moment, and E ₀ represents the average message delay of the current business at the current moment. Packet loss rate, k _W is the estimated level of W ₀ , k _T is the estimated level of T ₀ , k _E is the estimated level of E ₀ , d _tx represents the shortest number of hops required for node t to jump to node x.

In the above method, determining whether to trigger reconstruction of the slice instance according to the vector of the node includes:

Calculate the Euclidean distance between any two nodes in the slice instance according to the vector of the node;

When the Euclidean distance between any two nodes meets the preset condition, the reconstruction of the slice instance is triggered.

In the above method, when the Euclidean distance between any two nodes meets the preset conditions, the reconstruction of the slice instance is triggered, including:

When the Euclidean distance between any two nodes is smaller than the preset minimum Euclidean distance between nodes, reconstruction of the slice instance is triggered.

In the above method, when the Euclidean distance between any two nodes meets the preset conditions, the reconstruction of the slice instance is triggered, including:

When the Euclidean distance between any two nodes is greater than the preset maximum Euclidean distance between nodes, reconstruction of the slice instance is triggered.

In the above method, the method further includes:

When there is no Euclidean distance between any two nodes that is less than the preset minimum Euclidean distance between nodes, and there is no Euclidean distance between any two nodes that is greater than the preset maximum Euclidean distance between nodes. distance, prohibit triggering reconstruction of the slice instance.

In a second aspect, the present invention provides a triggering device, which includes:

Obtain module, used to obtain slice instances of network slices;

A processing module, configured to input the slice instance into the trained Node2Vec model, and output the vector of the nodes in the slice instance; wherein the vector elements of the node include: a quantity used to represent the node structure and a quantity used to represent the node structure. Indicates the amount of node load;

A triggering module, configured to determine whether to trigger reconstruction of the slice instance according to the vector of the node.

In the above device, the device is also used for:

Use the following method to obtain the trained Node2Vec model:

Obtain the starting node and the next hop node of the starting node from the slice instance to be trained, determine the next hop node as the current node, and set the initial value of the sampling number i to 1;

Get the adjacent nodes of the current node;

When i is smaller than the sampling step, the bias walk probability of the current node and the adjacent node is calculated according to the preset bias walk probability algorithm; wherein the preset bias walk probability is the same as the current bias walk probability. The network performance parameters of the current business when slicing instances are used at all times are positively correlated;

Sort the calculated bias wander probabilities from large to small, select the top M adjacent nodes, determine the selected adjacent nodes as the current node, update i to i+1, and return to execution Obtaining the adjacent nodes of the current node; wherein, M is a positive integer smaller than the number of adjacent nodes;

When i is greater than or equal to the sampling step, at least two sets of node sequences obtained by sampling are input into the preset Node2Vec model for training, and the trained Node2Vec model is obtained.

In the above device, the network performance parameters of the current service include one or more of the following:

The average occupied bandwidth of the current business, the average message delay of the current business, and the average packet loss rate of the current business.

In the above device, the device is also used to calculate the bias wander probability π _vx from the current node to the adjacent node using the following formula:

π _vx =α _pq (t,x)·w _vx

Among them, v represents the current node, t represents the adjacent node of the current node, W ₀ represents the average occupied bandwidth of the current business at the current moment, T ₀ represents the average message delay of the current business at the current moment, and E ₀ represents the average message delay of the current business at the current moment. Packet loss rate, k _W is the estimated level of W ₀ , k _T is the estimated level of T ₀ , k _E is the estimated level of E ₀ , d _tx represents the shortest number of hops required for node t to jump to node x.

In the above device, the trigger module is specifically used for:

Calculate the Euclidean distance between any two nodes in the slice instance according to the vector of the node;

When the Euclidean distance between any two nodes meets the preset condition, the reconstruction of the slice instance is triggered.

In the above device, when the Euclidean distance between any two nodes meets the preset condition, the triggering module triggers the reconstruction of the slice instance, including:

When the Euclidean distance between any two nodes is smaller than the preset minimum Euclidean distance between nodes, reconstruction of the slice instance is triggered.

In the above device, when the Euclidean distance between any two nodes meets the preset condition, the triggering module triggers the reconstruction of the slice instance, including:

When the Euclidean distance between any two nodes is greater than the preset maximum Euclidean distance between nodes, reconstruction of the slice instance is triggered.

In the above device, the device is also used for:

When there is no Euclidean distance between any two nodes that is less than the preset minimum Euclidean distance between nodes, and there is no Euclidean distance between any two nodes that is greater than the preset maximum Euclidean distance between nodes. distance, prohibit triggering reconstruction of the slice instance.

