Task offloading and virtual machine placement method in edge computing based on matching game

Abstract

The application discloses a task unloading and virtual machine placement method in edge computing based on matching game, which is based on matching game between edge servers and industrial internet time sequence task flow, and comprises the following steps: establishing an industrial internet network platform with edge computing capability, taking the balance between total completion time and total energy consumption as the target of the system, and jointly determining the optimal virtual machine deployment and task unloading decision in the time-sensitive time sequence task flow of the industrial internet equipment while meeting the storage capacity of the edge server, the completion time of the time sequence task flow and the deployment demand of the virtual machine. The method determines the average completion time of the industrial internet time sequence application, the total energy consumption of the system, the task unloading of the time sequence task flow and the deployment of the virtual machine through joint optimization.

Description

Technical Field

[0001] This invention pertains to network resource management technology, specifically relating to the offloading of collaborative computing tasks and placement of virtual machines in industrial internet systems with edge computing capabilities, and particularly to a method for task offloading and virtual machine placement in edge computing based on matching game theory. Background Technology

[0002] Today, the Internet of Things (IoT) is widely used in many existing industrial systems, such as transportation and manufacturing systems, and is further referred to as the Industrial Internet. Due to the generally limited computing power of Industrial Internet devices, edge computing has been proposed to alleviate this computational burden. Compared to remote cloud environments, edge servers are closer to Industrial Internet devices, thus enabling lower communication latency. In Industrial IoT with edge computing capabilities, how to schedule Industrial Internet task offloading and deploy virtual machines on edge servers is a key issue. Given the constraints of limited resources for Industrial Internet devices and application deadlines, finding suitable task offloading and virtual machine placement schemes to reduce computation time and energy consumption is crucial. While the combination of edge computing and the Industrial Internet brings many benefits, some issues closely related to practical implementation remain unresolved:

[0003] 1) In practice, an Industrial Internet application requires multiple operations to complete. That is, an application consists of multiple tasks. For example, opening an automatic door involves three operations (tasks): a sensor detects someone approaching the door, the controller controls the motor to run, and the motor opens the door via a toothed belt. However, these tasks are interdependent. The controller only knows that someone is approaching the door and needs to control the motor to open it when the sensor sends an opening signal. This is a system control method with feedback information, where the controlled output returns to the controlled input in some way and has a control effect on the input. This prompts us to consider the relationships between tasks from the same Industrial Internet application when building system models.

[0004] 2) For time-sensitive industrial internet applications, such as the opening and closing of automatic doors, the entire operation needs to be completed within a limited time to ensure that people do not wait in front of the door for a long time. This requires a fast-response control system with long-term performance guarantees.

[0005] 3) Most importantly, human / device intelligence allows Industrial Internet (IIoT) devices to engage in strategic and self-serving behaviors, even though they are rational. Clearly, IIoT devices strive for their own benefit when making offloading decisions. However, this behavior can lead to network congestion and computation time delays, impacting overall system performance. Furthermore, there is a growing trend towards deploying virtual machines (VMs) on edge networks to facilitate latency-sensitive IIoT applications. Therefore, due to the capacity limitations of edge servers, computation offloading decisions and VRC deployment need to be considered together to maximize overall network performance.

[0006] However, considering all the factors mentioned above, implementing this industrial internet system with edge computing capabilities will encounter certain difficulties for the following reasons:

[0007] a) Tasks in the time-series task flow of industrial internet devices are dependent on each other, and all tasks have time constraints for completion. Therefore, we need to consider not only whether previous tasks in the time-series task flow have been completed, but also that all tasks must be completed on time. This requires a good task scheduling solution.

[0008] b. Since both compute offloading decisions and virtual machine deployments contribute to the overall performance of the edge computing system, they must be jointly optimized to strike a balance between system completion time and energy consumption, while ensuring that all industrial internet devices and edge servers are satisfied with their decisions (i.e., their utility never decreases). This clearly presents an NP-hard problem, which is very difficult to solve.

