Edge computing resource collaborative allocation method and system for the Internet of Things

Abstract

The present invention provides an edge computing resource collaborative allocation method and system for the Internet of Things, which relates to the field of data processing. The method includes: obtaining test computing resource demand data of multiple Internet of Things terminals; establishing an Internet of Things terminal association graph of multiple Internet of Things terminals based on the test computing resource demand data of multiple Internet of Things terminals; generating an initial edge computing resource collaborative allocation plan for multiple Internet of Things terminals based on the Internet of Things terminal association graph of multiple Internet of Things terminals; obtaining historical computing resource demand data of multiple Internet of Things terminals; based on the historical computing resource demand data of multiple Internet of Things terminals and the Internet of Things terminal association graph, performing real-time optimization on the initial edge computing resource collaborative allocation plan to generate a real-time edge computing resource collaborative allocation plan; and scheduling computing resources for multiple edge computing nodes according to the real-time edge computing resource collaborative allocation plan, which has the advantage of achieving balanced distribution of edge computing resources.

Description

Technical Field

[0001] The present invention relates to the field of data processing, and in particular to a method and system for collaborative allocation of edge computing resources for the Internet of Things. Background Art

[0002] Edge computing technology, by deploying relatively powerful and resource-rich edge servers at the edge of the network, close to user access networks, can mitigate the uncertain latency inherent in traditional cloud service architectures due to internet transmission. Therefore, edge computing is considered a key technology for the future implementation of intelligent IoT. However, in actual edge computing applications, the processing power and resources of edge servers are relatively limited, necessitating the allocation and scheduling of processing tasks.

[0003] For large-scale IoT systems, the computing tasks and data volumes generated vary significantly across space and time. If a fixed IoT architecture is employed, whereby an edge node receives computing tasks and related data from a fixed number of sensors and smart devices, the load on each edge node can be highly uneven. For example, some edge nodes may be idle while others are overloaded with computing tasks, storage space, and communication capabilities. Therefore, how to properly schedule computing tasks and their associated data, and evenly distribute them to the appropriate edge nodes, is a pressing issue.

[0004] Therefore, it is necessary to provide an edge computing resource collaborative allocation method and system for the Internet of Things to achieve balanced allocation of edge computing resources. Summary of the Invention

[0005] The present invention provides an edge computing resource collaborative allocation method for the Internet of Things, including: obtaining test computing resource demand data of multiple Internet of Things terminals, wherein the test computing resource demand data of the Internet of Things terminals includes the demand for multiple computing resources at multiple test time points; establishing an Internet of Things terminal association graph of the multiple Internet of Things terminals based on the test computing resource demand data of the multiple Internet of Things terminals; generating an initial edge computing resource collaborative allocation plan for the multiple Internet of Things terminals based on the Internet of Things terminal association graph of the multiple Internet of Things terminals; obtaining historical computing resource demand data of the multiple Internet of Things terminals; based on the historical computing resource demand data of the multiple Internet of Things terminals and the Internet of Things terminal association graph, performing real-time optimization on the initial edge computing resource collaborative allocation plan to generate a real-time edge computing resource collaborative allocation plan; and scheduling computing resources for multiple edge computing nodes according to the real-time edge computing resource collaborative allocation plan.

[0006] Furthermore, based on the test computing resource demand data of multiple IoT terminals, an IoT terminal association graph of multiple IoT terminals is established, including: for each computing resource, based on the test computing resource demand data of multiple IoT terminals, calculating the demand similarity of the computing resource demands of any two IoT terminals, based on the demand similarity of the computing resources of any two IoT terminals, determining the first similar IoT terminal corresponding to the computing resource of each IoT terminal, and establishing a first IoT terminal association subgraph of the computing resources of multiple IoT terminals based on the first similar IoT terminal corresponding to the computing resource of each IoT terminal; for each IoT terminal, based on the test computing resource demand data of the IoT terminal, calculating the demand correlation coefficient of the IoT terminal corresponding to any two computing resources; based on the demand correlation coefficient of any two computing resources corresponding to each IoT terminal, calculating the demand correlation similarity of any two IoT terminals; based on the demand correlation similarity of any two IoT terminals, determining the second similar IoT terminal of each IoT terminal based on the demand correlation similarity of any two IoT terminals, and establishing a second IoT terminal association subgraph of multiple IoT terminals based on the second similar IoT terminal of each IoT terminal, wherein the IoT terminal association graph includes the first IoT terminal association subgraph and the second IoT terminal association subgraph corresponding to each computing resource demand.

[0007] Furthermore, based on the test computing resource demand data of multiple IoT terminals, the demand similarity of the computing resource demands of any two IoT terminals is calculated, including: for each IoT terminal, performing time-varying filtered empirical mode decomposition on the computing resource demands of the IoT terminal at multiple test time points, obtaining multiple target intrinsic mode functions of the computing resources corresponding to the IoT terminal, and extracting the functional characteristics of each target intrinsic mode function; for any two IoT terminals, based on the functional characteristics of each target intrinsic mode function of the computing resources corresponding to the two IoT terminals, the demand similarity of the computing resource demands of any two IoT terminals is calculated.

[0008] Furthermore, based on the computing resource demand association graph of multiple IoT terminals, an initial edge computing resource collaborative allocation plan for multiple IoT terminals is generated, including: establishing a sample database, wherein the sample database is used to store the computing resource demand characteristics of multiple sample IoT terminal groups, the first IoT terminal association subgraph and the second IoT terminal association subgraph corresponding to each computing resource, and the edge computing resource collaborative allocation plan; based on the computing resource demand association graph of multiple IoT terminals, similar sample IoT terminal groups are determined from multiple sample IoT terminal groups; based on the edge computing resource collaborative allocation plan of similar sample IoT terminal groups, an initial edge computing resource collaborative allocation plan for multiple IoT terminals is generated.