In a third aspect, embodiments of the present invention further provide a device, which device includes: a processor and a storage medium storing instructions executable by the processor, and the storage medium relies on the processor to perform operations through a communication bus. , when the instruction is executed by the processor, the triggering method described in one or more of the above embodiments is executed.

Embodiments of the present invention provide a computer storage medium that stores executable instructions. When the executable instructions are executed by one or more processors, the processor executes the trigger described in one or more embodiments. method.

The invention provides a triggering method, device, equipment and computer storage medium. The method includes: obtaining a slice instance of a network slice, inputting the slice instance into a trained Node2Vec model, and outputting a vector of nodes in the slice instance, Among them, the vector element of the node includes: a quantity used to represent the node structure and a quantity used to represent the node load. According to the vector of the node, it is determined to trigger the reconstruction of the slice instance; that is to say, in the present invention, after obtaining After slicing the slicing instance of the network slicing, input the slicing instance into the trained Node2Vec model to get the vector of the node of the slicing instance, because the obtained vector contains quantities used to represent the node structure and node load, and based on the node The vector can reflect the structure of the slice and the load of the slice. Then, in the reconstruction of the slice instance to determine whether to trigger the slice instance based on the vector of the node, since the vector of the node has two dimensions, using a vector of two dimensions reduces the time required to determine whether to trigger the slice instance. The complexity of reconstruction of slice instances improves the decision-making efficiency of whether network slices need to be reconstructed, thereby improving the reconstruction efficiency of network slices.

Description of the drawings

Figure 1 is a schematic flow chart of an optional triggering method in an embodiment of the present invention;

Figure 2 is a network topology diagram of a slice example marked with bias weights in related technologies;

Figure 3 is a schematic flow chart of an example of an optional triggering method provided by an embodiment of the present invention;

Figure 4 is a schematic structural diagram of an optional triggering device in an embodiment of the present invention;

Figure 5 is a schematic structural diagram of an optional device provided by an embodiment of the present invention.

Detailed ways

The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention.

Embodiment 1

An embodiment of the present invention provides a triggering method. Figure 1 is a schematic flowchart of an optional triggering in an embodiment of the present invention. As shown in Figure 1, the triggering method may include:

S101: Obtain the slice instance of the network slice;

Currently, for network slicing, after the slicing service starts, first select a slicing template that matches the current business requirements from the slicing template library, then create a slicing instance based on the selected slicing template, and then launch the slicing instance to carry the business requirements. The slicing instance After going online, the slice instance is represented based on the matrix to determine whether to trigger the reconstruction of the slice instance; and then the matrix is used to determine whether to trigger the reconstruction of the slice instance. Since the dimension of the matrix is high, the matrix is used to represent the slice information. This increases the computational complexity in determining whether to trigger reconstruction of the slice instance, thereby increasing the cost of monitoring the slice instance.

In order to reduce the complexity of determining whether to trigger the reconstruction of the slice instance and reduce the cost of monitoring the slice instance, embodiments of the present invention provide a triggering method, which is used to determine whether to trigger the reconstruction method of the slice instance. First, obtain the network A slice instance of the slice.

Here, it should be noted that the triggering method provided by the embodiment of the present invention is deployed after the slicing instance goes online, that is, the above-mentioned runtime stage. That is to say, after the slicing service starts, first select the current business from the slicing template library. The requirements match the slicing template, and then create a slicing instance based on the selected slicing template, and then go online to carry the business requirements. After the slicing instance is online, obtain the slicing instance.

The current business of the above-mentioned slicing instance may be a shopping application, a mobile call service, or a communication application. Here, the embodiment of the present invention does not specifically limit this.

S102: Input the slice instance into the trained Node2Vec model, and output the vector of the nodes in the slice instance;

In obtaining the slice instance, the topology map of the slice instance is obtained. The topology map of the slice instance includes the nodes in the slice instance and the connection relationships between nodes. Then, input the slice instance into the trained In the Node2Vec model, the vector of nodes in the slice instance can be output.

The above-trained Node2Vec model is obtained by sampling sample data, and then using the sample data to train the Node2Vec model. Among them, the Node2Vec model is a model used to generate node vectors in the network. The input of the model is the network structure, and the output is each node vector. vector of nodes.

Here, by introducing the processing of slicing problems into the Node2Vec model, the slicing information is reduced from N dimensions to 2 dimensions, which effectively reduces the space complexity of slicing information representation and storage, reduces the management and monitoring costs of slicing, and improves network operation and maintenance. benefit.

