Edge service transition methods suitable for task continuity

Abstract

This invention provides an edge service transition method, storage medium, and electronic device suitable for task continuity. [Solution] The method involves using a GAN (Generative Adversarial Network) model to predict the user's movement trajectory and obtain a predicted trajectory, determining whether the last point of the predicted trajectory is outside the server's service area, determining the transition time and transition location based on the predicted trajectory, constructing the service operation state as a service operation log, the server transmitting the service log file and the service transition time to the target server, issuing a transition command to request that the server complete the service restoration work, deciding to perform a communication link switching operation at the transition time, the target server restoring the service to the planned state, and switching the server.

Description

[Technical Field]

[0001] The present invention relates to the technical field of resource allocation, and more particularly to an edge service transition method suitable for task continuity. [Background technology]

[0002] Edge service transition refers to the process of transferring a service from one edge server to another edge server better suited to the user's needs in an edge computing environment. It primarily addresses issues such as discontinuous service response due to user movement, including a sudden decline in service quality due to a mismatch between the finite service area of ​​the edge server and the mobility of the user's terminal, and interruptions to running edge services. Service transitions can be divided into stateless transitions and state transitions. A stateless transition simply redirects the user's request to a new server, unaffected by previous requests or states. A state transition, also known as a real-time transition, requires ensuring that the service fully restores to its pre-transition state after the transition is complete, allowing it to continue providing services to the user. Performing a state transition is a prerequisite for ensuring service continuity.

[0003] Currently, service transition methods to address the discontinuity of services caused by user mobility and limitations in the service area of ​​edge servers require two main tasks: formulating transition decisions and implementing transitions. The objective of transition decisions is to identify when and where to transition services that are operational and require transition. The main trend in current research is to select the optimal target server using optimization algorithms or reinforcement learning algorithms, with optimization goals for latency and energy consumption during the transition process. Transition implementation requires determining how services will transition so that services on the original server and the target server are synchronized, and ensuring consistency of the service operational state before and after the transition is crucial. Recent research generally employs methods that use container technology to copy the service operational state, transmit it to the target server, and then restore it to continue operation. Methods of performing state transitions using containers can be classified into two categories: state replication and state reproduction.

[0004] State replication is a method of completely replicating the service state in the original environment to the target environment. During the transition, the original server creates a snapshot of the current service operating state, including data in memory, variable values, and file system state. This snapshot is then transmitted to the target server, where it is restored. State reproduction, rather than directly replicating the state of the original environment, is a method of recreating the service in the target environment. It is also called LR log reproduction transition because it replays the service operation log on the target server as a way to achieve service synchronization. This method typically involves first collecting service log information on the original server, then transmitting the log to the target server, where the target server analyzes the log and sequentially executes the instructions within it to restore the service to its original state.

[0005] In most prior art, transition decision-making and transition execution have been studied separately. For example, Patent Document 1 discloses an edge service transition method based on trajectory prediction. Specifically, by selecting the optimal target server according to the trajectory prediction result, the response time of user services due to user movement can be shortened, and the service can be dynamically updated based on the total transition time to compensate for the transition. This method shortens the overall service response time by acquiring the transition target based on movement trajectory prediction and determining the optimal service transition path according to the optimal number of path points of the transition base station. However, it does not make a fundamental improvement to the method of implementing the service transition, and it is necessary to temporarily suspend the service in order to transmit all data synchronized with the service during the transition, so the problems of service interruption and long service time during the transition process still remain. Patent Document 2 discloses an edge service transition method based on trajectory prediction and checkpoint recovery technology. Specifically, it acquires the next location the user will move to based on trajectory prediction, sets the edge server at that location as the target server for the transition, then uses the hierarchical storage characteristics of the container to first transmit static resource data to the target server, and then uses checkpoint recovery technology to realize the data transition of the memory state. This method reduces service interruption time during service transitions to some extent by transmitting static resource data and memory state data separately. However, it directly targets the server at the user's next location, determined by trajectory prediction, meaning that trajectory prediction is not involved in the execution of the service transition, and it requires interrupting service execution to transmit memory state data.

[0006] In summary, current research on service transition methods still faces the following problems: Existing service synchronization methods all operate by synchronizing the current service operating state on the original server as the target state for service synchronization. If a service is still running during synchronization, a synchronization difference in the service state occurs, requiring the service to be temporarily suspended to complete the final synchronization in order to ensure service consistency, thus leading to the problem of long downtime. State replication methods require a large amount of data to be transmitted, resulting in high transmission energy consumption, which is also one of the reasons for long service downtime. [Prior art documents] [Patent Documents]

[0007] [Patent Document 1] Chinese Patent Application Publication No. 114554420 Specification [Patent Document 2] Chinese Patent Application Publication No. 116208608 Specification [Overview of the project] [Problems that the invention aims to solve]