[0009] Furthermore, based on the computing resource demand association graphs of multiple IoT terminals, similar sample IoT terminal groups are determined from multiple sample IoT terminal groups, including: for each first IoT terminal association subgraph of multiple IoT terminals, extracting global features and local features of the first IoT terminal association subgraph; extracting global features and local node features of the second IoT terminal association subgraphs of multiple IoT terminals; based on the computing resource demand characteristics of multiple IoT terminals, the global features and local node features of each first IoT terminal association subgraph and the second IoT terminal association subgraph, determining similar sample IoT terminal groups from multiple sample IoT terminal groups.

[0010] Furthermore, based on the historical computing resource demand data and computing resource demand association graph of multiple IoT terminals, the initial edge computing resource collaborative allocation plan is optimized in real time to generate a real-time edge computing resource collaborative allocation plan, including: based on the historical computing resource demand data and computing resource demand association graph of multiple IoT terminals, predicting the future computing resource demand data of multiple IoT terminals; based on the future computing resource demand data and computing resource demand association graph of multiple IoT terminals, the initial edge computing resource collaborative allocation plan is optimized in real time to generate a real-time edge computing resource collaborative allocation plan.

[0011] Furthermore, based on the historical computing resource demand data and computing resource demand association graph of multiple IoT terminals, future computing resource demand data of multiple IoT terminals are predicted, including: for each IoT terminal, based on the historical computing resource demand data of the IoT terminal, initial future computing resource demand data of the IoT terminal is predicted, wherein the initial future computing resource demand data of the IoT terminal includes the initial demand of the IoT terminal for multiple computing resources at multiple future time points; for each computing resource, based on the initial demand of a first similar IoT terminal for the computing resource corresponding to the IoT terminal at multiple future time points, the initial demand of the IoT terminal for the computing resource at multiple future time points is iteratively corrected for the first time, and the corrected demand of the IoT terminal for the computing resource at multiple future time points is generated; for each IoT terminal, based on the corrected demand of a second similar IoT terminal for each computing resource at multiple future time points, the corrected demand of the IoT terminal for the computing resource at multiple future time points is iteratively corrected for the second time, and the future computing resource demand data of the IoT terminal is generated.

[0012] Furthermore, based on the future computing resource demand data of multiple IoT terminals and the computing resource demand association graph, the initial edge computing resource collaborative allocation plan is optimized in real time to generate a real-time edge computing resource collaborative allocation plan, including: determining the IoT terminals to be allocated and the edge computing nodes to be allocated based on the future computing resource demand data of multiple IoT terminals and the initial edge computing resource collaborative allocation plan; generating multiple candidate edge computing resource collaborative allocation plans based on the IoT terminals to be allocated and the edge computing nodes to be allocated through a Monte Carlo model; establishing an optimization evaluation function; and generating a real-time edge computing resource collaborative allocation plan based on the computing resource demand association graph, the optimization evaluation function and the multiple candidate edge computing resource collaborative allocation plans.

[0013] Furthermore, based on the computing resource demand association graph, the optimization evaluation function and multiple candidate edge computing resource collaborative allocation schemes, a real-time edge computing resource collaborative allocation scheme is generated, including: calculating the optimization value of each candidate edge computing resource collaborative allocation scheme based on the computing resource demand association graph and the optimization evaluation function; generating a real-time edge computing resource collaborative allocation scheme based on the optimization value of each candidate edge computing resource collaborative allocation scheme through a genetic algorithm.

[0014] The present invention provides an edge computing resource collaborative allocation system for the Internet of Things, which is used for the above-mentioned edge computing resource collaborative allocation method for the Internet of Things, including: a demand testing module, used to obtain test computing resource demand data of multiple Internet of Things terminals, wherein the test computing resource demand data of the Internet of Things terminals include the demand for multiple computing resources at multiple test time points; an association establishment module, used to establish an Internet of Things terminal association graph of multiple Internet of Things terminals based on the test computing resource demand data of multiple Internet of Things terminals; an initial allocation module, used to generate an initial edge computing resource collaborative allocation plan for multiple Internet of Things terminals based on the Internet of Things terminal association graph of multiple Internet of Things terminals; a demand acquisition module, used to obtain historical computing resource demand data of multiple Internet of Things terminals; a real-time optimization module, used to optimize the initial edge computing resource collaborative allocation plan in real time based on the historical computing resource demand data of multiple Internet of Things terminals and the Internet of Things terminal association graph, and generate a real-time edge computing resource collaborative allocation plan; a collaborative allocation module, used to schedule computing resources for multiple edge computing nodes according to the real-time edge computing resource collaborative allocation plan.

[0015] Compared with the existing technology, the method and system for collaborative allocation of edge computing resources for the Internet of Things provided by the present invention have at least the following beneficial effects:

[0016] 1. By acquiring and analyzing the test computing resource demand data of multiple IoT terminals, the actual resource demand of IoT terminals can be predicted and met more accurately. This helps avoid over-allocation or under-allocation of resources, thereby improving the overall utilization of edge computing resources. An initial resource allocation plan is generated based on the IoT terminal association graph, and the plan is optimized in real time to match historical data. This real-time nature enables edge computing to respond more quickly to changes in resource requirements of IoT terminals, improving the response speed and flexibility of edge computing. By comprehensively considering the test and historical computing resource demand data of IoT terminals, a more scientific and reasonable resource allocation strategy can be formulated. This helps balance the resource requirements between different IoT terminals and achieve fair and efficient allocation of resources. Real-time optimization of resource allocation plans can ensure that edge computing nodes can maintain stable operation when computing resource demands fluctuate. This helps reduce the risk of edge computing crashes or performance degradation due to insufficient or excessive resources.