Among them, the vector elements of the node include: the amount used to represent the node structure and the amount used to represent the node load; that is to say, the obtained vector of the node can reflect the structure of the node and the load of the node. In this way, according to the node's Vector is used to determine whether to trigger the reconstruction of the slice instance, mainly based on the structure of the node and the load of the node to determine whether to trigger the reconstruction of the slice instance. The node vector obtained in this way is conducive to ensuring a low amount of calculation. Under certain conditions, combining the structure of the node itself and the current load situation can accurately determine whether to trigger the reconstruction of the slice instance, which increases the accuracy of decision-making.

Further, in order to obtain the trained Node2Vec model, in an optional embodiment, the following method is used to obtain the trained Node2Vec model:

Obtain the starting node and the next hop node of the starting node from the slice instance to be trained, determine the next hop node as the current node, and set the initial value of the sampling number i to 1;

Get the adjacent nodes of the current node;

When i is smaller than the sampling step, the bias walk probability of the current node and adjacent nodes is calculated according to the preset bias walk probability algorithm;

Sort the calculated bias wander probabilities from large to small, select the top M adjacent nodes, determine the selected adjacent nodes as the current node, update i to i+1, and return to execution Get the adjacent nodes of the current node;

When i is greater than or equal to the sampling step, input at least two sets of sampled node sequences into the preset Node2Vec model for training to obtain a trained Node2Vec model.

Specifically, slice instances to be trained are obtained, where the number of slice instances to be trained may be one or more. Here, the embodiment of the present invention does not specifically limit this.

In order to obtain sample data for model training, the following method is used for sampling: for the slice instance to be trained, a starting node is first randomly determined, and then the next node of the starting node is randomly determined from the adjacent nodes of the starting node. One hop node, and the next hop node is determined as the current node. Since the sampling step is preset, in order to sample according to the sampling step, the initial value of the sampling number i is set to 1.

Then obtain the adjacent nodes of the current node, and determine the relationship between i and the sampling step. When i is less than the sampling step, calculate the offset walk between the current node and the adjacent node according to the preset offset walk probability algorithm. Take probability.

Figure 2 is a network topology diagram of a slice example marked with bias weight in related technologies. As shown in Figure 2, t represents the starting node, v represents the next hop node of the starting node, t, x1, x2 and x3 represent Adjacent nodes of v node, in related technology, the walking method in the Node2Vec model uses the following formula to calculate the bias walking probability α _pq (t,x):

Among them, combined with Figure 2, parameters p and q are used to control the tendency of the sampling process to Breadth First Search (BFS, Breadth First Search) and Depth First Search (DFS, Depth First Search) respectively. p is the return parameter, the larger p, The smaller the probability of getting the same node during the sampling process. q is the entry and exit parameter. If q>1, the sampling process will tend to BFS. If q<1, the sampling process will tend to DFS. d _tx represents the need for node t to jump to node v. The shortest number of jumps.

In the embodiment of the present invention, in order to trigger the reconstruction of the slice instance, a preset bias walking probability is provided here, where the preset bias walking probability is the same as when the slice instance is used at the current moment. is positively correlated with the network performance parameters of the current service of the slice instance. That is to say, the value of the bias wandering probability increases with the increase of the network performance parameters of the current service of the slice instance. As the network performance parameters of the current service of the slice instance increase, It decreases with the decrease of performance parameters. In this way, the offset wander probability is related to the network performance parameters of the current service.

Then sort the calculated bias wander probabilities from large to small, and select the top M adjacent nodes as the current node, where M is a positive integer smaller than the number of adjacent nodes; thus, The sampled training samples are node sequences with better network performance parameters of the current business, which improves the quality of the training samples, so that a Node2Vec model that better meets the requirements can be trained.

After updating the current node by selecting adjacent nodes, i is updated to i+1, the next sampling is performed, and execution returns to obtain the adjacent nodes of the current node.

In addition, when i is greater than or equal to the sampling step, it means that the sampling has been completed. Then, at least two sets of node sequences jumped during the sampling are used as training samples and input into the preset Node2Vec model for training to obtain the trained Node2Vec model.

In the embodiment of the present invention, the bias parameter w _vx is introduced in the sampling stage of the trained Node2Vec model, so that the sampled node sequence samples can train a mapping model that integrates the slice structure and load status, completing the mapping of the slice structure and load information. Fusion mapping can realize multi-factor-oriented reconstruction decision-making.

In order to more accurately determine whether to trigger the reconstruction of the slice instance, in an optional embodiment, the network performance parameters of the current service include one or more of the following:

The average occupied bandwidth of the current business, the average message delay of the current business, and the average packet loss rate of the current business.

