Synchronization method and device for collaborative stream data processing based on edge assistance

Abstract

The present invention discloses a synchronization method and device for collaborative stream data processing based on edge assistance. The present invention constructs a method to minimize the overall system delay variance and the reconnection loss accompanying task migration as the overall optimization goal according to the service status of each server at the start of the current time slot, the network propagation delay and transmission delay between the edge and the terminal user, and the load of the edge server, and confirms the overall constraints; decomposes the overall optimization goal into the time slot, constructs the time slot optimization goal and the time slot constraints, solves the time slot optimization goal using a convex optimization strategy, and randomly rounds the solution result as a decision for the redistribution of stream data computing tasks, and deploys and migrates user stream data computing tasks according to the decision for the redistribution of stream data computing tasks. The present invention can effectively improve the synchronization efficiency and service quality of stream data, and has a high application value.

Description

Technical Field

[0001] The present invention relates to the technical field of edge-assisted multi-party stream data computing task deployment, and specifically to a synchronization method and device for edge-assisted collaborative stream data processing. Background Art

[0002] With the continuous progress of fifth-generation mobile communication technology (5G) and edge computing technology, applications that rely on collaborative streaming data processing, such as high-definition real-time video communication, interactive online education, remote medical collaborative surgery, virtual reality (VR), and digital twin technology, are becoming mainstream. These applications are characterized by the large amount of data transmitted and the involvement of multiple parties, and the streaming data computing tasks are coordinated by multiple parties. This requires not only that the application can provide a high degree of real-time performance to ensure the quality of service, but also that data and operations are synchronized to ensure effective collaboration across multiple participants and devices. However, meeting the balance between real-time and synchronization requirements in the above collaborative streaming data applications, especially in resource-limited environments, constitutes a complex challenge to be solved.

[0003] Edge servers can act as relays between central cloud servers and terminals. With the help of edge computing, the real-time problem of applications is alleviated to a certain extent by allocating computing tasks to nodes approximately near the user's geographical location; on the other hand, it also alleviates the computing pressure of central cloud servers and terminal devices. However, given that the network conditions faced by all network participants may show unpredictable fluctuations, as well as the significant heterogeneity between devices and the uneven distribution of computing power, the combined effect of these factors may cause challenges in streaming data synchronization while ensuring real-time response of the system:

[0004] First, for users who rely on streaming data, what is needed is a synchronization state that is as continuous as possible, rather than synchronization limited to a short period of time, which is obviously not enough to maintain a high-quality user experience. Given the inherent instability of network status, there are undoubtedly significant challenges in achieving a long-term stable synchronization mechanism. From another perspective, if an online scheduling strategy is adopted, the task migration accompanying task deployment may have a series of negative effects on the user experience, which not only affects the immediate task execution efficiency, but may also trigger a series of impacts in subsequent time slots, forming a complex cascade effect. How to overcome the above difficulties and ensure data synchronization on the basis of ensuring real-time performance has become an important topic in the current research field.

[0005] Therefore, a new scheduling method for cloud-edge collaborative environment is needed to achieve collaborative stream data synchronization on the basis of ensuring real-time performance, while ensuring that user experience is not affected by task migration. Summary of the invention

[0006] The purpose of the present invention is to provide a synchronization method and device for collaborative stream data processing based on edge assistance, which improves the synchronization efficiency of stream data on the basis of ensuring real-time performance, while ensuring that the user experience is not affected by task migration. In order to solve the above technical problems, the present invention provides the following technical solutions:

[0007] The synchronization method of edge-assisted collaborative stream data processing includes the following steps:

[0008] According to the above technical solution, at the beginning of each time slot, the service status of each server at the beginning of the current time slot, the network propagation delay and transmission delay between the edge and the terminal user, and the load situation of the edge are obtained;

[0009] According to the service status of each server at the start of the current time slot, the network propagation delay and transmission delay between the edge and the end user, and the load of the edge server, the overall delay variance of each user flow data computing task deployment is calculated;

[0010] Minimizing the overall system latency variance and the reconnection loss associated with task migration is taken as the overall optimization goal, and the overall constraints are confirmed;

[0011] The overall optimization goal is decomposed into time slots, the time slot optimization goal and time slot constraints are constructed, the time slot optimization goal is solved using the convex optimization strategy, and the solution result is randomly rounded as the decision for the redistribution of the streaming data computing task.