[0017] 2. By performing time-varying filtered empirical mode decomposition on the computing resource requirements of IoT terminals at multiple test time points and extracting functional features, the similarity of the requirements of any two IoT terminals for the same computing resource can be accurately calculated. This helps identify terminals with similar resource requirements, enabling more precise resource matching and allocation. Calculating the correlation coefficient of the requirements of IoT terminals for any two computing resources reveals the correlation between different resource requirements of the terminals. This helps consider the diverse needs of terminals during resource allocation and achieve comprehensive resource optimization. By constructing the first IoT terminal association subgraph and the second IoT terminal association subgraph, the correlation between the computing resource requirements of IoT terminals can be comprehensively reflected. This provides a strong basis for resource collaborative allocation and helps achieve optimal resource configuration and efficient utilization. Leveraging historical data and experience in the sample database, an initial edge computing resource collaborative allocation plan for multiple IoT terminals can be quickly generated. This approach not only improves resource allocation efficiency but also reduces the cost of manual intervention.

[0018] 3. Real-time optimization based on future computing resource demand data from IoT terminals can more accurately predict and meet the terminals' future resource needs. This helps avoid over- or under-allocation of resources and ensures efficient resource utilization. Using a Monte Carlo model, multiple candidate edge computing resource co-allocation schemes are generated, providing the system with a variety of possible resource allocation strategies. This allows the system to quickly adjust allocation plans to meet new requirements when resource demands change. By establishing an optimization evaluation function, each candidate resource allocation scheme can be evaluated to select the optimal one. This helps ensure the rationality and efficiency of resource allocation. Using a genetic algorithm to optimize the candidate edge computing resource co-allocation schemes, a resource allocation strategy closer to the optimal solution is found. Genetic algorithms, as optimization algorithms that mimic biological evolution, possess strong global search capabilities and adaptability, enabling them to find optimal solutions to complex resource allocation problems. They can optimize resource allocation schemes in real time and offer powerful dynamic adjustment capabilities. This helps the system respond quickly to changes in IoT terminal resource demands, ensuring system stability and performance. More accurate real-time resource allocation and efficient optimization algorithms can significantly improve edge computing resource utilization. This helps reduce resource waste and lower operating costs. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] This specification will be further described in the form of exemplary embodiments, which will be described in detail with reference to the accompanying drawings. These embodiments are not limiting, and in these embodiments, like numbers represent like structures, wherein:

[0020] Figure 1 This is a flowchart of a collaborative allocation method for edge computing resources for the Internet of Things according to some embodiments of this specification;

[0021] Figure 2 is a schematic diagram of a first IoT terminal association subgraph according to some embodiments of this specification;

[0022] Figure 3 This is a module diagram of an edge computing resource collaborative allocation system for the Internet of Things according to some embodiments of this specification. DETAILED DESCRIPTION

[0023] To more clearly illustrate the technical solutions of the embodiments of this specification, the following briefly describes the drawings required for describing the embodiments. Obviously, the drawings described below are merely examples or embodiments of this specification. Those skilled in the art can apply this specification to other similar scenarios based on these drawings without inventive effort. Unless otherwise apparent from the context or otherwise noted, the same reference numerals in the figures represent the same structure or operation.

[0024] Figure 1 This is a flow chart of a method for collaboratively allocating edge computing resources for the Internet of Things according to some embodiments of this specification, such as Figure 1 As shown, the collaborative allocation method of edge computing resources for the Internet of Things may include the following process.

[0025] S101. Obtain test computing resource demand data for multiple IoT terminals.

[0026] Among them, the test computing resource demand data of the IoT terminal includes the demand for various computing resources (for example, CPU, memory, storage, bandwidth, GPU, etc.) at multiple test time points.

[0027] Understandably, when multiple IoT terminals connected to the same edge computing node process data collected using the same type of computing resources, one type of resource will be rapidly consumed, leaving a large number of other types of computing resources idle and unused. However, since this type of computing resource has been consumed and cannot be used by other applications, the other idle resources will also be unused, resulting in low resource utilization.

[0028] S102: Establish an IoT terminal association graph of the multiple IoT terminals based on the test computing resource demand data of the multiple IoT terminals.

[0029] In some embodiments, S102 specifically includes:

[0030] For each computing resource, based on the test computing resource demand data of multiple IoT terminals, the demand similarity of the computing resource requirements of any two IoT terminals is calculated; based on the demand similarity of the computing resource requirements of any two IoT terminals, a first similar IoT terminal corresponding to the computing resource of each IoT terminal is determined; based on the first similar IoT terminal corresponding to the computing resource of each IoT terminal, a first IoT terminal association subgraph corresponding to the computing resources of the multiple IoT terminals is established;

[0031] For each IoT terminal, based on the IoT terminal's test computing resource demand data, calculate the demand correlation coefficient between any two computing resources corresponding to the IoT terminal;

[0032] Based on the demand correlation coefficient of each IoT terminal corresponding to any two computing resources, calculate the demand correlation similarity of any two IoT terminals;

[0033] Based on the demand-related similarity of any two IoT terminals, a second similar IoT terminal is determined for each IoT terminal. Based on the second similar IoT terminal of each IoT terminal, a second IoT terminal association subgraph of multiple IoT terminals is established, wherein the IoT terminal association graph includes a first IoT terminal association subgraph and a second IoT terminal association subgraph corresponding to each computing resource demand.