Specifically, the network performance parameters of the current service may include one or more of the average occupied bandwidth of the current service, the average message delay of the current service, and the average packet loss rate of the current service. Here, this embodiment of the present invention No specific limitation is made.

Further, in order to determine more accurately whether to trigger the reconstruction of the slice instance, the network performance parameters of the current service may include the average occupied bandwidth of the current service, the average message delay of the current service and the average packet loss rate of the current service. In an optional embodiment, the following formula is used to calculate the offset walking probability π _vx from the current node to the adjacent node:

π _vx =α _pq (t,x)·w _vx (2)

Among them, v represents the current node, t represents the adjacent node of the current node, W ₀ represents the average occupied bandwidth of the current business at the current moment, T ₀ represents the average message delay of the current business at the current moment, and E ₀ represents the average message delay of the current business at the current moment. Packet loss rate, k _W is the estimated level of W ₀ , k _T is the estimated level of T ₀ , and k _E is the estimated level of E ₀ .

Among them, in practical applications, the sum of the above k _W , k _T and k _E is 1, and the importance of occupied bandwidth, message delay and packet loss rate in sampling decisions can be adjusted by adjusting these three estimation level parameters. degree of sex.

Here, based on the original bias walk probability using the above formula (1), formula (3) is introduced, that is, multiplying the original bias walk probability by w _vx so that the preset bias walk probability is used. The π _vx obtained by the walking probability algorithm is integrated into the structure of the node and the load condition of the node, so that the obtained bias walking probability takes into account the network performance of the current business, so that the selected adjacent nodes are network nodes with better network performance. , which is conducive to improving the quality of training samples, which is conducive to training a more accurate Node2Vec model.

S103: Determine whether to trigger reconstruction of the slice instance according to the vector of the node.

After determining the node vector of the slice instance, in order to determine whether to trigger the reconstruction of the slice instance, it may be determined whether to trigger the reconstruction of the slice instance according to the vector of the node. For example, one of the vectors of the node may be used. component to determine whether to trigger the reconstruction of the slice instance, or another component in the node vector to determine whether to trigger the reconstruction of the slice instance, or two components of the node vector to determine whether Triggering the reconstruction of the slice instance, here, the embodiment of the present invention does not specifically limit this.

In order to determine whether to trigger reconstruction of the slice instance, in an optional embodiment, S102 may include:

Calculate the Euclidean distance between any two nodes in the slice instance based on the node's vector;

When the Euclidean distance between any two nodes meets the preset conditions, the reconstruction of the slice instance is triggered.

Specifically, according to the vector of the node, the distance formula between two points is used to calculate the Euclidean distance between any two nodes in the slice instance, and then it is judged whether the Euclidean distance between any two nodes meets the preset conditions. , only when the preset conditions are met, the reconstruction of the slice instance is triggered. When the preset conditions are not met, the reconstruction of the slice instance is prohibited from being triggered.

Further, in order to trigger the reconstruction of the slice instance, in an optional embodiment, when the Euclidean distance between any two nodes meets the preset conditions, the reconstruction of the slice instance is triggered, including:

When the Euclidean distance between any two nodes is less than the preset minimum Euclidean distance between nodes, the reconstruction of the slice instance is triggered.

Specifically, a preset minimum Euclidean distance between nodes is preset in this triggering method. The minimum Euclidean distance represents the minimum acceptable vector distance in the vector space. Its practical significance is that between the two corresponding nodes in the current slice instance Then, when the Euclidean distance between any two nodes is less than the preset minimum Euclidean distance between nodes, it means that the load between the two nodes in the slice instance is less than The current slice instance corresponds to the upper limit of load that can be carried between two nodes. It can be seen that the current slice instance cannot guarantee the stable operation of the current business, so the reconstruction of the slice instance is triggered.

Further, in order to trigger the reconstruction of the slice instance, in an optional embodiment, when the Euclidean distance between any two nodes meets the preset conditions, the reconstruction of the slice instance is triggered, including:

When the Euclidean distance between any two nodes is greater than the preset maximum Euclidean distance between nodes, the reconstruction of the slice instance is triggered.

Specifically, a preset maximum Euclidean distance between nodes is preset in this triggering method. The maximum Euclidean distance represents the maximum acceptable vector distance in the vector space. Its practical significance is that one of the two corresponding nodes in the current slice instance The lower limit of the acceptable resource occupancy rate between any two nodes. Then, when the Euclidean distance between any two nodes is greater than the preset maximum Euclidean distance between nodes, it means that there are resources between the two nodes in the slice instance. The occupancy rate is greater than the lower limit of the acceptable resource occupancy rate between the corresponding two nodes in the current slicing instance. It can be seen that the current slicing instance cannot guarantee the stable operation of the current business, so the reconstruction of the slicing instance is triggered.

