Internet of Things edge server service caching method and device

Abstract

The invention provides an internet of things edge server service caching method and device, and the method comprises the steps: obtaining the caching information of each edge server corresponding to each piece of user equipment in the internet of things, obtaining the link state between the edge servers, and taking the caching information of each edge server as each node, constructing a topological relation graph by taking each link state as each edge; the method comprises the following steps: acquiring first task information of a current time step of each user equipment and second task information of at least part of time steps before the current time step, and generating a task sequence based on the first task information and the second task information, inputting the topological relation graph and the task sequence into a trained service cache prediction model to predict and obtain third task information of the next time step of each user equipment; and determining a service caching strategy based on the third task information corresponding to each user equipment and the caching information of each edge server. According to the method, the service caching efficiency and the system performance are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of Internet of Things devices, and in particular to a method and device for caching Internet of Things edge servers. Background Art

[0002] The Internet of Things (IoT) is a network form that connects various devices through the Internet to achieve data collection and intelligent management. In recent years, with the advancement of sensor technology and communication technology, the Internet of Things has rapidly become popular in smart homes, smart cities, and industrial automation. In industrial automation, the use of sensors and communication technology for real-time collection and control can monitor the status of equipment and detect faults in a timely manner, thereby ensuring the coordinated work of production line equipment and reducing downtime; this method can improve system response speed, enhance operational safety, and optimize resource allocation, ultimately bringing greater economic benefits and user experience.

[0003] With the spread of industrial automation, the number of sensors and other devices in the production environment is increasing, and the number of real-time acquisition and decision-making tasks generated is also increasing. In this case, the use of cloud computing is no longer appropriate; first, cloud computing often has a high latency. Data is transmitted from sensors to the cloud for processing, and then the results are fed back to the device, which may not meet the needs of real-time response; second, with the surge in the number of devices, the pressure on network bandwidth increases, and frequent data uploads will lead to bandwidth bottlenecks, increasing the cost of data transmission; in this case, edge computing has gradually become an important technology to meet the needs of real-time processing of massive data. Edge computing can sink part of the calculation to the edge server, and by processing data near the data source, the distance of data transmission is reduced, thereby significantly reducing the response time; at the same time, data processing and analysis tasks are performed at the edge, which can reduce the amount of data that needs to be sent to the cloud, saving bandwidth and transmission costs; edge computing is combined with cloud computing to support distributed architecture and better respond to the needs of devices and data in different geographical locations. Therefore, by offloading computing tasks from cloud servers to network edge devices for processing, edge computing can reduce data transmission delays, improve real-time performance and efficiency, and is particularly suitable for tasks with high requirements for real-time response.

[0004] At present, many scenarios (such as autonomous driving, industrial automation, and smart healthcare) have increasingly higher requirements for real-time response. At the same time, edge nodes usually have limited computing and storage resources. If the resource utilization rate is not high, the overall performance of the system will be reduced and the operating cost will be increased. In recent years, in order to minimize the completion time of task offloading and maximize the utilization rate of edge node resources, service caching technology has become one of the research hotspots. Service caching technology is an important module that cannot be missing in high-performance scenarios. An excellent service caching technology can greatly improve the efficiency of the entire system. Service caching technology can share some of the pressure for task offloading strategies by offloading computing services in advance, reducing the total delay of task completion; it can also avoid repeated uploading of cache services by continuously optimizing cache hits, effectively reducing network congestion and the total cost of the system.

[0005] Although the existing service caching methods can achieve service caching, there are a series of problems. For example, in the publicly available voice command task deployment method for IoT smart homes, the voice commands are first classified through federated learning, and then LSTM is used to predict the voice commands in order to minimize the computational load in the command reasoning process. The tasks corresponding to the predicted commands will be actively stored at the edge device. This method uses a neural network for prediction at a single node and caches the computing services of the prediction tasks in advance; when there are multiple user devices, this method is obviously more complicated and brings additional CPU and storage burdens to the system, which reduces system performance.