[0008] The present invention aims to provide an edge service transition method suitable for task continuity. This method optimizes the entire transition operation process by combining trajectory prediction and state reproduction technology. Trajectory prediction predetermines the transition time and transition location, and based on the transition location, an optimal transition strategy is obtained using a reinforcement learning algorithm. Next, the service operation state at the time of transition is calculated, and all service synchronization work is completed in advance on the target server using this state, while giving full consideration to user mobility and avoiding service interruptions due to data transmission. Furthermore, by employing state reproduction state transitions, the amount of data transmitted between servers is reduced, and data transmission energy consumption is lowered. [Means for solving the problem]

[0009] The technical means employed in this invention are as follows:

[0010] One aspect of the present invention provides an edge service transition method suitable for task continuity. Specifically, this transition method includes: Step S1 involves using a GAN model to predict the user's movement trajectory under a transition field, obtaining the predicted trajectory, and then using a deep learning model to predict the user's movement trajectory. S2 determines whether the last point of the predicted trajectory is outside the server's service area; if not, it returns to S1; if it is, it executes S3. Step S3: Determine the transition time and transition position based on the predicted trajectory. Step S4 involves using a Q-Learning algorithm to determine the optimal transition strategy based on the transition location and server information, calculating the service operating state at the time of transition according to the current service operating status and server information, constructing the service operating state as a service operation log, and writing the service operation log containing the operating time and service operating state to the service log file. S5, the original server transmits the service log file and the service transition point to the target server, issues a transition command to the target server requesting the target server to complete the service restoration work, and ensures that the communication link switching work is performed at the transition point. S6, after the target server receives the transition instructions and service log file, creates a new container, then allocates memory and processor resources to the new container, then parses the service log file to extract the instructions therein, collects the relevant resources, executes the instructions sequentially within the container, and when all instructions are executed, it indicates that the service is in the planned state on the target server. S7. When the time reaches the transition point and the user has already entered the service area of the target server, execute the switching operation of the communication link, redirect the request to the new target server through the update of the network path configuration, ensure that the target server continues to provide services to the user, and after ensuring that the service can respond normally, delete the application on the original server.

[0011] Furthermore, in S1, the transition scenario consists of a mobile user, an edge server, and a running service.

[0012] Furthermore, in S1, the predicted user movement trajectory is composed of the positions of the user at eight future time points, Tra = {(t1,loc1),…(t mig ,loc i ),…(t8,loc8)}, where t i represents the time at the i-th point, and loc i represents the position at the i-th time point.

[0013] Furthermore, S2 <http: / / www.w3.org / 1999 / xhtml# calculates the distance l between the position loc8 of the last point of the predicted user movement trajectory and the position loc mec of the currently serving server, compares the distance l with the radius r of the server's service area, executes S3 if l > r, and returns to S1 otherwise.

[0014] Furthermore, S3 [[ID=?]] For each trajectory point in the predicted trajectory, calculate the distance l between each trajectory point and the server one by one in reverse order from the seventh trajectory point, compare it with the radius r of the server's service area, and if l < r, set the position of the trajectory point as the transition position point loc mig , and set the time corresponding to the transition position point loc mig as the transition time t mig .

[0015] Furthermore, S4 <0000?25>It should be noted that there seems to be an error in the original text where the tag in line 21 is "<http: / / www.w3.org / 1999 / xhtml# " which is not in the correct format. Also, the tag in line 41 "<0000?25>" seems to be incorrect. Please check and correct these if possible for a more accurate translation.Step S41 of obtaining an optimal transition strategy and step S42 of calculating the service operation state at the transition time, In S41, Define the transition strategy D as a matrix of size M×N, as shown in Equation (1),

Number

number

[0016] In a second aspect of the present invention, a storage medium is provided. The storage medium stores a program and, when the program is running, executes a task continuity-oriented edge service transition method as described above.

[0017] In a third aspect of the present invention, an electronic device is provided. The electronic device comprises a memory, a processor, and a computer program stored in the memory and operable on the processor, wherein the processor executes the task continuity-oriented edge service transition method described in any of the above through the operation of the computer program. [Effects of the Invention]

[0018] The present invention has the following beneficial effects compared to the prior art.

[0019] This invention optimizes the entire transition operation process by combining trajectory prediction and state reproduction technology. Trajectory prediction predetermines transition times and transition locations, completing tasks that would result in excessively long service interruptions before the official service transition, thus fully considering user mobility and reducing service interruption time. Finally, by employing a state reproduction method for state transitions, the amount of data transmitted between servers is reduced, thereby reducing data transmission energy consumption. [Brief explanation of the drawing]

[0020] To further clarify the embodiments of the present invention or the technical means in the prior art, the accompanying drawings necessary for describing the embodiments or the prior art are briefly introduced below. The following accompanying drawings represent some embodiments of the present invention, and it goes without saying that those skilled in the art can obtain other accompanying drawings from these without any creative effort.