In order to achieve accurate triggering of reconstruction of slice instances, in an optional embodiment, the above method further includes:

When there is no Euclidean distance between any two nodes that is less than the preset minimum Euclidean distance between nodes, and there is no Euclidean distance between any two nodes that is greater than the preset maximum Euclidean distance between nodes. distance, prohibit triggering reconstruction of slice instances.

When there is no Euclidean distance between any two nodes that is smaller than the preset minimum Euclidean distance between nodes, it means that there is no load between the two nodes in the slice instance that is lower than the corresponding two nodes in the current slice instance. The upper limit of the load that can be carried between them. When the Euclidean distance between any two nodes does not exist greater than the preset maximum Euclidean distance between nodes, it means that there is no resource occupation between the two nodes in the slice instance. The rate is greater than the lower limit of the acceptable resource occupancy rate between the corresponding two nodes in the current slice instance. It can be seen that the current slice instance can ensure the stable operation of the current business, so it is prohibited to trigger the reconstruction of the slice instance.

The following is an example to describe the triggering method described in one or more of the above embodiments.

Figure 3 is a schematic flow chart of an example of an optional triggering method provided by an embodiment of the present invention. As shown in Figure 3, the slice reconstruction triggering method of this example is deployed after the slice instance goes online and before the slice instance goes offline. The evolution from static slicing to elastic slicing under the standard slicing life cycle is realized; the triggering method can include:

S301: After the slicing service is started, first select a slicing template from the slicing template library; where the selected slicing template matches the current business requirements;

S302: Create a slice instance according to the selected slice template;

S303: The slicing instance goes online and carries business requirements.

Specifically, after the slice instance goes online, the slice reconstruction triggering strategy proposed by the present invention is started.

S304: Determine whether the slicing service has ended; if yes, execute S305; if no, execute S306;

S305: The slice instance goes offline; execute S311;

S306: Use S strategy sampling, that is, use the slice instances to be trained to obtain training samples;

S307: Use the training samples to train the Node2Vec model and obtain the trained Node2Vec model;

S308: Input the slice instance into the trained Node2Vec model, map the slice instance to the vector space, and obtain the vector of the node in the slice instance;

Specifically, the slice reconstruction triggering strategy first samples the current slice based on the S sampling strategy (equivalent to the sampling method described in one or more embodiments above), then uses the collected samples to train the Node2Vec model, and then inputs the slice into the training A good model obtains the mapping results of slice structure and load status in low-dimensional vector space.

It should be pointed out that the training data obtained by the S strategy is used to train the Node2Vec model. This model can map slices and slice loads from high-dimensional information space to two-dimensional vector space. Each virtual node has its corresponding two-dimensional vector representation. . Since the virtual link load status information is integrated into the biased wandering strategy of Node2Vec sampling, the Euclidean distance between vectors can not only represent the connection similarity of the topology, but also the load status between two virtual nodes.

S309: Calculate the Euclidean distance between nodes and determine whether there is a distance exceeding the threshold in the Euclidean distance; if there is a distance that exceeds the threshold, execute S310; if not, execute S304;

S310: Reconstruct the slice, which triggers the reconstruction of the slice instance;

S311: Recover resources and end the service.

Specifically, after the mapping is completed, a threshold check is started to see if there are any vector pairs that exceed the threshold range. If it exists, reconstruct the current slice, replace the currently running slice with the reconstructed slice, and continue to monitor the reconstruction trigger strategy for the reconstructed slice; if it does not exist, continue to execute the reconstruction trigger for the current slice. Monitoring of strategies. If the slicing service ends, the slicing instance will be taken offline, all resources allocated to the slicing will be recovered, and the service will end.

For the threshold check, specifically, if there is a virtual link connection between two nodes or the business load between them is heavy, it may cause the Euclidean distance of their corresponding vector pairs in the vector space to become closer. Based on this changing relationship , the slice status monitoring algorithm of this example will set up two thresholds. The threshold T _min represents the minimum acceptable vector distance in the vector space. Its practical significance is the upper limit of the load that can be carried between the corresponding two nodes in the current slice instance. When When a Euclidean distance smaller than T _min is detected in the vector space, slice reconstruction will be triggered. The threshold T _max represents the maximum acceptable vector distance in the vector space. Its practical significance is the lower limit of acceptable resource usage between the corresponding two nodes in the slice instance. When a Euclidean distance greater than T _max is detected in the vector space, , slice reconstruction will be triggered. The difference between triggering slice reconstruction in the two cases is that the former will allocate more bearing resources to the new slice instance than the existing slice instance during reconstruction, while the latter will allocate less bearing resources than the existing slice instance.