[0006] In another technical solution, in view of the shortcomings of the multi-region collaborative caching method in the above technical solution, a new collaborative caching algorithm is proposed. This caching method calculates the popularity of tasks based on the frequency of occurrence of tasks. The popularity will be updated with the arrival of tasks. The system will cache the most popular tasks each time, thereby realizing service caching based on this method. Although this service caching method takes into account the collaborative caching of computing services by nodes in a certain area, it lacks flexibility and cannot utilize the timing rules of task arrival, and has the disadvantage of low service caching efficiency.

[0007] In summary, the service caching method in the prior art has the disadvantages of low service caching efficiency and low system performance; therefore, how to improve the efficiency of service caching and system performance is a technical problem that needs to be solved urgently. Summary of the invention

[0008] In view of this, an embodiment of the present invention provides an Internet of Things edge server service caching method and device to eliminate or improve one or more defects in the prior art.

[0009] One aspect of the present invention provides an Internet of Things edge server service caching method, the method comprising:

[0010] Obtain cache information of each edge server corresponding to each user device in the Internet of Things, obtain the link status between each edge server, and construct a topological relationship diagram with the cache information of each edge server as each node and the link status as each edge;

[0011] Obtaining first task information of a current time step of each user device and second task information of at least part of the time steps before the current time step, generating a task sequence based on the first task information and the second task information, and inputting the topological relationship diagram and the task sequence into a trained service cache prediction model to predict and obtain third task information of the next time step of each user device;

[0012] A service cache strategy is determined based on the third task information corresponding to each of the user devices and the cache information of each edge server.

[0013] In some embodiments of the present invention, determining a service cache strategy based on the third task information corresponding to each user device and the cache information of each edge server includes:

[0014] Based on the third task information corresponding to each of the user devices and the cache information of each edge server, determining whether a service corresponding to the third task information is cached in the edge server;

[0015] When the service corresponding to the third task information is not cached in the edge server, the service corresponding to the third task information is cached in the corresponding edge server.

[0016] In some embodiments of the present invention, when the service corresponding to the third task information is not cached in the edge server, caching the service corresponding to the third task information in the corresponding edge server includes:

[0017] When no service corresponding to the third task information is cached in the edge server, determining whether the total number of services cached in the edge server is less than a threshold;

[0018] When the value is not less than the threshold, the infrequently used services cached in the edge server are identified and deleted.

[0019] In some embodiments of the present invention, infrequently used services cached in the edge server are determined and deleted based on the LRU algorithm.

[0020] In some embodiments of the present invention, the method further comprises:

[0021] Constructing a sample data set and a loss function, wherein the sample data in the sample data set includes topology relationship graph sample data, task sequence sample data, and user equipment task information sample data;

[0022] The initial service cache prediction model is pre-trained based on the sample data set and the loss function to obtain a trained service cache prediction model.

[0023] In some embodiments of the present invention, the loss function is:

[0024]

[0025] Among them, v l represents the real task information of the user device at the lth time step, v l+1 represents the real task information of the user device at the l+1th time step, represents the user equipment task information predicted by the model, W θ represents the model parameters and M represents the total number of time steps.

[0026] In some embodiments of the present invention, the service cache prediction model is a ST-GCN model.

[0027] According to another aspect of the present invention, an Internet of Things edge server service cache system is also disclosed, the system comprising a processor, a memory, and a computer program stored in the memory, the processor is used to execute the computer program, and when the computer program is executed, the system implements the steps of the method described in any of the above embodiments.

[0028] Correspondingly, the present invention further discloses a computer-readable storage medium on which a computer program is stored. When the computer program is executed by a processor, the steps of the method described in any of the above embodiments are implemented.

[0029] According to yet another aspect of the present invention, a computer program product is disclosed, including a computer program, which implements the steps of the method described in any of the above embodiments when executed by a processor.