[0009] c. Since multi-server edge computing includes compute offloading, compute task transfer, and virtual machine deployment, the design of this system needs to consider all design requirements or constraints, such as the capacity of the edge servers, the completion deadline of the time-series task flow, and the deployment conditions of virtual machines on the edge servers, which further increases the complexity of task allocation. Summary of the Invention

[0010] Purpose of the invention: To address the shortcomings in industrial internet device application analysis and resource optimization in the existing technologies, this invention provides a method for task offloading and virtual machine placement in edge computing based on matching game theory.

[0011] Technical Solution: A method for task offloading and virtual machine placement in edge computing based on matching game theory. The method establishes an industrial internet network system with edge computing capabilities based on matching game theory between edge servers and industrial internet time-series task flows. With the goal of balancing total completion time and total energy consumption, the system jointly determines the optimal virtual machine deployment and task offloading decisions in the time-sensitive time-series task flows of industrial internet devices, while meeting the storage capacity of edge servers, the completion time of time-series task flows, and the deployment requirements of virtual machines.

[0012] The method includes the following steps:

[0013] (1) Build an industrial internet system with edge computing capabilities, including an industrial internet device layer, an edge server layer and a cloud layer. The industrial internet device layer deploys industrial internet device groups with time-sequenced task flows, the edge server layer deploys edge servers, and the cloud layer deploys virtual machines.

[0014] (2) Establish a task offloading and virtual machine placement model based on the industrial internet system, including constructing system functions to control the offloading of computing tasks;

[0015] The task unloading and virtual machine placement include the following processes:

[0016] (a) The task set consisting of the kth task of all time-series task flows Each task in the process checks its own completion deadline; if the task's current completion time is less than or equal to the completion deadline of the task flow to which it belongs, then... j When the difference reaches the set warning value, the task is unloaded and the calculation is performed in the cloud;

[0017] (b) The remaining tasks in the set are determined according to the benefit function U. j，k (i, S) i \N{s j，k}) Calculate preference values ​​for edge servers to construct a preference table; edge servers are configured according to the benefit function U i (j, k) Calculate the preference values ​​for the task to construct a preference table;

[0018] (c) Each task sends a request to the top-ranked edge server based on its own preference table;

[0019] (d) The edge server that receives the request accepts a maximum of n tasks based on its own preference table, while rejecting other task requests;

[0020] (e) Exchange matching process: The task selected by the edge server checks whether there is an exchange blocking pair; if so, the matching status is updated to... Otherwise, maintain the existing matching state μ; repeat this step until there are no more swap blocking pairs;

[0021] (f) Virtual machine deployment phase: The edge server matched with the task checks whether it meets the virtual machine deployment requirements based on the type of virtual machine required by the task; if it does, the task is processed and the corresponding type of virtual machine is deployed; otherwise, the task is rejected and added to the next round of matching.

[0022] (g) Repeat steps (a)-(g) until task set L k All tasks in the process are processed;

[0023] (h) Task set Once finished, proceed to the next task set. And so on.

[0024] Furthermore, in step (1), each industrial internet device has a time-sensitive sequential task flow, consisting of multiple tasks, and the types of tasks are different, requiring different types of virtual machines. Sequential means that because the output of the previous task is used as the input of the next task, the next task can only begin execution after the previous task has finished. Therefore, the start time of a task in a sequential task flow is the time when the previous task completes its computation; it should be noted that when the unloading targets of two tasks in a sequential task flow are the same, the next task can begin transmission only after the previous task has completed its unloading and transmission. The start time of the computation task execution on the local device. On the edge server Start time of execution and the start time of execution in the cloud. The formulas are expressed as follows:

[0025]

[0026]

[0027]

[0028] Among them, computational task s j，k From industrial internet devices Fully uploaded to the target edge server time

[0029]

[0030] Computational tasks s j，k From industrial internet devices Time to fully upload to the cloud

[0031]

[0032] In the formula, Represents the time-series task flow w j Tasks j，k-1 Type of virtual machine required Is it deployed on an edge server? superior, They represent tasks s respectively j，k-1 Is it on the edge server? Execution, whether it is executed in the cloud, or whether it is executed on a local device. Execute on, They represent tasks s respectively j，k-1 On the edge server Cloud and local devices Completion time on They represent computational tasks s respectively. j，k Transmission time from the local device layer to the edge server layer, and from the edge server layer to the cloud layer.