In other words, this example proposes a slice status monitoring method based on Node2Vec. By mapping the slice structure and slice load from a high-dimensional information space to a two-dimensional vector space, it reduces the space complexity required to store the slice structure and slice status. degree, which reduces the space cost of monitoring slice structure and slice status, improves the efficiency of managing slice structure and slice status, and adopts slice reconstruction triggering decision-making based on two-dimensional vector space. In the two-dimensional vector image output by the Node2Vec algorithm, the vector There is a one-to-one correspondence with the nodes in the slice topology. The vector relationship reflects the topology structure and load status between the nodes in the slice Chile. The relationship operation results between vectors are used as the judgment criteria for reconstruction triggering to realize the slicing structure and load. Multi-dimensional information fusion decision-making improves the accuracy of decision-making. The sampling strategy S is adopted. By introducing the bias parameter w _vx in the Node2Vec model sampling stage, the sampled node sequence samples can train a mapping model that fuses the slice structure and load status. Provides an implementation solution for the application of Node2Vec model in the field of slice representation and monitoring.

The invention provides a triggering method, which method includes: obtaining a slice instance of a network slice, inputting the slice instance into a trained Node2Vec model, and outputting a vector of nodes in the slice instance, where the vector elements of the node include: The amount used to represent the node structure and the amount used to represent the node load are determined to trigger the reconstruction of the slice instance according to the vector of the node; that is to say, in the present invention, after obtaining the slice instance of the network slice, the When the slice instance is input into the trained Node2Vec model, the vector of the node of the slice instance can be obtained, because the obtained vector contains quantities used to represent the node structure and node load, and the node-based vector can reflect the structure and structure of the slice. The load condition of the slice, then, determine whether to trigger the reconstruction of the slice instance based on the vector of the node. Since the vector of the node has two dimensions, using the vector of two dimensions reduces the complexity of determining whether to trigger the reconstruction of the slice instance. , thereby improving the decision-making efficiency of whether network slices need to be reconstructed, and thus improving the reconstruction efficiency of network slices for reconstruction.

Embodiment 2

Based on the same inventive concept, an embodiment of the present invention provides a trigger device. Figure 4 is a schematic structural diagram of an optional trigger device in the embodiment of the present invention. As shown in Figure 4, the trigger device includes: an acquisition module 41, processing module 42 and trigger module 43;

Among them, the acquisition module 41 is used to obtain the slice instance of the network slice;

The processing module 42 is used to input the slice instance into the trained Node2Vec model, and output the vector of the node in the slice instance; wherein the vector elements of the node include: an amount used to represent the node structure and an amount used to represent the node load. ;

The trigger module 43 is used to determine whether to trigger the reconstruction of the slice instance according to the vector of the node.

In an optional embodiment, the above device is also used for:

Use the following method to get the trained Node2Vec model:

Obtain the starting node and the next hop node of the starting node from the slice instance to be trained, determine the next hop node as the current node, and set the initial value of the sampling number i to 1;

Get the adjacent nodes of the current node;

When i is less than the sampling step size, the bias walk probability of the current node and the adjacent node is calculated according to the preset bias walk probability algorithm; where the preset offset walk probability is the same as the current moment using the slice instance are positively correlated with the network performance parameters of the current business;

Sort the calculated bias wander probabilities from large to small, select the top M adjacent nodes, determine the selected adjacent nodes as the current node, update i to i+1, and return to execution Get the adjacent nodes of the current node; where M is a positive integer smaller than the number of adjacent nodes;

When i is greater than or equal to the sampling step, input at least two sets of sampled node sequences into the preset Node2Vec model for training to obtain a trained Node2Vec model.

In an optional embodiment, the network performance parameters of the current service include one or more of the following:

The average occupied bandwidth of the current business, the average message delay of the current business, and the average packet loss rate of the current business.

In an optional embodiment, the above device is also used for:

The bias walking probability π _vx from the current node to the adjacent node is calculated using the above formula (1)-formula (3).

In an optional embodiment, the above-mentioned trigger module 43 is specifically used to:

Calculate the Euclidean distance between any two nodes in the slice instance based on the node's vector;

When the Euclidean distance between any two nodes meets the preset conditions, the reconstruction of the slice instance is triggered.