[0034] In some embodiments, based on the test computing resource demand data of multiple IoT terminals, calculating the demand similarity of computing resource demands corresponding to any two IoT terminals includes:

[0035] For each IoT terminal, perform time-varying filtered empirical mode decomposition on the IoT terminal's demand for computing resources at multiple test time points to obtain multiple target intrinsic mode functions corresponding to the IoT terminal's computing resources, and extract the functional characteristics of each target intrinsic mode function;

[0036] For any two IoT terminals, the demand similarity of the computing resource demands of the two IoT terminals is calculated based on the function characteristics of each target intrinsic modal function of the computing resources corresponding to the two IoT terminals.

[0037] Specifically, the multiple target intrinsic mode functions may be the first few intrinsic mode functions obtained according to the decomposition order, for example, the first three intrinsic mode functions. Functional characteristics of the target intrinsic mode functions may include at least variance, amplitude, phase, instantaneous frequency, spectral characteristics, bandwidth, etc.

[0038] The similarity of computing resource requirements of two IoT terminals can be calculated using the following formula:

[0039]

[0040] in, is the demand similarity between the i-th IoT terminal and the j-th IoT terminal corresponding to the e-th computing resource demand, N is the total number of target intrinsic mode functions of computing resources corresponding to an IoT terminal, is the functional characteristic of the nth target intrinsic mode function corresponding to the eth computing resource requirement of the i-th IoT terminal, is the functional characteristic of the nth target intrinsic mode function corresponding to the eth computing resource requirement of the jth IoT terminal, It is the cosine similarity between the function characteristics of the nth target intrinsic modal function corresponding to the eth computing resource demand of the i-th IoT terminal and the function characteristics of the nth target intrinsic modal function corresponding to the eth computing resource demand of the j-th IoT terminal.

[0041] Figure 2is a schematic diagram of a first IoT terminal associated subgraph according to some embodiments of this specification, such as Figure 2 As shown, the first IoT terminal association subgraph includes nodes representing multiple IoT terminals. When the demand similarity of two IoT terminals is greater than the demand similarity threshold, the two nodes representing the two IoT terminals in the first IoT terminal association subgraph are connected by an edge. The length of the edge represents the demand similarity of the two IoT terminals. The greater the demand similarity, the shorter the edge.

[0042] For each IoT terminal, based on the test computing resource demand data of the IoT terminal, the Pearson correlation coefficient between any two computing resources corresponding to the IoT terminal is calculated as the demand correlation coefficient.

[0043] The demand-related similarity of two IoT terminals can be calculated according to the following formula:

[0044]

[0045] in, is the demand-related similarity between the i-th IoT terminal and the j-th IoT terminal, is the demand correlation coefficient of the ith IoT terminal corresponding to any e-th computing resource and the f-th computing resource, is the demand correlation coefficient of the j-th IoT terminal corresponding to any e-th computing resource and the f-th computing resource, and E is the total number of computing resource types.

[0046] The second IoT terminal association subgraph includes nodes representing multiple IoT terminals. When the demand-related similarity between two IoT terminals is greater than a demand-related similarity threshold, two nodes representing the two IoT terminals in the first IoT terminal association subgraph are connected by an edge. The length of the edge represents the demand-related similarity between the two IoT terminals. The greater the demand-related similarity, the shorter the edge.

[0047] S103: Based on the IoT terminal association graph of the multiple IoT terminals, generate an initial edge computing resource collaborative allocation plan for the multiple IoT terminals.

[0048] In some embodiments, S103 specifically includes:

[0049] Establishing a sample database, wherein the sample database is used to store computing resource demand characteristics of multiple sample IoT terminal groups, a first IoT terminal association subgraph and a second IoT terminal association subgraph corresponding to each computing resource, and an edge computing resource collaborative allocation scheme. The edge computing resource collaborative allocation scheme for the sample IoT terminal groups can be determined manually or using experimental data;

[0050] Based on the computing resource demand association graph of multiple IoT terminals, determine similar sample IoT terminal groups from multiple sample IoT terminal groups;

[0051] Based on the edge computing resource collaborative allocation scheme of similar sample IoT terminal groups, an initial edge computing resource collaborative allocation scheme for multiple IoT terminals is generated.

[0052] In some embodiments, determining a similar sample IoT terminal group from a plurality of sample IoT terminal groups based on a computing resource requirement association graph of a plurality of IoT terminals includes:

[0053] For each first IoT terminal association subgraph of the plurality of IoT terminals, extract global features (e.g., size, density, global clustering coefficient, etc.) and local features (e.g., degree of each node, mean value of edges of each node, local clustering coefficient, etc.) of the first IoT terminal association subgraph;

[0054] Extracting global features (e.g., size, density, global clustering coefficient, etc.) and local node features (e.g., degree of each node, mean value of edges of each node, local clustering coefficient, etc.) of a second IoT terminal association subgraph of the plurality of IoT terminals;

[0055] Based on the computing resource demand characteristics of the plurality of IoT terminals, the global characteristics and the local node characteristics of each first IoT terminal associated subgraph and the second IoT terminal associated subgraph, a similar sample IoT terminal group is determined from the plurality of sample IoT terminal groups.

[0056] Specifically, the size of the global feature represents the number of nodes in the IoT terminal association subgraph, which can reflect the complexity and information content of the IoT terminal association subgraph. The density represents the ratio of the number of edges to the number of nodes in the IoT terminal association subgraph. The IoT terminal association subgraph with high density may indicate that there is a strong mutual correlation between the nodes. The global clustering coefficient can measure the connection tightness between the neighboring nodes of all nodes in the IoT terminal association subgraph. The number of all triples (structures composed of three nodes) in the IoT terminal association subgraph can be counted, and the number of closed triples (that is, three nodes are connected to each other) can be counted. The global clustering coefficient is equal to the number of closed triples divided by the total number of triplets.