[0033] Furthermore, in step (1) of the industrial internet system with edge computing capabilities, the industrial internet devices offload computing tasks to edge servers for execution, requiring the deployment of corresponding virtual machines on those edge servers. Each edge server has a certain amount of storage space z. i ;

[0034] The total storage space of virtual machines deployed on each edge server cannot exceed the total storage space of that edge server.

[0035]

[0036] In the formula: Represents the time-series task flow w j Tasks j，k The type of virtual machine required Is it deployed on an edge server? superior.

[0037] Furthermore, in step (2), the task completion time of the time-series task flow is the time when the task finishes its computation, including the transmission time for task unloading, the computation time, and the waiting time. Corresponding to the start time, the completion time of the computation task executed on the local device is... On the edge server Completion time of execution And the completion time when executed in the cloud. The formulas are expressed as follows:

[0038]

[0039]

[0040]

[0041] Among them, computational task s j，k Local device On the calculation time

[0042]

[0043] Computational tasks s j，k On the edge server On the calculation time

[0044]

[0045] Computational tasks s j，k Computing time in the cloud

[0046]

[0047] In the formula: c j，k This indicates the completion of a task s. j，k Number of CPU cycles required f C These represent industrial internet devices. Edge server CPU cycles per time unit in the cloud, S i Indicates assignment to edge server A set of tasks.

[0048] Furthermore, in step (b), the task constructs a benefit function U for the preference table of edge servers. j，k (i, S) i \{s j，k The benefit function U for constructing a preference table for tasks by the edge server. i (j, k) represent the following respectively:

[0049]

[0050]

[0051] Among them, edge servers Execution time sequence task flow w j Tasks j，k Energy consumption The expression is:

[0052]

[0053] In the formula: Represents the time-series task flow w j Tasks j，k The type of virtual machine required Is it deployed on an edge server? superior, Indicates task s j，k Is it on the edge server? implement, Represents type m j，k Tasks j，k From the edge server connected to the local device Transmitted to the target edge server Transmission energy consumption, Indicates task s j，k On the edge server Computational energy consumption, Indicates task s j，k From the equipment To and equipment The transmission energy consumption of the connected edge servers.

[0054] Furthermore, in step (f), tasks rejected by the edge server need to update their preference tables by updating the waiting time before being added to the next round of matching. The preference value depends not only on the assigned edge server but also on other tasks assigned to that edge server. This type of matching is defined as externality matching in the prior art.

[0055] Beneficial effects: Compared with the prior art, the substantial progress and significant effects of the present invention are as follows:

[0056] 1) This invention takes into account the strict time requirements for the completion of industrial internet device applications, dividing them into multiple tasks. These tasks are dependent and sequential, meaning that the next task can only be executed after the previous task is completed, and the output of the previous task serves as the input for the next task. To further meet the time constraints, we propose an industrial internet system with edge computing capabilities.

[0057] 2) The method described in this invention enables joint task offloading and virtual machine placement in edge computing for time-series industrial internet applications based on matching game theory, forming an industrial internet system with edge computing capabilities. By establishing a matching game with externalities between the task set and the edge server set of the time-series task flow, the system achieves a stable matching state, thereby achieving a balance between system completion time and energy consumption. Externalities mean that the preference values ​​of the task players participating in the matching process are influenced not only by the matched edge server but also by other task players matched to that edge server. Attached Figure Description

[0058] Figure 1 This is a schematic diagram of an industrial internet system structure with edge computing capabilities constructed by the method described in this invention.

[0059] Figure 2 The figure shows a comparison of the average completion time of the embodiments and the prior art with different numbers of industrial IoT devices;

[0060] Figure 3 The diagram shows a comparison of the energy consumption of the embodiment and existing technologies with different numbers of industrial IoT devices. Detailed Implementation

[0061] To illustrate the technical solutions disclosed in this invention in detail, further explanation will be provided below with reference to the accompanying drawings and specific embodiments.