In an optional embodiment, when the Euclidean distance between any two nodes meets the preset condition, the above-mentioned triggering module 43 triggers the reconstruction of the slice instance, including:

When the Euclidean distance between any two nodes is less than the preset minimum Euclidean distance between nodes, the reconstruction of the slice instance is triggered.

In an optional embodiment, when the Euclidean distance between any two nodes meets the preset condition, the above-mentioned triggering module 43 triggers the reconstruction of the slice instance, including:

When the Euclidean distance between any two nodes is greater than the preset maximum Euclidean distance between nodes, the reconstruction of the slice instance is triggered.

In an optional embodiment, the above device is also used for:

When there is no Euclidean distance between any two nodes that is less than the preset minimum Euclidean distance between nodes, and there is no Euclidean distance between any two nodes that is greater than the preset maximum Euclidean distance between nodes. distance, prohibit triggering reconstruction of slice instances.

In practical applications, the above-mentioned acquisition module 41, processing module 42 and trigger module 43 can be implemented by a processor located on the device, specifically a central processing unit (CPU, Central Processing Unit), a microprocessor (MPU, Microprocessor Unit), a digital Signal processor (DSP, Digital Signal Processing) or Field Programmable Gate Array (FPGA, Field Programmable Gate Array) and other implementations.

Figure 5 is a schematic structural diagram of an optional device provided by an embodiment of the present invention. As shown in Figure 5, an embodiment of the present invention provides a device 500, which includes:

The processor 51 and the storage medium 52 storing instructions executable by the processor 51. The storage medium 52 relies on the processor 51 to perform operations through the communication bus 53. When the instructions are executed by the processor 51, Execute the triggering method described in the first embodiment above.

It should be noted that in actual application, various components in the terminal are coupled together through the communication bus 53 . It can be understood that the communication bus 53 is used to implement connection communication between these components. In addition to the data bus, the communication bus 53 also includes a power bus, a control bus and a status signal bus. However, for the sake of clarity, the various buses are labeled as communication bus 53 in FIG. 5 .

An embodiment of the present invention provides a computer storage medium that stores executable instructions. When the executable instructions are executed by one or more processors, the processor executes the triggering method described in Embodiment 1.

Among them, the computer-readable storage medium can be magnetic random access memory (ferromagnetic random access memory, FRAM), read only memory (Read Only Memory, ROM), programmable read-only memory (Programmable Read-Only Memory, PROM), erasable memory In addition to programmable read-only memory (ErasableProgrammable Read-Only Memory, EPROM), electrically erasable programmable read-only memory (Electrically Erasable Programmable Read-Only Memory, EEPROM), flash memory (Flash Memory), magnetic surface memory, optical disks, Or memory such as Compact Disc Read-Only Memory (CD-ROM).

Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Accordingly, the invention may take the form of a hardware embodiment, a software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the invention may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, magnetic disk storage, optical storage, etc.) embodying computer-usable program code therein.

The invention is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each process and/or block in the flowchart illustrations and/or block diagrams, and combinations of processes and/or blocks in the flowchart illustrations and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing device to produce a machine, such that the instructions executed by the processor of the computer or other programmable data processing device produce a use A device for realizing the functions specified in one process or multiple processes of the flowchart and/or one block or multiple blocks of the block diagram.

These computer program instructions may also be stored in a computer-readable memory that causes a computer or other programmable data processing apparatus to operate in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including the instruction means, the instructions The device implements the functions specified in a process or processes of the flowchart and/or a block or blocks of the block diagram.

These computer program instructions may also be loaded onto a computer or other programmable data processing device, causing a series of operating steps to be performed on the computer or other programmable device to produce computer-implemented processing, thereby executing on the computer or other programmable device. Instructions provide steps for implementing the functions specified in a process or processes of a flowchart diagram and/or a block or blocks of a block diagram.

The above descriptions are only preferred embodiments of the present invention and are not intended to limit the scope of the present invention.

Claims (9)

1. A method of triggering, comprising:

after the slice instance is online, acquiring a slice instance of the network slice; the slice instance is created according to a slice template matched with the current service requirement;