[0021] [Figure 1]This is a flowchart of the method of the present invention. [Figure 2] This figure shows the transition field landscape of the method of the present invention. [Figure 3] This is a workflow diagram for the cold transition method. [Figure 4] This is a flowchart illustrating the workflow for pre-dump service transitions. [Figure 5] This is a flowchart illustrating the workflow of the method of the present invention. [Modes for carrying out the invention]

[0022] To further clarify the object, technical means, and merits of the embodiments of the present invention, the technical means in the embodiments of the present invention will be described clearly and completely below with reference to the drawings of the embodiments, and it goes without saying that the embodiments described are not all embodiments but only a selection of embodiments of the present invention. Any other embodiments that a person skilled in the art could obtain without creative work based on the embodiments of the present invention shall all be included within the scope of the present invention.

[0023] It should be noted that terms such as “first,” “second,” etc., in the specification, claims, and drawings of the present invention are used to distinguish similar subjects and not to indicate a specific order or priority. Since the embodiments of the present invention described herein can be carried out in an order other than that illustrated herein, it is possible to understand that the data thus adopted may be interchangeable where appropriate. Furthermore, the terms “includes” and “has,” and any variations thereof, refer to non-exclusive inclusion. For example, a process, method, system, product, or apparatus that includes a series of steps or units does not need to be limited to those steps or units explicitly listed, but further includes other steps or units that are not explicitly listed or that make those processes, methods, products, or apparatus unique.

[0024] Conventional edge service transition methods for task continuity in edge computing environments have the problem of long service interruptions because server switching operations, including file copying, data transmission, service restoration, and communication link switching, must be performed after temporarily suspending the service. This invention provides an edge service transition method based on trajectory prediction and improved LR (Log-Replay Migration) to address data transmission type service transition scenarios. In this method, the future movement trajectory of the current mobile user is first predicted using a GAN (Generative Adversarial Network) algorithm, and the service transition time and transition location are determined based on the predicted trajectory and the server's service area. Next, the selection of the optimal transition decision and preparation work for transition implementation are performed in parallel, namely, the optimal target server is selected using a Q-Learning algorithm based on the transition location, and service synchronization work is performed using improved LR. The improved LR proactively determines the service operational status at the time of communication link switching based on the transition point, and prepares the service in advance on the target server to reach that state. This makes it possible to complete all service synchronization work before the formal transition, solving the problem of service synchronization failures that occur when the service is still running during data transmission, and ensuring normal service response during service synchronization. Finally, since only a communication link switch is required at the transition point, service interruption time is reduced, continuity of service response is ensured, and energy consumption during data transmission is also reduced.

[0025] This invention addresses the problem of dynamic transitions of resource-transfer type services in an edge computing environment. The entire transition system consists of multiple edge servers and users using mobile terminals, with overlapping service areas between servers to ensure all areas are covered. Users utilize online services provided by edge servers, and it is assumed that each edge server possesses the relevant resources necessary for the service; in other words, each server can provide the service to the user.

[0026] This invention improves upon the LR log playback transition method by having the server actively determine the service operating state at the time of the communication link switchover before the transition, and having the target server prepare the service to that state in advance. Once the service on the original server is also operating to that state, it indicates that the services on both servers are synchronized. To achieve this objective, this invention determines a future point in time as the transition point based on trajectory prediction, calculates the service operating state at the transition point according to information such as the server's current processing capacity, network transmission speed, and the current operating state of the service, and sets that state as the service operating state at the time of the communication link switchover. Based on the characteristic of using the service operation log to represent the service operating state in the state reproduction method, this invention writes the calculated service operating state to the service operation log in the form of an instruction, then transmits the log to the target server, and once the target server has executed all state reproduction instructions according to the log, it indicates that the services on both servers are synchronized. In the entire process of service synchronization employing the improved LR method, each operation does not require service interruption, and the service can always respond normally, thus ensuring service continuity during the process.

[0027] As shown in Figure 1, the task continuity-oriented edge service transition method provided by the present invention includes the following steps.

[0028] S1 uses a GAN model to predict the user's movement trajectory.

[0029] First, the overall transition scene is a mobile edge computing network communication scene consisting of mobile users, edge servers, and running services.

[0030] To avoid energy consumption resulting from frequent transitions and to prevent service interruptions caused by user movement exceeding the current server's service area, the present invention uses a real-time user movement trajectory prediction method to determine whether or not to transition the service. The server collects the user's movement trajectory in real time and performs trajectory prediction once at each time interval. The specific process is as follows:

[0031] This invention employs a GAN model to predict a user's movement trajectory. GAN is a deep learning model based on generators and discriminators, and by having two networks compete against each other, it constantly approaches the true value in its prediction results. This model is an effective method in the field of movement trajectory prediction due to its advantages such as complex data processing, diversity of generation, and strong adaptability. The server makes a trajectory prediction once at time intervals according to the user's past movement trajectory. The trajectory predicted by this invention is composed of the user's positions at eight future points in time, and Tra={(t1,loc1),…(t i ,loc i ), ... (t8, loc8)} is expressed as follows: Here, t i represents the time of the i-th point, loc i This represents the position at the i-th time point.