[0062] To address the severe network congestion caused by forwarding large amounts of data to remote cloud for processing, which prevents time-sensitive industrial internet applications from being processed within the effective time and hinders system optimization, this invention provides a task offloading and virtual machine placement method in edge computing based on matching game theory. This method is based on matching game theory between edge servers and industrial internet time-series task flows. It includes establishing an industrial internet network platform with edge computing capabilities, aiming to balance the system's total completion time and total energy consumption. While meeting the storage capacity of edge servers, the completion time of time-series task flows, and the deployment requirements of virtual machines, it jointly determines the optimal virtual machine deployment and task offloading decisions in the time-sensitive time-series task flows of industrial internet devices.

[0063] In the method described in this invention, the method is based on matching game theory to achieve joint task offloading and virtual machine placement in edge computing for time-series industrial internet applications. Unlike existing technologies, this invention considers the processing method of time-series task flows in edge computing for industrial internet applications, and also takes into account the interests of both tasks and edge servers within the time-series task flow. To achieve a balance between system completion time and energy consumption, this invention constructs a joint optimization to determine the average completion time of industrial internet time-series applications, the total energy consumption of the system, task offloading in the time-series task flow, and virtual machine deployment.

[0064] Based on the aforementioned technical solution and existing technologies well-known to those skilled in the art, further, combining... Figure 1 The implementation process of the method described in this invention is explained in detail below.

[0065] Step 1: Build an industrial internet system with edge computing capabilities and construct the function model of the system.

[0066] This step mathematically analyzes the practical problems addressed by the joint task offloading and virtual machine placement methods in edge computing for time-series industrial internet applications.

[0067] Based on actual application scenarios, an industrial internet system with edge computing capabilities consists of an industrial internet device layer, an edge server layer, and a cloud layer. The industrial internet device layer comprises a set of industrial internet devices running latency-sensitive applications. (Using a collection...) The cardinality of a set. The edge server layer consists of a set of edge servers with heterogeneous processing power and storage capacity. (Using a collection...) The cardinality of a set. A set of virtual machines holding the target application is stored in the cloud. (Using a collection) The cardinality of a set.

[0068] In the set In the middle, each industrial internet device Each industrial internet device (IIoT) employs a time-sensitive, sequential task flow, comprised of multiple distinct tasks that require different types of virtual machines for processing. The term "sequential" refers to the fact that the output of one task is used as the input for the next, meaning the next task can only begin execution after the previous one has finished. While all IIoT devices can offload computationally intensive tasks, reducing their computational burden, this introduces challenges related to virtual machine deployment.

[0069] For each edge server Industrial Internet devices offload computing tasks to edge servers, requiring the deployment of corresponding virtual machines on those servers. Each edge server has a certain amount of storage space. i Therefore, the total storage space of virtual machines deployed on each edge server cannot exceed the total storage space of that edge server. in, Represents the time-series task flow w j Tasks j，k The type of virtual machine required Is it deployed on an edge server? superior.

[0070] In the aforementioned industrial internet system with edge computing capabilities, in the joint task offloading and virtual machine placement model for edge computing oriented towards time-series industrial internet applications, the start time of a task in a time-series task flow is defined as the time when the previous task completes its computation. It's important to note that when two tasks in a time-series task flow have the same offloading target, the next task can begin transmission only after the previous task has completed its offloading and transmission. The start time of the computation task execution on the local device is also considered. On the edge server Start time of execution and the start time of execution in the cloud. The formulas are expressed as follows:

[0071]

[0072]

[0073]

[0074] Among them, computational task s j，k From industrial internet devices Fully uploaded to the target edge server time

[0075]

[0076] Computational tasks s j，k From industrial internet devices Time to fully upload to the cloud

[0077]