[0030] The IoT edge server service caching method and device disclosed in the above-mentioned embodiment of the present invention use edge nodes and node arrival tasks as input information of a service cache prediction model, and obtain prediction results of subsequent arrival tasks of all nodes in the system through the service cache prediction model. This method can predict the task information of each user device in the next time step based on the cache prediction model, and then cache the services corresponding to each user device in the corresponding edge server respectively by pre-caching. This method not only improves the efficiency of service caching, reduces task completion delay, but also enhances system performance.

[0031] Additional advantages, purposes, and features of the present invention will be described in part in the following description, and will become apparent to those skilled in the art after studying the following, or may be learned from the practice of the present invention. The purposes and other advantages of the present invention may be achieved and obtained by the structures specifically indicated in the specification and the accompanying drawings.

[0032] Those skilled in the art will appreciate that the objectives and advantages that can be achieved with the present invention are not limited to the above specific description, and the above and other objectives that can be achieved by the present invention will be more clearly understood from the following detailed description. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] The drawings described herein are used to provide a further understanding of the present invention, constitute a part of this application, and do not constitute a limitation of the present invention. The components in the drawings are not drawn to scale, but are only for illustrating the principles of the present invention. In order to facilitate the illustration and description of some parts of the present invention, the corresponding parts in the drawings may be enlarged, that is, they may become larger relative to other components in the exemplary device actually manufactured according to the present invention. In the drawings:

[0034] Figure 1 A flowchart of an IoT edge server service caching method according to an embodiment of the present application.

[0035] Figure 2 This is a flow chart of an IoT edge server service caching method according to another embodiment of the present application. DETAILED DESCRIPTION

[0036] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the embodiments and the accompanying drawings. Here, the illustrative embodiments of the present invention and their descriptions are used to explain the present invention, but are not intended to limit the present invention.

[0037] It should also be noted that, in order to avoid obscuring the present invention due to unnecessary details, only structures and / or processing steps closely related to the solutions according to the present invention are shown in the accompanying drawings, while other details that are not closely related to the present invention are omitted.

[0038] It should be emphasized that the term “include / comprises” when used herein refers to the presence of features, elements, steps or components, but does not exclude the presence or addition of one or more other features, elements, steps or components.

[0039] It should also be noted here that, unless otherwise specified, the term "connection" in this article can refer not only to a direct connection, but also to an indirect connection with an intermediary, not only to a wired connection, but also to a wireless connection, and the specific connection can be changed based on the actual application scenario.

[0040] In this application, an IoT edge network is considered, which consists of n user devices, m edge devices and cloud servers. The user device set is represented by N = {1, 2, 3, ..., n}. The processing power of this type of device is weak and can only generate and transmit data and tasks. The number of edge servers is equal to the number of user devices. The edge servers are close to the corresponding user devices, and the communication delay with the user devices is small. The cloud server is represented by C, which has powerful processing power and storage capacity that is considered to be infinite. It caches all services and executes all tasks. However, the distance between the cloud server and the user device is far, which means that the communication between the user device and the cloud server will produce a large delay. In addition, the user device must communicate with the cloud server through the edge server.

[0041] Furthermore, there are q types of tasks in this scheme, which are generated by user devices; due to the limitation of their resources and low computing power, user devices cannot process these tasks and need to offload these tasks to edge devices for processing. In order to increase the speed of task processing, the edge server can process tasks by means of cache services; if there is a computing service corresponding to the task in the edge server, when the user device offloads the task to the edge device, it can only transmit the data required for the task without transmitting the computing service required for the task. The task can only be executed on the computing node that caches the corresponding service. In this application, a binary variable δ can be used. k , σ k Represents the deployment of the service. Indicates that the type of service k is q; Indicates that service k is deployed on node m, otherwise it is equal to 0. In addition, the storage and computing resources of the edge server are also limited. Assume that each edge server m has processing resources (CPU cycles) C(m) and storage size S m At the same time, when task v is offloaded to server m for execution, it consumes L per unit time. v CPU cycles, while occupying storage; when service k is cached on server m, it takes up storage storage.