[0032] S2 determines whether the last point of the predicted trajectory is outside the server's service area.

[0033] The predicted trajectory Tra is obtained, and the position of the 8th point, loc8, and the current position of the server providing the service, loc mec Based on this, the distance l between the two is calculated and compared with the radius r of the server's service area. If l > r, S3 is executed; otherwise, the process returns to S1.

[0034] S3. Determine the transition time point and the transition position based on the predicted trajectory.

[0035] When to transition and where to transition are two important issues that the transition decision-making should solve. Since the user is always in a moving state regardless of whether to transition or not, minimizing the service interruption time during the transition while minimizing the service response delay after the transition as much as possible, and ensuring that the user can obtain service responses in both cases before and after the transition are essential elements for ensuring service continuity.

[0036] In this invention, the predicted trajectory is used to determine the transition time point and the transition position, and the purpose includes the following three purposes. First, fully considering the user's mobility, make the transition position closer to the user's position after the transition is completed, and ensure the reliability of the transition decision-making. Second, determine the transition time point in advance to reserve time for completing service synchronization. Third, calculate the service operating state based on the transition time point to prepare for the execution of service transition using the subsequent improved LR.

[0037] Based on the predicted trajectory Tra, calculate the distance l between each trajectory point and the server one by one in reverse order from the seventh trajectory point, and compare it with the radius r of the service area of the server. When l < r, the position of the trajectory point is set as the transition position point loc mig and the time corresponding to this point is set as the transition time point t mig . By selecting the target server based on this position, the user can surely obtain service responses in both cases before and after the transition.

[0038] S4. Obtain the optimal transition strategy and calculate the service operating state at the transition time point.

[0039] (1) Obtain the optimal transition strategy.

[0040] First, define the transition strategy D as a matrix of size M×N, as shown in Equation (1).

Equation

[0041] In the formula, M represents the maximum number of tasks, N represents the number of edge servers, and D j,i This is the j-th service Task j Server MEC for execution i This indicates whether or not to transition to D j,i The value can be 0 or 1, where 0 is Task j MEC for execution i This indicates that it will not transition to MEC, and 1 is for execution. i This indicates a transition to [a different state].

[0042] Next, we construct a service transition model with the optimization goals of service data transmission delay and data transmission energy consumption. j and server MEC i The data transmission delay between the two is shown in equation (2). T j,i =data j / V i (2)

[0043] In the formula, T j,i This represents data transmission delay, and data j This is the service Task requested by the user. j This represents the required amount of data to be transmitted, V i The server is MEC i V represents the data transmission speed between the user and the system. i This is calculated according to equation (3).

number

[0044] In the formula, B i The server is MEC i This represents the available bandwidth of the corresponding transmission channel, p i The server is MEC i g represents the transmit power of the base station corresponding to d, g represents the power gain of the channel, and d i The transition position is loc mig and server MECi represents the distance to, δ represents the channel loss coefficient, and σ represents the noise power.

[0045] Service Task j When transitioning to the server MEC i The data transmission energy consumption is calculated according to Equation (4). E j,i = data mig d mig P i (4)

[0046] In the formula, E j,i represents the energy consumption generated during the transition, data mig represents the amount of data transferred, d mig represents the distance between the original server and the target server, and P i represents the transmission power of the wired link between the servers.

[0047] The objective function of the model is shown in Equation (5) and Equation (6).

Equation

[0048] In the formula, D represents the transition strategy, α represents the weight value of the delay, β represents the weight value of the energy consumption, and the goal of optimization is to minimize the total consumption per transition.

[0049] Next, based on the related information of the transition location and the server, the Q-Learning algorithm is used to find the optimal target server. The Q-Learning algorithm is a traditional reinforcement learning algorithm and is suitable for solving the decision problem of discrete action spaces. The reinforcement learning algorithm can better adapt to complex and diverse edge environments and consider the problem of long-term return compared with the case of finding the transition strategy using the optimization algorithm, so it is advantageous for improving the performance of the entire transition system.

[0050] (2) Calculate the service operating status at the time of transition.

[0051] The server collects relevant information such as memory, CPU, and bandwidth, as well as the current operational progress of the service, and determines the service operational progress at the transition point according to equations (3) and (7). S'=S+(V i *t / data task ) (7)

[0052] In the formula, S represents the current service operational progress, S' represents the service operational progress at the transition point, and V i represents the current data transmission speed between the server and the user, and t is the transition time t. mig This represents the difference between the current time and the data. task This represents the total amount of data for the service in question.