[0078] In the formula, Represents the time-series task flow w j Tasks j，k-1 The type of virtual machine required Is it deployed on an edge server? superior, They represent tasks s respectively j，k-1 Is it on the edge server? Execution, whether it is executed in the cloud, or whether it is executed on a local device. Execute on, They represent tasks s respectively j，k-1 On the edge server Cloud and local devices Completion time on They represent computational tasks s respectively. j，k The transmission time from the local device layer to the edge server layer, and from the edge server layer to the cloud layer. Correspondingly, the transmission of computing tasks also incurs certain energy consumption. Here, Indicates task s j，k From the equipment To and equipment The transmission power consumption of the connected edge servers, Indicates task s j，kEnergy consumption for transmission from the edge server layer to the cloud. The type is m j，k Tasks j，k From the edge server connected to the local device Transmitted to the target edge server The transmission energy consumption is expressed by the following formula:

[0079]

[0080] in, Indicates in network link e i Transmit a type m on ∈ε j，k Tasks j，k The energy consumed is from the associated edge server. To the target edge server The path consists of multiple network links. Indicates link e i Whether it is on the corresponding path. Here, ε represents the set of links between edge servers.

[0081] Furthermore, a task in a time-sequential task flow has a start time and a corresponding end time. Task s j，k The end time is defined as the time when the task completes, including the waiting time required for all tasks previously assigned to that edge server to complete execution. The completion time of the task executed on the local device is calculated. On the edge server Completion time of execution And the completion time when executed in the cloud. The formulas are expressed as follows:

[0082]

[0083]

[0084]

[0085] Among them, computational task s j，k Local device On the calculation time

[0086]

[0087] Computational tasks s j，k On the edge server On the calculation time

[0088]

[0089] Computational tasks s j，k Computing time in the cloud

[0090]

[0091] In the formula: c j，k This indicates the completion of a task s. j，k Number of CPU cycles required f C These represent industrial internet devices. Edge server CPU cycles per time unit in the cloud, S i Indicates assignment to edge server A set of tasks.

[0092] Correspondingly, tasks in a time-series task flow reside on the local device. On the edge server The mathematical expressions for computing energy consumption in the cloud are as follows:

[0093]

[0094]

[0095]

[0096] In the formula: and P C These represent local devices. Edge server And the effective capacitance parameters in the cloud; c j，k This indicates the completion of a task s. j，k Number of CPU cycles required f C These represent industrial internet devices. Edge server CPU cycles per time unit in the cloud.

[0097] In summary, a time-series task flow w can be derived. j Completion time and energy consumption The mathematical expressions for are as follows:

[0098]

[0099]

[0100] Among them, |w j | Represents a time-series task flow w jThe last task in the series.

[0101] Next, we consider the balance between system completion time and energy consumption on a system-wide scale, and attach all relevant constraints. Ultimately, we can construct a joint task offloading and virtual machine placement problem in edge computing for time-series industrial internet applications:

[0102]

[0103] Constraints:

[0104]

[0105]

[0106]

[0107]

[0108]

[0109]

[0110]

[0111]

[0112]

[0113]

[0114]

[0115]

[0116]

[0117]

[0118] In the above formula: |w j | Represents a time-series task flow w j The last task in the series.

[0119] Step 2: Solve using the matching game method

[0120] First, we define two sets as matching objects, and allocate computing resources to tasks according to the time order in the task flow. Let... Let represent the set of k tasks in all time-series task flows. This indicates the maximum length of the task flow. We first determine... The unloading plan will be determined later. And so on. The definition of a match is as follows:

[0121] A matching μ is a set arrive A function of the set of all subsets, such that for all... as well as 1) Head If tasks s j，k No matching of any edge servers; 2) If edge server i does not match any task, 3)s j，k ∈μ(i) if and only if i∈μ(s) j，k ).

[0122] Based on the benefit function value, each selfish device will have its own list of preferred tasks. The benefit function for task distribution to edge servers is... Note that the preference value depends not only on the assigned edge server, but also on other tasks assigned to that edge server. This type of matching is defined in the literature as externality matching.

[0123] Each edge server is also selfish, preferring lower energy consumption. Therefore, edge servers are more suited to specific tasks. j，k The utility function is in, It is an edge server Execution time sequence task flow w j Tasks j，k The energy consumed.