[0057] The degree of each node in the local node feature represents the number of edges connecting it to other nodes, which can reflect the node's activity in the IoT terminal association subgraph. The mean of the edges associated with each node can be calculated as the mean length of the edges connecting it to other nodes. The local clustering coefficient measures the degree of connectivity between a single node's neighbors in the IoT terminal association subgraph. For a node, all its neighboring nodes are identified and the number of edges actually existing between them is counted. The local clustering coefficient is calculated by dividing the actual number of edges by the maximum possible number of edges between neighboring nodes. For example, if there is an edge between nodes A and B, then node A is a neighbor of node B, and node B is also a neighbor of node A. The number of node A's neighboring nodes, k, is 4. The maximum possible number of edges between node A's neighbors is 4(4-1) / 2 = 6. If there are edges connecting nodes B and C, C and D, and D and E, respectively, then the actual number of edges is 3. Therefore, the local clustering coefficient of node A is 3 / 6 = 0.5.

[0058] The computing resource demand characteristics of the IoT terminal may include the mean and variance of the IoT terminal's demand for each computing resource, which can be calculated based on the IoT terminal's demand for computing resources at multiple test time points.

[0059] The extracted features of each first IoT terminal associated subgraph and the extracted features of the second IoT terminal associated subgraph are quantified to form a feature vector. This feature vector includes the features of each first IoT terminal associated subgraph and the global features, local features, and computing resource requirement features of the second IoT terminal associated subgraph.

[0060] For each sample IoT terminal group, features extracted from each first IoT terminal association subgraph and second IoT terminal association subgraph of the sample IoT terminal group may be quantized to form a feature vector of the sample IoT terminal group.

[0061] The Euclidean distance between the feature vectors of multiple IoT terminals and the feature vector of any sample IoT terminal group can be calculated, and the sample IoT terminal group whose Euclidean distance is less than a Euclidean distance threshold is regarded as a similar sample IoT terminal group.

[0062] An initial edge computing resource collaborative allocation scheme for multiple IoT terminals can be generated through a scheme generation model based on feature vectors of multiple IoT terminals, feature vectors of similar sample IoT terminal groups, and edge computing resource collaborative allocation schemes for similar sample IoT terminal groups. The scheme generation model can be a feedforward neural network model.

[0063] S104: Obtain historical computing resource demand data of multiple IoT terminals.

[0064] Specifically, the historical computing resource demand data of the IoT terminal may include the IoT terminal's demand for various computing resources (e.g., CPU, memory, storage, bandwidth, GPU, etc.) at multiple historical time points in the current cycle (e.g., one day, one week, etc.).

[0065] S105. Based on the historical computing resource demand data of multiple IoT terminals and the IoT terminal association graph, the initial edge computing resource collaborative allocation plan is optimized in real time to generate a real-time edge computing resource collaborative allocation plan.

[0066] In some embodiments, S105 specifically includes:

[0067] Based on the historical computing resource demand data and computing resource demand association graph of multiple IoT terminals, the future computing resource demand data of multiple IoT terminals is predicted;

[0068] Based on the future computing resource demand data and computing resource demand association graph of multiple IoT terminals, the initial edge computing resource collaborative allocation plan is optimized in real time to generate a real-time edge computing resource collaborative allocation plan.

[0069] In some embodiments, predicting future computing resource demand data of multiple IoT terminals based on historical computing resource demand data of multiple IoT terminals and a computing resource demand association graph includes:

[0070] For each IoT terminal, based on the historical computing resource demand data of the IoT terminal, predicting initial future computing resource demand data of the IoT terminal, wherein the initial future computing resource demand data of the IoT terminal includes initial demands of the IoT terminal for multiple computing resources at multiple future time points;

[0071] For each computing resource, based on the initial demands for the computing resource by a first similar IoT terminal corresponding to the computing resource at multiple future time points, the initial demands for the computing resource by the IoT terminal at multiple future time points are first iteratively revised to generate revised demands for the computing resource by the IoT terminal at multiple future time points.

[0072] For each IoT terminal, based on the revised requirements of a second similar IoT terminal for each computing resource at multiple future time points, the revised requirements of the IoT terminal for computing resources at multiple future time points are iteratively revised for the second time to generate future computing resource requirement data for the IoT terminal.

[0073] Specifically, for each IoT terminal, a demand prediction model corresponding to the IoT terminal can be established. The demand prediction model corresponding to the IoT terminal can be used to predict the initial future computing resource demand data of the IoT terminal based on the historical computing resource demand data of the IoT terminal. The demand prediction model can be a long short-term memory network model.

[0074] The initial demand for computing resources by IoT terminals at multiple future time points can be revised for the first iteration according to the following process:

[0075] S1051. Establish a global correction model corresponding to the computing resources, wherein the global correction model may be a long short-term memory network model;

[0076] S1052. Perform a first round of corrections on the initial demand for computing resources by each IoT terminal at multiple future time points based on the initial demand for computing resources by a first similar IoT terminal corresponding to the computing resource of each IoT terminal at multiple future time points using a global correction model corresponding to the computing resources, thereby generating the demand for computing resources by each IoT terminal at multiple future time points after the first round of corrections.

[0077] S1053. Calculate a global revised difference based on the difference between the initial demand for computing resources of each IoT terminal at multiple future time points and the demand for computing resources of each IoT terminal at multiple future time points after the first round of revision.

[0078] S1054: Determine whether the global correction difference is less than the global correction difference threshold. If so, complete the first iterative correction corresponding to the computing resource, proceed to the first iterative correction of the next computing resource, and execute S1051. If not, execute S1055.