[0053] This invention defines the service operation status at the time of transition as the progress of service operation at the time of transition in a resource transmission type service transition scenario. A state reproduction state transition method is employed, and the virtualization tool used is a Docker container. Each service runs on the server as a container unit, and each container generates a log when it starts up. After obtaining the service operation status at the time of transition, the server collects the log file of the service, then constructs the calculated service operation status as a service operation log, and writes the service operation log, which contains the operating time and service operation status, to the log file of the service.

[0054] S5 transmits data and notifies the target server.

[0055] After obtaining the transition strategy and service log files at the time of transition, the server transmits the log files and service transition information to the target server via the wired link between servers, issues a transition command to the target server requesting it to complete the service restoration work, and decides to perform the communication link switching operation at the time of transition.

[0056] Data transmission in this process is the main source of energy consumption, and service regeneration can be achieved simply by transmitting the service's operational logs, which the target server then uses to regenerate the service according to. Log files are typically much smaller and consume less energy than the service's data and resources.

[0057] S6, the target server restores the service to its planned state.

[0058] After receiving instructions and log information, the target server first creates a new container, allocates memory and processor resources to the new container, then parses the log to extract the instructions, collects the relevant resources, and executes the instructions sequentially within the container. Once all instructions have been executed, it indicates that the service has reached its planned state on the target server.

[0059] Compared to state replication, the method of regeneration using log commands takes longer for service restoration. However, in this invention, a future point in time is used as the transition point using trajectory prediction, and the service operating state at the transition point is calculated from there. Since the service restoration work is performed before the official transition, although service restoration still takes a long time, it does not result in excessively long service interruptions.

[0060] S7, switch servers.

[0061] When the time or transition point arrives and the user is already within the target server's service area, the communication link switching operation is performed, the request is redirected to the new target server through a network routing update, the service continues to be provided to the user by the new target server, and the application on the original server is deleted after it is confirmed that the service can be responded to normally. Because this process inherently involves switching communications, it may cause a temporary service interruption, and the duration of this interruption is the communication link switching time.

[0062] In summary, the task continuity-oriented edge service transition method provided by the present invention, through trajectory prediction and improved LR, enables the completion of tasks such as data transmission and service restoration during service transitions before the formal transition. Service interruption time is compressed to include only the communication link switching time, and this portion of the time is affected by factors such as network topology, physical distance, server load, and network traffic. In this invention, results obtained after performing multiple service transition simulations under a 5G environment show that the communication link switching time range is 30ms to 80ms. However, since the service normally has a buffer function, the buffered data can continue to provide service to the user for approximately 1000ms when the network is disconnected. Therefore, by performing service transitions using the method of the present invention, service interruption time can be significantly reduced, and the user's service experience is not significantly affected during this period, thus improving the user's service experience and effectively ensuring service continuity.

[0063] This invention provides a novel dynamic edge service transition method based on trajectory prediction and state reproduction to address the service interruption problem during the service transition process. This invention evaluates the method primarily in terms of energy consumption and delay, where delay includes transition delay and service response delay, and energy consumption mainly refers to energy consumption resulting from data transmission that may be involved in the transition process. Since this invention achieves better results primarily by improving the entire transition process, the effectiveness of process optimization and its impact on energy consumption and delay is demonstrated by comparing it with two other transition methods in terms of both the process and experimental data. The transition scenario of this invention is shown in Figure 2.

[0064] 1. Comparison of work processes Typical service transition methods include the cold migration method and the pre-dump service transition method. Below, we will explain how the transition method provided by the present invention and the work processes of these two methods have been improved compared to conventional methods, resulting in a significant reduction in service interruption delay.

[0065] 1. Cold transition method Figure 3 shows the transition process for the cold transition method. In this method, server selection is initiated when the transition is decided, and then operations such as service suspension, data backup, and data transmission are gradually performed according to the state replication service transition method. After the target server receives the relevant data, it begins to reproduce the service, and after the service synchronization is completed, the server is switched over, the target server continues to provide services to the user, and the application that has already stopped on the original server is deleted. In the cold transition method, after selecting the target server, the service must be stopped until the entire transition process is completed, and the transition delay includes data copy time, data transmission time, service restoration time, and communication link switching time. In the cold transition method, the server is selected using the user's location at the time of the scheduled transition, but since the user is still moving during the transition, the selected server is not necessarily optimal, which may lead to a certain degree of increased service response delay. In addition, the full state replication method requires a large amount of data to be transmitted, which also causes excessive energy consumption.

[0066] 2. Pre-dump service transition method Pre-dump service transition is an improvement over the cold transition method, and its transition process is shown in Figure 4. This method addresses the problem of excessively long service interruptions in the cold transition method by utilizing the container's hierarchical storage characteristics to transmit all mirror layer data and most of the underlying layer data to the target server before the service is temporarily interrupted, thereby significantly reducing the amount of data transmitted during service interruptions and shortening transition delays. The transition delay in this method includes the synchronization time for the remaining underlying layer data and the switching time for the communication link. Since the underlying layer data is generated during service operation and constantly changes, this method transmits the modified data multiple times during the previous transmission, stopping the service only once to transmit the remaining data. While this method shortens transition delays, it does increase energy consumption for some data transmission.