[0124] The method described in this invention supports swapping operations. This means that... Any two tasks in the set can swap their edge servers, and The matching of other tasks in the set with the edge server remains unchanged. The definition of swap matching is as follows:

[0125]

[0126] Where i = μ(s) j，k ), i' = μ(s) j‘，k ).

[0127] Furthermore, this invention also relates to the concept of exchange-blocking pairs. Defined as follows:

[0128] (s j，k s j′，k A pair is a swap match if and only if 1) and 2)

[0129] According to the definition mentioned above, a swap operation can only occur when there is a swap blocking pair. Condition (1) means that after the swap operation, the utility of all participants should not decrease, and condition (2) means that after the swap operation, the utility of at least one participant increases.

[0130] Finally, a stable matching state of the system is obtained, which is the unique solution to the problem of this invention after optimization. If the pairing does not contain any exchange-blocking pairs, then the pairing is a bilateral transactional stable pairing. When there are no exchange-blocking pairs, the matching result is satisfactory in the set of tasks and the set of edge servers involved. In other words, their benefit value does not increase, which violates the second term in the definition of an exchange-matching pair. Therefore, when there are no exchange-blocking pairs, the final pairing result is a bilateral transactional stable pairing.

[0131] like Figure 2 As shown, for all schemes, the average completion time (ACT) increases with the number of devices because more time-series task flows need to be processed by the edge servers, resulting in longer waiting times. Furthermore, SEHPA has the highest ACT because each virtual machine can be deployed on a maximum of two edge servers, and all tasks requiring that virtual machine are offloaded to these edge servers, potentially leading to long task waiting times. In addition, CDGA outperforms SEHPA because tasks requiring that virtual machine can be distributed to multiple available edge servers, with relatively less workload on each. MTA performs worse than MGVDA because each edge server can accept multiple proposals per round, resulting in longer waiting times and wasted computational resources. Our proposed matching game algorithm, MGVDA, effectively addresses the limitations of the other three algorithms, achieving the best performance.

[0132] For example Figure 3 As shown, system energy consumption (EC) increases with the number of Industrial Internet devices. Naturally, processing more sequential task flows with multiple tasks requires more energy. Furthermore, we can see that when the number of Industrial Internet devices is small (less than 300), SEHPA, CDGA, and MTA perform similarly in minimizing system EC. However, when the number of Industrial Internet devices is large, the system EC of CDGA and SEHPA increases significantly. This is because... Figure 2 Higher ACT (Active Task Response) in the middle of the timeframe leads to more task streams expiring soon, resulting in more tasks being sent to the cloud. Cloud computing is more powerful than edge computing, but it consumes more energy. Furthermore, because this algorithm effectively balances the energy consumption of local computing, edge computing, and cloud computing on the system's energy efficiency (EC), its performance surpasses other algorithms.

[0133] Combination Figure 2 , Figure 3 It can be seen that the superior performance of the matching game described in this invention has been verified.