[0012] Deploy and migrate user stream data computing tasks based on the decision of stream data computing task redistribution;

[0013] At the beginning of the next time slot, a new round of stream data computing task redistribution is performed.

[0014] According to the above technical solution, the overall optimization goal is:

[0015]

[0016] In the formula, is the set of all time slots, is the set of all users, ε is the set of all edge servers, and t represents the set The elements in u represent the set The elements in the set ε, e represents the elements in the set ε, and x u,e,t represents a binary decision variable, indicating whether the stream data computing task of user u is executed by edge server e at time slot t. If x u,e,t =1, then the stream data computing task is executed by the edge server, otherwise the stream data computing task is not executed by the edge server. is the sum of the delays perceived by user u at the beginning of time slot t, The value is the network transmission delay Propagation Delay And the computational transmission delay of the edge server The sum of represents the average delay of all users in time slot t, x u,e,t-1 Indicates the decision result of whether the stream data computing task of user u in the previous time slot is assigned to edge server e. is the delay value when user u reconnects with edge server e, The value can be 3 times the network transmission delay The value indicates that ω is the importance of the reconnection loss caused by task migration relative to the synchronization state.

[0017] According to the above technical solution, the overall constraints include:

[0018] The perceived delay of each user in each time slot must meet the requirements of interactive applications:

[0019]

[0020] The streaming data computing tasks undertaken by each server in each time slot should be within the load range:

[0021]

[0022] For each scheduled stream data computing task in each time slot, each scheduled stream data computing task can only be distributed on one server and cannot be served by multiple servers:

[0023]

[0024] In the formula, is the set of all time slots, is the set of all users, ε is the set of all edge servers, and t represents the set The elements in u represent the set The elements in the set ε, e represents the elements in the set ε, and x u,e,t is a binary decision variable, indicating whether the stream data computing task of user u is executed by edge server e at time slot t. is the sum of the delays perceived by user u at the beginning of time slot t, α is the threshold of user perceived delay, r u,e,t is the resource value requested by user u from edge server e in time slot t, β e is the available resource value of edge server e.

[0025] According to the above technical solution, the time slot optimization target is:

[0026]

[0027] in, is the set of all time slots, is the set of all users, ε is the set of all edge servers, and t represents the set The elements in u represent the set The elements in the set ε, e represents the elements in the set ε, and x u,e,t is a binary decision variable, indicating whether the stream data computing task of user u is executed by edge server e at time slot t, x u,e,t-1 Indicates the decision result of whether the stream data computing task of user u in the previous time slot is assigned to edge server e. is the sum of the delays perceived by user u at the beginning of time slot t, represents the mean delay of all users in time slot t, is the delay value when user u reconnects with edge server e, o e,t is a time-varying penalty term.

[0028] According to the above technical solution, the time slot constraint condition is:

[0029] in, It is the decision matrix composed of all task deployment decision variables at time slot t, and its domain is the continuous interval [0,1] in the real number field; h(.) is the vector composed of the constraints of each time slot t, including the vector composed of the overall constraints.

[0030] According to the above technical solution, the nonlinear term x u,e,t-1 ⊕x u,e,t The equivalent conversion is performed using an equivalent function, wherein the equivalent function g(x u,e,t ,x u,e,t-1 ) The conditions to be met include:

[0031] When x u,e,t =1,x u,e,t-1 =1, g(1,1)=0;

[0032] When x u,e,t =1,x u,e,t-1 =0, g(1,0)=1;

[0033] When x u,e,t =0,x u,e,t-1 =1, g(0,1)=0;

[0034] When x u,e,t =0,x u,e,t-1 =0, g(0,0)=0;

[0035] In the formula, xu,e,t-1 Indicates the decision result of whether the stream data computing task of user u in the previous time slot is assigned to edge server e. is the delay value when user u reconnects with edge server e, x u,e,t is a binary decision variable, indicating whether the stream data computing task of user u is executed by edge server e at time slot t.

[0036] According to the above technical solution, the penalty term o that varies with time e,t The determination process includes: for any time slot t, the total delay of all users served by each edge server e before time slot t Calculate the variance and normalize the variance value to get o e,t , o e,t The value is updated according to time as the task migration penalty item in each time slot t;

[0037] Said o e,t The calculation formula is:

[0038]

[0039] Among them, the collection represents the set of users served by edge server e in the historical time slot; represents the set of time slots served by edge server e for user u; is the average delay of all users in time slot t, is the sum of the delays perceived by user u at the beginning of time slot t.