[0067] 3. Method of the present invention The transition process according to the present invention is as shown in Figure 5. Trajectory prediction is used to obtain the user's movement trajectory over a certain period in the future. Based on the trajectory, a suitable transition point is selected. Service reproduction is ensured to be completed on the target server before reaching that point. The service continues to operate without interruption during this period, and the service response is temporarily suspended only when the transition point is reached and the communication link is switched. Therefore, the transition delay in the present invention includes only the communication link switching time. Furthermore, since the target server is selected using the transition position determined by trajectory prediction, this helps in selecting the optimal target server whose position is closer to the user's position when the transition is completed in the actual situation, thereby reducing the delay in service response. The present invention employs a state reproduction service synchronization method, eliminating the need to transmit large amounts of static and dynamic data. Only service operation logs need to be transmitted to the target server, thus reducing energy consumption to some extent.

[0068] 2. Comparative analysis of experimental data The comparative experiment of the present invention is carried out on a Linux (registered trademark) system installed on a Windows 11 system. The virtual machine is allocated 2GB of virtual memory and two virtual CPU cores, and the physical machine is equipped with a Core i7-12900 processor and 16GB of memory. The version of CRIU used is 3.16, and the version of Docker is 20.10. The migration decision-making experiment is carried out under a Python 3.8 environment.

[0069] (1) Experimental evaluation indicators The indicators compared in this article are delay and energy consumption respectively. The delay includes migration delay and response delay. Among them, the migration delay represents the service interruption time, which is mainly determined by the amount of work that needs to be executed after the service interruption. In the experiment, the migration delay is obtained by calculating the time difference between the start and end of the service interruption, and its calculation method is shown in Equation (8). The response delay represents the response time of the request made by the user to the target server, which is determined by factors such as the performance of the server, the output of the base station, and the distance between the user and the server, and is calculated according to Equations (2) and (3). The energy consumption refers to the energy consumption caused by transmitting data from the original server to the target server during service migration, which is mainly affected by the amount of data transferred during migration, and is calculated according to Equation (4). T mig =t over -t<M:math xmlns:M="http: / / www.w3.org / 1998 / Math / MathML"><M:mrow><M:mi>T< / M:mi>< / M:mrow>< / M:math> start (8)

[0070] In the formula, T mig represents the migration delay, t over represents the end point of the service interruption, t start represents the start point of the service interruption, T j represents the response delay between MEC j and the user, and data represents the amount of data of the user request.

[0071] (2) Comparative experiment This invention addresses the problem of excessively long service interruptions during service transitions by improving conventional transition methods based on trajectory prediction and state reproduction technologies, thereby optimizing the work process, reducing delays during service transitions, and lowering energy consumption. To demonstrate the effectiveness of the present invention, an experimental comparison was conducted with two other representative methods, a cold transition method and a pre-dump service transition method. Twenty transition tests were performed using each of the three methods, and the average values ​​of each indicator were calculated. The experimental results are shown in Table 1.

[0072] Table 1 Experimental Comparison Results [Table 1]

[0073] As can be seen from Table 1, in this invention, by completing the service restoration work in advance, the transition delay is reduced by 99.3% compared to the cold transition method and by 95.3% compared to the pre-dump service transition method. Analyzing from the perspective of response delay, in this invention, the user's position obtained by trajectory prediction is closer to the user's position when the transition is completed in the actual situation. Therefore, the response delay of the server selected based on this position is reduced by 6.6% compared to the other two methods, improving the quality of service and avoiding the user's movement during the transition process exceeding the service area of ​​the current server. Analyzing from the perspective of energy consumption, in this invention, a state reproduction method is used, and data is transmitted mainly in log format, so the amount of data transmitted is reduced compared to the container method. The energy consumption of this invention is reduced by 75.1% compared to the cold transition method and by 76.2% compared to the pre-dump service transition method.

[0074] In summary, the method of the present invention significantly reduces latency and energy consumption compared to the other two transition methods. The reduction in energy consumption helps service providers conserve energy and alleviate energy pressure on edge servers. The reduction in latency improves service quality and enhances the user experience, and in particular, significantly reduces latency during service interruptions, effectively ensuring service continuity.

[0075] Finally, the following should be explained. The above embodiments are merely for illustrating, and not limiting, the technical means of the present invention. Although the present invention has been described in detail with reference to the embodiments described above, it is also possible to modify the technical means described in the embodiments described above, or to make equivalent substitutions to some or all of their technical features. It will be obvious to those skilled in the art that such modifications or substitutions will not cause the essence of the corresponding technical means to deviate from the scope of the technical means of the embodiments of the present invention.