Claims

1. A method for task offloading and virtual machine placement in edge computing based on matching game theory, characterized in that: The method is based on the matching game between edge servers and industrial internet time-series task flows to establish an industrial internet network system with edge computing capabilities. The system aims to balance the total completion time and total energy consumption. While meeting the storage capacity of edge servers, the completion time of time-series task flows, and the virtual machine deployment requirements, it jointly determines the optimal virtual machine deployment and task offloading decisions in the time-sensitive time-series task flows of industrial internet devices. The method includes the following steps: (1) Build an industrial internet system with edge computing capabilities, including an industrial internet device layer, an edge server layer and a cloud layer. The industrial internet device layer deploys industrial internet device groups with time-series task flows, the edge server layer deploys edge servers, and the cloud layer deploys virtual machines. In step (1), each industrial internet device has a time-sensitive sequential task flow, which consists of more than one task. The types of tasks are different, and the types of virtual machines required are also different. Set the start time for the computing task to execute on the local device. On edge servers Start time of execution and the start time of execution in the cloud. The formulas are expressed as follows: ； ； Among them, computing tasks From industrial internet devices Fully uploaded to the target edge server time : Computational tasks From industrial internet devices Time to fully upload to the cloud : In the formula, Representing a time-series task flow Task Type of virtual machine required Is it deployed on an edge server? superior, Representing tasks Is it on the edge server? Execution, whether it is executed in the cloud, or whether it is executed on a local device. Execute on, , , Representing tasks On the edge server Cloud and local devices Completion time on , Representing computational tasks Transmission time from the local device layer to the edge server layer, and from the edge server layer to the cloud layer; (2) Establish a task offloading and virtual machine placement model based on the industrial internet system, including constructing system functions to control the offloading of computing tasks; The task unloading and virtual machine placement include the following processes: (a) The first of all time-series task flows A task set consisting of [number] tasks Each task in the process checks its own completion deadline; if the task's current completion time is less than the completion deadline of the task sequence to which it belongs, then... When the difference reaches the set warning value, the task is unloaded and the calculation is performed in the cloud; (b) The remaining tasks in the set are determined according to the benefit function. Calculate preference values ​​for edge servers to construct a preference table; edge servers are then configured according to a benefit function. Calculate preference values ​​for tasks to construct a preference table; In step (b), the task constructs a benefit function for the preference table of edge servers. Benefit function for constructing a preference table for tasks with edge servers They are represented as follows: ， Among them, edge servers Execution sequence task flow Task Energy consumption The expression is: In the formula: Representing a time-series task flow Task Type of virtual machine required Is it deployed on an edge server? superior, Indicates task Is it on the edge server? implement, Representation type Task From the edge server connected to the local device Transmitted to the target edge server Transmission energy consumption, Indicates task On the edge server Computational energy consumption, Indicates task From the equipment To and equipment The transmission power consumption of the connected edge servers; (c) Each task sends a request to the top-ranked edge server based on its own preference table; (d) The edge server that receives the request accepts a maximum of n tasks according to its own preference table, while rejecting other task requests; (e) Exchange matching process: The task selected by the edge server checks whether there is an exchange blocking pair; if so, the matching status is updated to... Otherwise, maintain the existing matching state. Repeat this step until there are no more swap blocking pairs. (f) Virtual machine deployment phase: The edge server matched with the task checks whether it meets the virtual machine deployment requirements based on the type of virtual machine required by the task; if it does, the task is processed and the corresponding type of virtual machine is deployed; otherwise, the task is rejected and added to the next round of matching. (g) Repeat steps (a)-(g) until the task set is complete. All tasks in the process are processed; (h) Task set Once finished, proceed to the next task set. +1 And so on.

2. The method for task offloading and virtual machine placement in edge computing based on matching game theory as described in claim 1, characterized in that: In step (1), in an industrial internet system with edge computing capabilities, industrial internet devices offload computing tasks to edge servers for execution, requiring the deployment of corresponding virtual machines on those edge servers; each edge server has a certain amount of storage space. ; The total storage space of virtual machines deployed on each edge server cannot exceed the total storage space of that edge server, satisfying the following expression: ; In the formula: Representing a time-series task flow Task Type of virtual machine required Is it deployed on an edge server? superior.

3. The method for task offloading and virtual machine placement in edge computing based on matching game theory as described in claim 1, characterized in that: In step (2), the task completion time of the time-series task flow is the time when the task finishes its calculation, including the transmission time for task unloading, the calculation time, and the waiting time; corresponding to the start time, the completion time of the calculation task executed on the local device. On edge servers Completion time of execution And the completion time when executed in the cloud. The formulas are expressed as follows: ； ； Among them, computing tasks Local device On the calculation time : Computational tasks On the edge server On the calculation time : Computational tasks Computing time in the cloud : In the formula: Indicates the completion of a task Number of CPU cycles required , , These represent industrial internet devices. Edge servers The CPU cycle of each time unit in the cloud.

4. The method for task offloading and virtual machine placement in edge computing based on matching game theory according to claim 1, characterized in that: In step (f), tasks rejected by the edge server need to update their own preference table by updating the waiting time before being added to the next round of matching; Its preference value depends on the edge server assigned and the other tasks assigned to that edge server.

Images