[0079] S1055. Perform a t-round correction on the computing resource demands of each IoT terminal at multiple future time points after the t-1 round of corrections using a global correction model corresponding to the computing resources, based on the computing resource demands of a first similar IoT terminal corresponding to each IoT terminal at multiple future time points after the t-1 round of corrections, to generate the computing resource demands of each IoT terminal at multiple future time points after the t-1 round of corrections.

[0080] S1056. Calculate a global correction difference based on the difference between the demand for computing resources of each IoT terminal at multiple future time points after the t-th round of correction and the demand for computing resources of each IoT terminal at multiple future time points after the t-1-th round of correction, and execute S1054.

[0081] For each IoT terminal, a corresponding association correction model is established. The association correction model corresponding to each IoT terminal is used to correct the IoT terminal's correction requirements for computing resources at multiple future time points based on the correction requirements of a second similar IoT terminal for each computing resource at multiple future time points. The association correction model can be a short-term memory network model.

[0082] In some embodiments, based on the future computing resource demand data and computing resource demand association graph of multiple IoT terminals, the initial edge computing resource collaborative allocation plan is optimized in real time to generate a real-time edge computing resource collaborative allocation plan, specifically including:

[0083] Based on the future computing resource demand data of multiple IoT terminals and the initial edge computing resource collaborative allocation plan, determine the IoT terminals and edge computing nodes to be allocated;

[0084] Through the Monte Carlo model, multiple candidate edge computing resource collaborative allocation schemes are generated based on the IoT terminals and edge computing nodes to be allocated.

[0085] Establish optimization evaluation function;

[0086] Based on the computing resource demand association graph, optimization evaluation function and multiple candidate edge computing resource collaborative allocation schemes, a real-time edge computing resource collaborative allocation scheme is generated.

[0087] Specifically, for each edge computing node, it can be determined whether some computing resources of the edge computing node are overloaded based on the initial edge computing resource collaborative allocation plan and the future computing resource demand data of multiple IoT terminals. If so, the IoT terminal with the smallest future computing resource demand data in the edge computing node is used as the IoT terminal to be allocated. Based on the initial edge computing resource collaborative allocation plan and the future computing resource demand data of multiple IoT terminals, it can be determined whether some computing resources of the edge computing node are idle. If so, the edge computing node is used as the edge computing node to be allocated.

[0088] You can set the initial parameters for the Monte Carlo simulation, such as the number of simulations and the randomness range for resource allocation. Based on the probability model, IoT terminals and edge computing nodes are randomly sampled to determine which node each terminal is assigned to and how much resources are allocated. The result of each random sampling is used as a candidate for edge computing resource collaborative allocation. This process is repeated multiple times to generate multiple candidate solutions.

[0089] The optimization evaluation function can be related to multiple factors, such as the average completion processing time or maximum completion time of the data collected by the IoT terminal, the utilization rate of computing, storage, network and other resources of the edge computing node, and the load balancing between the edge computing nodes.

[0090] The optimization evaluation function can also be correlated with the similarity of computing resource requirements between any two IoT terminals and the related similarity of requirements between any two IoT terminals. For example, in a candidate edge computing resource collaborative allocation scheme, the smaller the mean similarity of computing resource requirements between any two IoT terminals assigned to each edge computing node and the smaller the related similarity of requirements, the greater the optimization value of the candidate edge computing resource collaborative allocation scheme.

[0091] In some embodiments, a real-time edge computing resource collaborative allocation scheme is generated based on a computing resource demand association graph, an optimization evaluation function, and a plurality of candidate edge computing resource collaborative allocation schemes, including:

[0092] Based on the computing resource demand association graph and the optimization evaluation function, the optimization value of each candidate edge computing resource collaborative allocation scheme is calculated;

[0093] A real-time edge computing resource collaborative allocation scheme is generated based on the optimization value of each candidate edge computing resource collaborative allocation scheme through a genetic algorithm.

[0094] Specifically, each candidate edge computing resource collaborative allocation solution is encoded as an individual in a genetic algorithm, typically using binary or real number encoding. Based on the optimization value of each candidate edge computing resource collaborative allocation solution, a certain number of outstanding individuals are selected from multiple candidate edge computing resource collaborative allocation solutions using methods such as roulette wheel selection and tournament selection to serve as parents. A crossover operation is performed on the selected parent individuals to generate new offspring individuals. The crossover operation simulates the gene exchange during biological reproduction and can generate new edge computing resource collaborative allocation solutions. A mutation operation is performed on the offspring individuals to randomly change some of their gene values to increase the diversity of the population. The mutation operation helps to escape from local optimal solutions and explore a wider solution space. Repeated selection, crossover, and mutation operations continuously generate new populations. As iterations proceed, individuals in the population gradually approach the optimal solution. Termination conditions are set, such as reaching the maximum number of iterations or a fitness threshold. When the termination conditions are met, iterations are terminated and the optimal individual is output as the real-time edge computing resource collaborative allocation solution.

[0095] S106. Schedule computing resources for multiple edge computing nodes according to the real-time edge computing resource collaborative allocation plan.

[0096] Specifically, according to the real-time edge computing resource collaborative allocation scheme, the required computing resources are reserved on each edge computing node for the corresponding IoT terminal. After the resource reservation is completed, computing resources are dynamically allocated to the IoT terminal based on actual needs to ensure that its computing tasks can be carried out smoothly.

[0097] Figure 3 This is a module diagram of an edge computing resource collaborative allocation system for the Internet of Things according to some embodiments of this specification, such as Figure 3 As shown, the edge computing resource collaborative allocation system for the Internet of Things can include a demand testing module, an association establishment module, an initial allocation module, a demand acquisition module, a real-time optimization module and a collaborative allocation module.