[0076] (Note) (Note 1) Step S1 involves using a GAN model to predict the user's movement trajectory under a transition field and obtaining the predicted trajectory. S2 determines whether the last point of the predicted trajectory is outside the server's service area; if not, it returns to S1; if it is, it executes S3. Step S3: Determine the transition time and transition position based on the predicted trajectory. Step S4 involves using a Q-Learning algorithm to determine the optimal transition strategy based on the transition location and server information, calculating the service operating state at the time of transition according to the current service operating status and server information, constructing the service operating state as a service operation log, and writing the service operation log containing the operating time and service operating state to the service log file. S5, the original server transmits the service log file and the service transition point to the target server, issues a transition command to the target server requesting the target server to complete the service restoration work, and decides to perform a communication link switchover operation at the transition point. S6, after the target server receives the transition instructions and service log file, creates a new container, allocates memory and processor resources to the new container, then parses the service log file to extract the instructions, collects the relevant resources, executes the instructions sequentially within the container, and once all instructions have been executed, indicates that the service has reached the planned state on the target server. A task continuity-oriented edge service transition method characterized by the steps of: S7, when the transition time has arrived and the user is already in the service area of ​​the target server, performing a communication link switching operation, redirecting the request to the new target server through updating the network routing, the target server continuing to provide services to the user, and deleting the application on the original server after the service has been successfully responded to.

[0077] (Note 2) The method according to Appendix 1, characterized in that, in S1, the transition scene consists of a mobile user, an edge server, and a running service.

[0078] (Note 3) In S1, the predicted user movement trajectory is composed of the user's positions at eight future points in time, where Tra={(t1,loc1),…(t i ,loc i ), ... (t8, loc8)}, t i represents the time of the i-th point, loc i The method according to Appendix 1, characterized in that represents the position at the i-th time point.

[0079] (Note 4) S2, The distance l between the position loc8 of the last point of the predicted user movement trajectory and the position loc of the server currently providing the service mec is calculated, and if l > r where r is the radius of the service area of the server, S3 is executed; otherwise, return to S1. The method according to appendix 1, characterized by including this.

[0080] (Appendix 5) S3 includes For the trajectory points in the predicted trajectory, the distance l between each trajectory point and the server is calculated one by one in reverse order from the seventh trajectory point and compared with the radius r of the service area of the server. If l < r, the position of the trajectory point is set as the transition position point loc mig and the time corresponding to the transition position point loc mig is set as the transition time t mig . The method according to appendix 1, characterized by including this.

[0081] (Appendix 6) S4 includes S41, the step of obtaining the optimal transition strategy, and S42, the step of calculating the service operation state at the transition time, In S41 The transition strategy D is defined as a matrix of size M×N and is shown in formula (1), [Number] In the formula, M represents the maximum number of tasks, N represents the number of edge servers, and D j,i represents whether the j-th service Task j transitions to the server MEC i for execution. The value of D j,i takes 0 or 1. 0 represents that Task j does not transition to MEC i for execution, and 1 represents that it transitions to MEC i for execution. Construct a service transition model with the optimization goals of service data transmission delay and data transmission energy consumption, and for service Task j and server MEC iThe data transmission delay between and is calculated according to equation (2). T j,i =data j / V i (2) In the formula, T j,i This represents data transmission delay, and data j This is the service Task requested by the user. j This represents the required amount of data to be transmitted, V i The server is MEC i V represents the data transmission speed between the user and the system. i This is calculated according to equation (3),

number

number

[0082] (Note 7) A storage medium characterized by storing a program and executing a task continuity-oriented edge service transition method described in any one of the appendices 1 to 6 when the program is running.

[0083] (Note 8) An electronic device comprising memory, a processor, and a computer program stored in the memory and operable on the processor, wherein the processor executes a task continuity-oriented edge service transition method described in any one of appendices 1 to 6 via the computer program.

Claims

1. S1, a step in which a GAN model is used to predict the user's movement trajectory under a transition field and the predicted trajectory is obtained, S2 determines whether the last point of the predicted trajectory is outside the server's service area. If it is not, it returns to S1; if it is, it executes S3. S3, a step in which the transition time and transition position are determined based on the predicted trajectory, S4, The process involves using a Q-Learning algorithm to determine the optimal transition strategy based on the transition location and server information, calculating the service operating state at the time of transition according to the current service operating status and server information, constructing the service operating state as a service operation log, and writing the service operation log containing the operating time and service operating state to the service log file. S5, the original server transmits the service log file and the service transition point to the target server, issues a transition command to the target server requesting the target server to complete the service restoration work, and decides to perform a communication link switchover operation at the transition point. S6, after the target server receives the transition instructions and service log file, it creates a new container, allocates memory and processor resources to the new container, then parses the service log file to extract the instructions, collects the relevant resources, executes the instructions sequentially within the container, and once all instructions have been executed, it indicates that the service has reached the planned state on the target server. A task continuity-oriented edge service transition method characterized by the following steps: S7, when the transition time arrives and the user is already in the service area of ​​the target server, perform a communication link switching operation, redirect the request to the new target server through a network routing update, the target server continues to provide services to the user, and deletes the application on the original server after the service has been successfully responded to.