[0042] Hereinafter, embodiments of the present invention will be described with reference to the accompanying drawings. In the accompanying drawings, the same reference numerals represent the same or similar components, or the same or similar steps.

[0043] Figure 1 This is a flow chart of an IoT edge server service caching method according to an embodiment of the present application, such as Figure 1 As shown, the method at least includes steps S10 to S30.

[0044] Step S10: Obtain cache information of each edge server corresponding to each user device in the Internet of Things, obtain the link status between each edge server, and construct a topological relationship diagram with the cache information of each edge server as each node and the link status as each edge.

[0045] In this step, the edge device topology is regarded as a topological relationship graph, so the present application can be regarded as a data set consisting of M topological relationship graphs at M time steps. Exemplarily, G is used to represent the spatiotemporal graph, and the graph at the tth time step can be mapped to a directed or undirected graph, which can be defined as G t =(V t , E, W); V t is a point set in the topological relationship graph, E represents an edge set, and W represents an adjacency matrix. In the topological relationship graph, the cache information of each edge server is used as each node, and each edge is determined by the link state between each edge server. In this embodiment, the number of edge servers is equal to the number of user devices in the Internet of Things, and the relationship between nodes and edges in the topological relationship graph corresponding to different time steps remains unchanged, and the adjacency matrix W also remains unchanged, but the characteristics of the node will change over time, and the characteristics of the node refer to the cache information of the node at the corresponding time step.

[0046] Step S20: Obtain the first task information of the current time step of each user device and the second task information of at least part of the time step before the current time step, generate a task sequence based on the first task information and the second task information, and input the topology relationship diagram and the task sequence into the trained service cache prediction model to predict the third task information of each user device in the next time step.

[0047] After obtaining the topology relationship diagram in step S10, the topology relationship diagram is further input into the trained service cache prediction model. In addition, the input information of the service cache prediction model includes not only the topology relationship diagram but also a task sequence composed of the task information generated by the user device at the current time step and the task information generated at least part of the time steps between the current time step. The output information of the service cache prediction model is the task information to be generated by the user device at the next time step predicted. It can be understood that "task information generated at least part of the time steps between the current time step" can be task information generated by all time steps before the current time step, or task information generated by part of all time steps, which is specifically determined based on the length of the task sequence.

[0048] Exemplarily, the service cache prediction model can specifically be an ST-GCN model, which is a deep learning model specifically used to process spatiotemporal data, especially in graph structured data (such as traffic networks, social networks, etc.). The ST-GCN model combines the characteristics of graph convolutional networks (GCN) and temporal convolutional networks (TCN), and can simultaneously capture the spatial and temporal dependencies of data; the structure of the ST-GCN model includes two spatiotemporal graph convolution blocks (ST-Conv Block) and an output fully connected layer (Output Layer), wherein the spatiotemporal graph convolution block consists of two time-gated convolutions and one spatial gated convolution. In this embodiment, the input information of the ST-GCN model is the feature vector of the graph corresponding to M time steps. The adjacency matrix W∈R corresponding to the graph n×n ; Where M is the time step, n is the number of nodes (the number of edge servers), c i is the number of channel features; in the ST-GCN model, the input information passes through two spatiotemporal graph convolution blocks and a fully connected layer, and the output That is, the output is the predicted features of each node corresponding to the M+1th time step. It can be understood that in the topological relationship diagram, the features of the nodes represent the cache information of the edge server.

[0049] In some embodiments of the present invention, in order to obtain a trained service cache prediction model, the Internet of Things edge server service cache method also includes the following steps: constructing a sample data set and a loss function, the sample data in the sample data set including topology relationship graph sample data, task sequence sample data, and user device task information sample data; based on the sample data set and the loss function, pre-training the initial service cache prediction model to obtain a trained service cache prediction model.