Inputting the topological graph of the slice example into a trained Node2Vec model, and outputting a vector of a Node in the slice example; wherein the vector elements of the node include: an amount representing the structure of the node and an amount representing the load of the node;

determining whether to trigger reconstruction of the slice instance according to the vector of the node; the trained Node2Vec model is obtained by the following steps:

acquiring a starting node and a next-hop node of the starting node from a slice example to be trained, determining the next-hop node as a current node, and setting an initial value of sampling times i to be 1;

acquiring adjacent nodes of the current node;

when i is smaller than the sampling step length, calculating the offset migration probability of the current node and the adjacent node according to a preset offset migration probability algorithm; the preset bias migration probability is positively correlated with network performance parameters of the current service when the slice instance is adopted at the current moment;

sequencing the calculated offset walk probability according to the order from large to small, selecting the adjacent nodes ranked in the previous M, determining the selected adjacent nodes as current nodes, updating i to be i+1, and returning to execute the adjacent nodes for acquiring the current nodes; wherein M is a positive integer less than the number of adjacent nodes;

When i is greater than or equal to the sampling step length, inputting at least two groups of Node sequences obtained by sampling into a preset Node2Vec model for training to obtain the trained Node2Vec model;

the determining whether to trigger the reconstruction of the slice instance according to the vector of the node comprises:

according to the vector of the nodes, calculating the Euclidean distance between any two nodes in the slice example;

and triggering the reconstruction of the slice instance when the Euclidean distance between any two nodes meets a preset condition.

2. The method of claim 1, wherein the network performance parameters of the current service include one or more of:

the average occupied bandwidth of the current service, the average message delay of the current service and the average packet loss rate of the current service.

3. A method according to claim 1 or 2, wherein the bias walk probability of the current node to the adjacent node is calculated using the formula：

；

；

；

Wherein v represents the current node, t represents the neighboring node of the current node, W ₀ Representing the average occupied bandwidth of the current service at the current moment, T ₀ Average message delay representing current business at current moment E ₀ Represents the average packet loss rate, k of the current service at the current moment _W Is W ₀ Is (k) the estimated level of (k) _T Is T ₀ Is (k) the estimated level of (k) _E Is E ₀ Is used to determine the estimated level of (1),representing the shortest number of hops that node t needs to jump to node x.

4. The method of claim 1, wherein triggering the reconstruction of the slice instance when the euclidean distance between any two nodes satisfies a preset condition comprises:

and triggering the reconstruction of the slice instance when the Euclidean distance between any two nodes is smaller than the preset minimum Euclidean distance between the nodes.

5. The method of claim 1, wherein triggering the reconstruction of the slice instance when the euclidean distance between any two nodes satisfies a preset condition comprises:

and triggering the reconstruction of the slice instance when the Euclidean distance between any two nodes is larger than the preset maximum Euclidean distance between the nodes.

6. The method according to claim 1, wherein the method further comprises:

and when no distance smaller than the preset minimum Euclidean distance between the nodes exists in the Euclidean distance between any two nodes, and no distance larger than the preset maximum Euclidean distance between the nodes exists in the Euclidean distance between any two nodes, triggering the reconstruction of the slice instance is forbidden.

7. A triggering device, the device comprising:

the acquisition module is used for acquiring the slice instance of the network slice after the slice instance is online; the slice instance is created according to a slice template matched with the current service requirement; the processing module is used for inputting the topological graph of the slice example into a trained Node2Vec model and outputting a vector of a Node in the slice example; wherein the vector elements of the node include: an amount representing the structure of the node and an amount representing the load of the node;

the triggering module is used for determining whether to trigger the reconstruction of the slice instance according to the vector of the node; triggering the reconstruction of the slice instance when the Euclidean distance between any two nodes meets a preset condition;

the trained Node2Vec model is obtained by adopting the following modes:

acquiring a starting node and a next-hop node of the starting node from a slice example to be trained, determining the next-hop node as a current node, and setting an initial value of sampling times i to be 1;

acquiring adjacent nodes of the current node;

when i is smaller than the sampling step length, calculating the offset migration probability of the current node and the adjacent node according to a preset offset migration probability algorithm; the preset bias migration probability is positively correlated with network performance parameters of the current service when the slice instance is adopted at the current moment;

Sequencing the calculated offset walk probability according to the order from large to small, selecting the adjacent nodes ranked in the previous M, determining the selected adjacent nodes as current nodes, updating i to be i+1, and returning to execute the adjacent nodes for acquiring the current nodes; wherein M is a positive integer less than the number of adjacent nodes;

when i is greater than or equal to the sampling step length, inputting at least two groups of Node sequences obtained by sampling into a preset Node2Vec model for training, and obtaining the trained Node2Vec model.

8. An apparatus, the apparatus comprising:

a processor and a storage medium storing instructions executable by the processor, the storage medium performing operations in dependence on the processor through a communication bus, the instructions, when executed by the processor, performing the triggering method of any one of claims 1 to 6.

9. A computer storage medium storing executable instructions which, when executed by one or more processors, perform the triggering method of any one of claims 1 to 6.