2. The method according to claim 1, characterized in that in S1, the transition scene consists of a mobile user, an edge server, and a running service.

3. In S1, the predicted user movement trajectory is composed of the user's position at eight future points in time, where Tra = {(t １ ,loc １ ), ... (t ｉ ,loc ｉ ), ... (t ８ ,loc ８ )} is expressed as, t ｉ represents the time of the i-th point, and loc ｉ The method according to claim 1, characterized in that represents the position at the i-th time point.

4. S2 is The position loc of the last point of the predicted user movement trajectory ８ and the position loc of the server currently providing the service ｍｅｃ calculate the distance l between them, compare the distance l with the radius r of the service area of the server, and if l > r, execute S3, otherwise return to S1. The method according to claim 1, characterized by including this.

5. S3 is For each trajectory point in the predicted trajectory, the distance l between each trajectory point and the server is calculated one by one in reverse order starting from the 7th trajectory point and compared with the radius r of the server's service area. If l < r, the position of that trajectory point is set to the transition position point lc. ｍｉｇ Let the transition position point be loc ｍｉｇ The time corresponding to the transition point t ｍｉｇ The method according to claim 1, characterized in that it includes the following:

6. S4 is S41 includes the step of determining the optimal transition strategy, and S42 includes the step of calculating the service operating status at the time of transition. In S41, The transition strategy D is defined as an M x N size matrix and is shown in equation (1). [Math 1] In the formula, M represents the maximum number of tasks, N represents the number of edge servers, and D ｊ,ｉ This is the jth service Task ｊ For execution, the server MEC ｉ This indicates whether or not to transition to D ｊ,ｉ The value can be 0 or 1, where 0 is Task ｊ MEC for execution ｉ This indicates that it will not transition to MEC, and 1 means that MEC will be executed. ｉ This represents a transition to, A service transition model was constructed with the goal of optimizing the data transmission delay and data transmission energy consumption of the service, and the service task... ｊ and server MEC ｉ The data transmission delay between and is calculated according to equation (2). T ｊ，ｉ =data ｊ / V ｉ (2) In the ceremony, T ｊ,ｉ This represents data transmission delay, and data ｊ This is the service Task requested by the user. ｊ This represents the required amount of data to be transmitted, V ｉ is a server MEC ｉ This represents the data transmission speed between the user and the system, V ｉ This is calculated according to equation (3), [Math 2] In the formula, B ｉ is a server MEC ｉ This represents the available bandwidth of the corresponding transmission channel, p ｉ is a server MEC ｉ g represents the transmit power of the base station corresponding to d, g represents the power gain of the channel, and d ｉ is the transition position loc ｍｉｇ and server MEC ｉ This represents the distance between the two points, δ represents the channel loss coefficient, and σ represents the noise power. Service Task ｊ is server MEC ｉ The data transmission energy consumption during the transition is calculated according to equation (4). E ｊ，ｉ =data ｍｉｇ d ｍｉｇ P ｉ (4) In the formula, E ｊ,ｉ This represents the amount of energy consumed during the transition, and data ｍｉｇ represents the amount of data that has been transitioned, and d ｍｉｇ This represents the distance between the original server and the target server, P ｉ This represents the transmission power of the wired link between servers. The target function of the service transition model is given by equations (5) and (6), [Math 3] α＋β＝１，（α＞０，β＞０） （６） In the formula, D represents the transition strategy, α represents the delay weight, and β represents the energy consumption weight. The optimization goal is to minimize the total consumption per transition. Next, the optimal target server is determined using the Q-Learning algorithm based on the transition location and server relationship information. In S42, The server collects memory, CPU, and bandwidth-related information and the current service uptime. ｉ The service operational progress at the transition point is determined by S', and S' is calculated according to equation (7). S’=S+(V ｉ *t / data ｔａｓｋ ) (7) In the formula, S represents the current service operational progress, S' represents the service operational progress at the transition point, and V ｉ represents the current data transmission speed between the server and the user, and t is the transition time t. ｍｉｇ This represents the difference between the current time and the past time, and data ｔａｓｋ The method according to claim 1, characterized in that represents the total amount of data for the service.

7. A storage medium characterized by storing a program and executing a task continuity-oriented edge service transition method described in any one of claims 1 to 6 when the program is running.

8. An electronic device comprising memory, a processor, and a computer program stored in the memory and operable on the processor, wherein the processor executes a task continuity-oriented edge service transition method described in any one of claims 1 to 6 via the computer program.

Images