[0050] For example, when the ST-GCN model is used as the service cache prediction model, the ST-GCN model is trained so that it can accurately predict the arrival rule of tasks in the scenario of the present invention. As shown in Algorithm 1, the input data in the ST-GCN training algorithm is the training data The output data is the model parameter θ; during the training process, the neural network parameters and the counter value are first initialized (line 1 of Algorithm 1); then a batch of data is taken from the training data, the predicted value of the batch of data is predicted by ST-GCN, and then the loss function is calculated, and the gradient is calculated and the neural network parameters are updated. At the same time, the counter is incremented by one (lines 3 to 7 of Algorithm 1); all steps are looped to update the parameters until the counter reaches the set value. In Algorithm 1, the training data is the sample data, N represents the number of sample data in the sample data set, and X represents the number of sample data in the sample data set. (i) It can be understood as data including topology diagram sample data and task sequence sample data, and Y (i)It can be understood as user equipment task information sample data.

[0051]

[0052] Furthermore, the service cache prediction model uses L2 loss as the loss function. Specifically, the model loss function can be:

[0053]

[0054] Among them, v l represents the real task information of the user device at the lth time step, v l+1 represents the real task information of the user device at the l+1th time step, represents the user equipment task information predicted by the model, W θ represents the trainable model parameters and M represents the total number of time steps.

[0055] Step S30: determining a service cache strategy based on the third task information corresponding to each of the user devices and the cache information of each edge server.

[0056] In this step, the cache strategy of the edge server is further determined based on the third task information corresponding to each user equipment predicted by the service cache prediction model.

[0057] In one embodiment, the service cache strategy is determined based on the third task information corresponding to each user device and the cache information of each edge server, including the following steps: judging whether the edge server has a service corresponding to the third task information cached therein based on the third task information corresponding to each user device and the cache information of each edge server; when the edge server has not a service corresponding to the third task information cached therein, the service corresponding to the third task information is cached in the corresponding edge server. In this embodiment, the third task information corresponding to each user device obtained based on the prediction is compared with the cache information of each edge server corresponding to the current time step to judge whether the edge server has a service corresponding to the task generated by the user device in the next time step cached therein; when the edge server has a service corresponding to the task generated by the user device in the next time step cached therein, the edge server does not need to re-cache the predicted service; and if the edge server has not a service corresponding to the task generated by the user device in the next time step cached therein, the predicted service needs to be cached on the corresponding edge server.

[0058] Furthermore, when the service corresponding to the third task information is not cached in the edge server, the service corresponding to the third task information is cached in the corresponding edge server, including: when the service corresponding to the third task information is not cached in the edge server, it is determined whether the total number of services cached in the edge server is less than a threshold; when it is not less than the threshold, the infrequently used services cached in the edge server are determined and deleted. In this embodiment, for the edge server that needs to cache the service, it is determined whether it has cache space. If there is cache space, the corresponding service can be directly cached in the edge server. If there is no cache space in the cache server, cache space can be provided for the service that needs to be cached by deleting the infrequently used service. Optionally, when deleting the infrequently used service in the edge server, the infrequently used service cached in the edge server can be determined and deleted based on the LRU algorithm; the LRU algorithm improves the cache hit rate by retaining the most recently used data and eliminating the longest unused data.