[0098] A demand testing module is used to obtain test computing resource demand data of multiple IoT terminals, wherein the test computing resource demand data of the IoT terminals includes demands for multiple computing resources at multiple test time points;

[0099] an association establishing module, for establishing an IoT terminal association graph of the plurality of IoT terminals based on the test computing resource demand data of the plurality of IoT terminals;

[0100] An initial allocation module, configured to generate an initial edge computing resource collaborative allocation plan for multiple IoT terminals based on an IoT terminal association graph of the multiple IoT terminals;

[0101] Demand acquisition module, used to obtain historical computing resource demand data of multiple IoT terminals;

[0102] A real-time optimization module is used to optimize the initial edge computing resource collaborative allocation plan in real time based on the historical computing resource demand data of multiple IoT terminals and the IoT terminal association graph, and generate a real-time edge computing resource collaborative allocation plan;

[0103] The collaborative allocation module is used to schedule computing resources for multiple edge computing nodes based on the real-time edge computing resource collaborative allocation scheme.

[0104] The IoT-oriented edge computing resource collaborative allocation system can be used to execute the IoT-oriented edge computing resource collaborative allocation method, which will not be described in detail here.

[0105] Finally, it should be understood that the embodiments described in this specification are intended only to illustrate the principles of the embodiments of this specification. Other variations may also fall within the scope of this specification. Therefore, by way of example and not limitation, alternative configurations of the embodiments of this specification may be considered consistent with the teachings of this specification. Accordingly, the embodiments of this specification are not limited to the embodiments explicitly described and illustrated in this specification.

Claims

1. A collaborative allocation method for edge computing resources for the Internet of Things, characterized by: include: Obtaining test computing resource demand data of multiple IoT terminals, wherein the test computing resource demand data of the IoT terminals includes demands for multiple computing resources at multiple test time points; Based on the test computing resource demand data of multiple IoT terminals, an IoT terminal association graph of the multiple IoT terminals is established; Based on the IoT terminal association graph of multiple IoT terminals, an initial edge computing resource collaborative allocation plan for multiple IoT terminals is generated; Obtain historical computing resource demand data for multiple IoT terminals; Based on the historical computing resource demand data of multiple IoT terminals and the IoT terminal association graph, the initial edge computing resource collaborative allocation plan is optimized in real time to generate a real-time edge computing resource collaborative allocation plan; According to the real-time edge computing resource collaborative allocation scheme, multiple edge computing nodes are scheduled for computing resources; Wherein, based on the test computing resource demand data of multiple IoT terminals, an IoT terminal association graph of the multiple IoT terminals is established, including: For each computing resource, based on the test computing resource demand data of multiple IoT terminals, the demand similarity of the computing resource requirements of any two IoT terminals is calculated; based on the demand similarity of the computing resource requirements of any two IoT terminals, a first similar IoT terminal corresponding to the computing resource of each IoT terminal is determined; based on the first similar IoT terminal corresponding to the computing resource of each IoT terminal, a first IoT terminal association subgraph corresponding to the computing resources of the multiple IoT terminals is established; For each IoT terminal, based on the IoT terminal's test computing resource demand data, calculate the demand correlation coefficient between any two computing resources corresponding to the IoT terminal; For each IoT terminal, we perform time-varying filtered empirical mode decomposition on the IoT terminal's demand for computing resources at multiple test time points to obtain multiple target intrinsic mode functions (IMFs) of the IoT terminal's corresponding computing resources. We then extract the functional characteristics of each IMF. For any two IoT terminals, we calculate the similarity of the computing resource demands of the two IoT terminals based on the functional characteristics of each IMF of the computing resources of the two IoT terminals using the following formula: in, is the similarity of the demand for the e-th computing resource demand of the i-th IoT terminal and the j-th IoT terminal, N is the total number of target intrinsic mode functions of computing resources corresponding to an IoT terminal, is the functional characteristic of the nth target intrinsic mode function corresponding to the eth computing resource requirement of the i-th IoT terminal, is the functional characteristic of the nth target intrinsic mode function corresponding to the eth computing resource requirement of the jth IoT terminal, is the cosine similarity between the function characteristics of the nth target intrinsic modal function corresponding to the eth computing resource demand of the i-th IoT terminal and the function characteristics of the nth target intrinsic modal function corresponding to the eth computing resource demand of the j-th IoT terminal; Based on the demand-related similarity of any two IoT terminals, a second similar IoT terminal of each IoT terminal is determined. Based on the second similar IoT terminal of each IoT terminal, a second IoT terminal association subgraph of multiple IoT terminals is established, wherein the IoT terminal association graph includes a first IoT terminal association subgraph and a second IoT terminal association subgraph corresponding to each computing resource demand. The first IoT terminal association subgraph includes nodes representing multiple IoT terminals. When the demand similarity of two IoT terminals is greater than a demand similarity threshold, the two nodes representing the two IoT terminals in the first IoT terminal association subgraph are connected by an edge. The length of the edge represents the demand similarity of the two IoT terminals. The greater the demand similarity, the shorter the edge. The second IoT terminal association subgraph includes nodes representing multiple IoT terminals. When the demand-related similarity of two IoT terminals is greater than the demand-related similarity threshold, the two nodes representing the two IoT terminals in the first IoT terminal association subgraph are connected by an edge. The length of the edge represents the demand similarity of the two IoT terminals. The greater the demand similarity, the shorter the edge.