[0040] According to the above technical solution, the random rounding includes:

[0041] For real matrices The result of any row and any column in u on the integer domain The probability of The probability of This process is implemented by random sampling. Integrate to form the solution on the integer domain

[0042] In another embodiment, a synchronization device for edge-assisted collaborative stream data processing includes:

[0043] The delay state sampling module obtains the service status of each server at the beginning of each time slot, the network propagation delay and transmission delay between the edge and the end user, and the load of the edge server at the beginning of each time slot, and records the total delay data of all users served by each edge in each time slot;

[0044] Deploy the decision-making computing module to minimize the overall system delay variance and the reconnection loss associated with task migration as the overall optimization goal based on the current delay status and historical edge service conditions, and confirm the overall constraints; decompose the overall optimization goal into time slots, construct time slot optimization goals and time slot constraints, use the convex optimization strategy to solve the time slot optimization goal, and randomly round the solution result as the decision for the redistribution of streaming data computing tasks;

[0045] Deploy decision-making application modules to form a consensus on the cloud, edge, and end, execute decisions on the redistribution of stream data computing tasks in each time slot, and migrate user task instances.

[0046] Among them, the overall optimization goal in the deployment decision calculation module is:

[0047]

[0048] In the formula, is the set of all time slots, is the set of all users, ε is the set of all edge servers, and t represents the set The elements in u represent the set The elements in the set ε, e represents the elements in the set ε, and x u,e,t represents a binary decision variable, indicating whether the stream data computing task of user u is executed by edge server e at time slot t. If x u,e,t =1, then the stream data computing task is executed by the edge server, otherwise the stream data computing task is not executed by the edge server. is the sum of the delays perceived by user u at the beginning of time slot t, The value is the network transmission delay Propagation Delay And the computational transmission delay of the edge server The sum of represents the average delay of all users in time slot t, x u,e,t-1 Indicates the decision result of whether the stream data computing task of user u in the previous time slot is assigned to edge server e. is the delay value when user u reconnects with edge server e, The value can be 3 times the network transmission delay The value indicates that ω is the importance of the reconnection loss caused by task migration relative to the synchronization state.

[0049] The deployment decision calculation module also includes overall constraints for constraining the overall optimization goal, and the overall constraints include:

[0050] The perceived delay of each user in each time slot must meet the requirements of interactive applications:

[0051]

[0052] The streaming data computing tasks undertaken by each server in each time slot should be within the load range:

[0053]

[0054] For each scheduled stream data computing task in each time slot, each scheduled stream data computing task can only be distributed on one server and cannot be served by multiple servers:

[0055]

[0056] In the formula, is the set of all time slots, is the set of all users, ε is the set of all edge servers, and t represents the set The elements in u represent the set The elements in the set ε, e represents the elements in the set ε, and x u,e,t is a binary decision variable, indicating whether the stream data computing task of user u is executed by edge server e at time slot t. is the sum of the delays perceived by user u at the beginning of time slot t, α is the threshold of user perceived delay, r u,e,t is the resource value requested by user u from edge server e in time slot t, β e is the available resource value of edge server e.

[0057] The time slot optimization target in the deployment decision calculation module is:

[0058]

[0059] in, is the set of all time slots, is the set of all users, ε is the set of all edge servers, and t represents the set The elements in u represent the set The elements in the set ε, e represents the elements in the set ε, and x u,e,t represents a binary decision variable, indicating whether the stream data computing task of user u is executed by edge server e at time slot t. If x u,e,t =1, then the stream data computing task is executed by the edge server, otherwise the stream data computing task is not executed by the edge server, x u,e,t-1 Indicates the decision result of whether the stream data computing task of user u in the previous time slot is assigned to edge server e. is the sum of the delays perceived by user u at the beginning of time slot t, represents the mean delay of all users in time slot t, is the delay value when user u reconnects with edge server e, o e,t is a time-varying penalty term.

[0060] In addition to the time slot constraints in the deployment decision calculation module:

[0061] in, It is the decision matrix composed of all task deployment decision variables at time slot t, and its domain is the continuous interval [0,1] in the real number field; h(.) is the vector composed of the constraints of each time slot t, including the vector composed of the overall constraints.