[0059] Figure 2 This is a flow chart of an IoT edge server service caching method according to another embodiment of the present application, such as Figure 2 As shown, this embodiment predicts the third task information of each user device at the next time step based on the trained ST-GCN model, and performs cache calculation service according to the prediction result of the ST-GCN model. Specifically, this embodiment includes the following steps: S01: When each time step arrives, if the input sequence length is l, firstly convert the task information D of the current time step l t and the task information of the previous l-1 time steps are combined into the input sequence d t ; Step S02: d t Input into ST-GCN to obtain the predicted value of the task data at the l+1 time step; Step S03: based on the predicted value of the task r of node i at the l+1 time step i and the cache status of the node at time step l c i Perform computing service cache. If the current node has cached task r i If the current node does not have a computing service for the task and the cache is full, the least frequently used computing service is replaced using the LRU rule; if the cache is not full, task r is directly i The compute service is added to the cache.

[0060] For example, Algorithm 2 is a computing service cache algorithm based on ST-GCN. In Algorithm 2, the input data is model M (specifically, it can be a pre-trained ST-GCN model), time series data D, and node cache information {c1, c2, ..., c k}, node cache size C, current time step l; the output data is the updated node cache information {c1, c2, ..., c k It can be understood that the time series data D input in the algorithm is composed of the task information D at the current time step l. t The node cache information {c1, c2, ..., c k} is the cache information of the edge server corresponding to the current time step, and the node cache size C is the cache space C(m) of each edge server. The second line in Algorithm 2 represents the construction of the input sequence d t ; Line 3 indicates that d t Input into ST-GCN to obtain the predicted value of the task data at time step l+1; Lines 4 to 6 indicate that if the current node has cached task r i The computing service of the task is processed and the next node is processed; the 8th and 9th lines indicate that the current node does not have the computing service of the task and the cache is full, and the LRU rule is used to replace the least frequently used computing service; the 10th line indicates that the cache is not full, and the task r is replaced. i The compute service is added to the cache.

[0061]

[0062] In the IoT edge server service caching method of the above embodiment, the edge nodes and node arrival tasks are abstracted as the input of ST-GCN, and then the prediction results of all node subsequent arrival tasks are obtained through ST-GCN. In this way, a neural network model can be used to predict the task status of all nodes, and then the computing service of the predicted task is cached in advance to reduce the task completion delay and enhance the overall performance of the system. In addition, the service caching method not only effectively reduces the cloud server load, but also reduces the network delay by coordinating cloud servers and edge servers.

[0063] Correspondingly, the present invention also provides an Internet of Things edge server service cache system, the system comprising a processor, a memory and a computer program stored in the memory, the processor is used to execute the computer program, and when the computer program is executed, the system implements the steps of the method described in any of the above embodiments.

[0064] Through the above embodiments, it can be found that the present application combines edge computing and cloud computing, introduces computing service cache under the premise of limited node resources, and multiple edge servers of the present application only use one neural network model to predict the tasks of all nodes in the system, saving CPU and storage resources of nodes in the system. At the same time, the corresponding computing services can be cached in advance according to the prediction results, reducing the delay in task completion.

[0065] In summary, the IoT edge server service caching method of the present application combines the characteristics of edge computing and cloud computing scenarios, and considers the computing service caching of real-time acquisition decision-making tasks in a network structure with three types of devices: IoT devices, edge servers, and cloud servers; the present application does not simply divide the tasks into two types: edge and cloud deployment, but divides the tasks themselves into two parts: computing services and corresponding data, and reduces the task completion delay by caching computing services in advance, making subsequent task offloading more flexible and more in line with the actual production environment. Specifically, the ST-GCN-based service caching method of the present application abstracts the node topology and incoming tasks into a spatiotemporal graph when making predictions, and can predict the arrival of tasks for all nodes through only one ST-GCN model, solving the problem of deploying a neural network for each node for prediction in current research.

[0066] The embodiments of the present invention also provide a computer-readable storage medium and a computer program product, on which a computer program is stored, and when the computer program is executed by a processor, the steps of the method described in any of the above embodiments are implemented. The computer-readable storage medium can be a tangible storage medium, such as a random access memory (RAM), a memory, a read-only memory (ROM), an electrically programmable ROM, an electrically erasable programmable ROM, a register, a floppy disk, a hard disk, a removable storage disk, a CD-ROM, or any other form of storage medium known in the technical field.