2. The method for collaborative allocation of edge computing resources for the Internet of Things according to claim 1, characterized in that: Based on the computing resource demand association graph of multiple IoT terminals, an initial edge computing resource collaborative allocation plan for multiple IoT terminals is generated, including: Establishing a sample database, wherein the sample database is used to store computing resource demand characteristics of multiple sample IoT terminal groups, a first IoT terminal association subgraph and a second IoT terminal association subgraph corresponding to each computing resource, and an edge computing resource collaborative allocation scheme; Based on the computing resource demand correlation graph of multiple IoT terminals, determine similar sample IoT terminal groups from multiple sample IoT terminal groups; Based on the edge computing resource collaborative allocation scheme of similar sample IoT terminal groups, an initial edge computing resource collaborative allocation scheme for multiple IoT terminals is generated.

3. The method for collaborative allocation of edge computing resources for the Internet of Things according to claim 2, characterized in that: Based on the computing resource requirement association graph of multiple IoT terminals, similar sample IoT terminal groups are determined from multiple sample IoT terminal groups, including: For each first Internet of Things terminal associated subgraph of the plurality of Internet of Things terminals, extracting global features and local features of the first Internet of Things terminal associated subgraph; Extracting global features and local node features of a second IoT terminal associated subgraph of the plurality of IoT terminals; Based on the computing resource demand characteristics of the plurality of IoT terminals, the global characteristics and the local node characteristics of each first IoT terminal associated subgraph and the second IoT terminal associated subgraph, a similar sample IoT terminal group is determined from the plurality of sample IoT terminal groups.

4. The method for collaborative allocation of edge computing resources for the Internet of Things according to claim 1, characterized in that: Based on the historical computing resource demand data and computing resource demand association graph of multiple IoT terminals, the initial edge computing resource collaborative allocation plan is optimized in real time to generate a real-time edge computing resource collaborative allocation plan, including: Based on the historical computing resource demand data and computing resource demand association graph of multiple IoT terminals, the future computing resource demand data of multiple IoT terminals is predicted; Based on the future computing resource demand data and computing resource demand association graph of multiple IoT terminals, the initial edge computing resource collaborative allocation plan is optimized in real time to generate a real-time edge computing resource collaborative allocation plan.

5. The method for collaborative allocation of edge computing resources for the Internet of Things according to claim 4, characterized in that: Based on the historical computing resource demand data and computing resource demand correlation graph of multiple IoT terminals, the future computing resource demand data of multiple IoT terminals is predicted, including: For each IoT terminal, based on the historical computing resource demand data of the IoT terminal, predicting initial future computing resource demand data of the IoT terminal, wherein the initial future computing resource demand data of the IoT terminal includes initial demands of the IoT terminal for multiple computing resources at multiple future time points; For each computing resource, based on the initial demands for the computing resource by a first similar IoT terminal corresponding to the computing resource at multiple future time points, the initial demands for the computing resource by the IoT terminal at multiple future time points are first iteratively revised to generate revised demands for the computing resource by the IoT terminal at multiple future time points. For each IoT terminal, based on the revised requirements of a second similar IoT terminal for each computing resource at multiple future time points, the revised requirements of the IoT terminal for computing resources at multiple future time points are iteratively revised for the second time to generate future computing resource requirement data for the IoT terminal.

6. The method for collaborative allocation of edge computing resources for the Internet of Things according to claim 5, characterized in that: Based on the future computing resource demand data and computing resource demand association graph of multiple IoT terminals, the initial edge computing resource collaborative allocation plan is optimized in real time to generate a real-time edge computing resource collaborative allocation plan, including: Based on the future computing resource demand data of multiple IoT terminals and the initial edge computing resource collaborative allocation plan, determine the IoT terminals and edge computing nodes to be allocated; Through the Monte Carlo model, multiple candidate edge computing resource collaborative allocation schemes are generated based on the IoT terminals and edge computing nodes to be allocated. Establish optimization evaluation function; Based on the computing resource demand association graph, optimization evaluation function and multiple candidate edge computing resource collaborative allocation schemes, a real-time edge computing resource collaborative allocation scheme is generated.

7. The method for collaborative allocation of edge computing resources for the Internet of Things according to claim 6, characterized in that: Based on the computing resource demand association graph, the optimization evaluation function, and multiple candidate edge computing resource collaborative allocation schemes, a real-time edge computing resource collaborative allocation scheme is generated, including: Based on the computing resource demand association graph and the optimization evaluation function, the optimization value of each candidate edge computing resource collaborative allocation scheme is calculated; A real-time edge computing resource collaborative allocation scheme is generated based on the optimization value of each candidate edge computing resource collaborative allocation scheme through a genetic algorithm.

8. The edge computing resource collaborative allocation system for the Internet of Things is characterized by: The method for collaboratively allocating edge computing resources for the Internet of Things (IoT) according to any one of claims 1 to 7 comprises: A demand testing module is used to obtain test computing resource demand data of multiple IoT terminals, wherein the test computing resource demand data of the IoT terminals includes demands for multiple computing resources at multiple test time points; an association establishing module, for establishing an IoT terminal association graph of the plurality of IoT terminals based on the test computing resource demand data of the plurality of IoT terminals; An initial allocation module, configured to generate an initial edge computing resource collaborative allocation plan for multiple IoT terminals based on an IoT terminal association graph of the multiple IoT terminals; Demand acquisition module, used to obtain historical computing resource demand data of multiple IoT terminals; A real-time optimization module is used to optimize the initial edge computing resource collaborative allocation plan in real time based on the historical computing resource demand data of multiple IoT terminals and the IoT terminal association graph, and generate a real-time edge computing resource collaborative allocation plan; The collaborative allocation module is used to schedule computing resources for multiple edge computing nodes based on the real-time edge computing resource collaborative allocation scheme.

Images