[0067] It should be understood by those skilled in the art that the exemplary components, systems and methods described in conjunction with the embodiments disclosed herein can be implemented in hardware, software or a combination of the two. Whether it is performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of the present invention. When implemented in hardware, it can be, for example, an electronic circuit, an application specific integrated circuit (ASIC), appropriate firmware, a plug-in, a function card, etc. When implemented in software, the elements of the present invention are programs or code segments used to perform the required tasks. The program or code segment can be stored in a machine-readable medium, or transmitted on a transmission medium or a communication link via a data signal carried in a carrier.

[0068] It should be clear that the present invention is not limited to the specific configuration and processing described above and shown in the figures. For the sake of simplicity, a detailed description of the known method is omitted here. In the above embodiments, several specific steps are described and shown as examples. However, the method process of the present invention is not limited to the specific steps described and shown, and those skilled in the art can make various changes, modifications and additions, or change the order between the steps after understanding the spirit of the present invention.

[0069] In the present invention, features described and / or illustrated for one embodiment may be used in the same or similar manner in one or more other embodiments, and / or combined with features of other embodiments or replace features of other embodiments.

[0070] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the embodiments of the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A method for caching services of an edge server of the Internet of Things, characterized in that: The method comprises: Obtain cache information of each edge server corresponding to each user device in the Internet of Things, obtain the link status between each edge server, and construct a topological relationship diagram with the cache information of each edge server as each node and the link status as each edge; Obtaining first task information of a current time step of each user device and second task information of at least part of the time steps before the current time step, generating a task sequence based on the first task information and the second task information, and inputting the topological relationship diagram and the task sequence into a trained service cache prediction model to predict and obtain third task information of the next time step of each user device; A service cache strategy is determined based on the third task information corresponding to each of the user devices and the cache information of each edge server.

2. The IoT edge server service caching method according to claim 1, characterized in that: Determining a service cache strategy based on the third task information corresponding to each of the user devices and the cache information of each edge server includes: Based on the third task information corresponding to each of the user devices and the cache information of each edge server, determining whether a service corresponding to the third task information is cached in the edge server; When the service corresponding to the third task information is not cached in the edge server, the service corresponding to the third task information is cached in the corresponding edge server.

3. The IoT edge server service caching method according to claim 2, characterized in that: When the service corresponding to the third task information is not cached in the edge server, caching the service corresponding to the third task information in the corresponding edge server includes: When no service corresponding to the third task information is cached in the edge server, determining whether the total number of services cached in the edge server is less than a threshold; When the value is not less than the threshold, the infrequently used services cached in the edge server are identified and deleted.

4. The IoT edge server service caching method according to claim 3, characterized in that: Based on the LRU algorithm, the infrequently used services cached in the edge server are determined and deleted.

5. The IoT edge server service caching method according to claim 1, characterized in that: The method further comprises: Constructing a sample data set and a loss function, wherein the sample data in the sample data set includes topology relationship graph sample data, task sequence sample data, and user equipment task information sample data; The initial service cache prediction model is pre-trained based on the sample data set and the loss function to obtain a trained service cache prediction model.

6. The IoT edge server service caching method according to claim 5, characterized in that: The loss function is: Among them, v l represents the real task information of the user device at the lth time step, v l+1 represents the real task information of the user device at the l+1th time step, represents the user equipment task information predicted by the model, W θ represents the model parameters and M represents the total number of time steps.

7. The IoT edge server service caching method according to any one of claims 1 to 6, characterized in that: The service cache prediction model is the ST-GCN model.

8. An Internet of Things edge server service cache system, the system comprising a processor, a memory and a computer program stored in the memory, characterized in that: The processor is used to execute the computer program. When the computer program is executed, the system implements the steps of the method according to any one of claims 1 to 7.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

10. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

Images