[0062] Compared with the prior art, the beneficial effects achieved by the present invention are as follows: the present invention proposes a synchronization strategy for collaborative streaming data processing based on edge assistance, which continuously redistributes streaming data computing tasks based on the overall delay error of the current system and the reliability of the server, that is, decides whether the user task instance needs to be migrated and the target server for migration, so as to achieve the goal of maximizing the continuous synchronization of the application and ensure the synchronous interaction between multiple users as much as possible. The deployment device of the present invention can evaluate the reliability of the server by using the historical service data of the server when the subsequent impact of unknown task migration occurs, so as to avoid the impact of unnecessary migration on the user experience. Compared with the existing technology, the scheduling scheme of this article can capture the overall delay difference in real time, adjust the task deployment scheme in time, and have the estimated effect of migration benefits, which can improve the overall synchronization effect of the application while effectively avoiding invalid migration. BRIEF DESCRIPTION OF THE DRAWINGS

[0063] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings:

[0064] Figure 1 is a flowchart of the steps of the synchronization method of edge-assisted collaborative stream data processing of the present invention;

[0065] Figure 2 It is a schematic diagram of the cloud-edge collaborative asynchronous transmission stream data computing architecture implemented by the present invention;

[0066] Figure 3 It is the change of the time slot synchronization state after applying the synchronization method based on edge-assisted collaborative stream data processing;

[0067] Figure 4It is the cumulative distribution function curve of the perceived delay in all time slots of users after applying the synchronization method based on edge-assisted collaborative streaming data processing. DETAILED DESCRIPTION

[0068] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0069] The present invention provides a technical solution:

[0070] Example 1

[0071] A synchronization method for collaborative stream data processing based on edge assistance, the method steps include:

[0072] S1. At the beginning of each time slot (the length of the time slot is determined by the system's requirements for synchronization), the central cloud scheduler obtains the service status of each server at the beginning of the current time slot, the network propagation delay and transmission delay between the edge and the end user, and the load of the edge server based on sampling.

[0073] S2. Calculate the overall delay variance of each user flow data computing task deployment based on the service status of each server at the start of the current time slot, the network propagation delay and transmission delay between the edge and the terminal user, and the load of the edge server;

[0074] S3, taking minimizing the overall system delay variance and the reconnection loss associated with task migration as the overall optimization goal, and confirming the overall constraints;

[0075]

[0076] In the formula, is the set of all time slots, is the set of all users, ε is the set of all edge servers, and t represents the set The elements in u represent the set The elements in the set ε, e represents the elements in the set ε, and x u,e,t represents a binary decision variable, indicating whether the stream data computing task of user u is executed by edge server e at time slot t. If x u,e,t =1, then the stream data computing task is executed by the edge server, otherwise the stream data computing task is not executed by the edge server. is the sum of the delays perceived by user u at the beginning of time slot t, The value is the network transmission delay Propagation Delay And the computational transmission delay of the edge server The sum of represents the average delay of all users in time slot t, x u,e,t-1 Indicates the decision result of whether the stream data computing task of user u in the previous time slot is assigned to edge server e. is the delay value when user u reconnects with edge server e, and ω is the importance of the reconnection loss caused by task migration relative to the synchronization state.

[0077] Among them, the nonlinear term x u,e,t-1 ⊕x u,e,t The equivalent conversion is performed using an equivalent function, wherein the equivalent function g(x u,e,t ,x u,e,t-1 ) The conditions include:

[0078] When x u,e,t =1,x u,e,t-1 =1, g(1,1)=0;

[0079] When x u,e,t =1,x u,e,t-1 =0, g(1,0)=1;

[0080] When x u,e,t =0,x u,e,t-1 =1, g(0,1)=0;

[0081] When x u,e,t =0,x u,e,t-1 =0, g(0,0)=0;

[0082] Therefore, the equivalent function g(x u,e,t ,x u,e,t-1 ) can be: g(x u,e,t ,x u,e,t-1 )=x u,e,t ·(1-x u,e,t-1 );

[0083] In the formula, x u,e,t-1 Indicates the decision result of whether the flow data computing task of user u in the previous time slot is assigned to edge server e, x u,e,t represents a binary decision variable, indicating whether the stream data computing task of user u is executed by edge server e at time slot t. If x u,e,t =1, then the stream data computing task is executed by the edge server, otherwise the stream data computing task is not executed by the edge server.

[0084] The overall constraints include:

[0085] The perceived delay of each user in each time slot must meet the requirements of interactive applications:

[0086]

[0087] The streaming data computing tasks undertaken by each server in each time slot should be within the load range:

[0088]

[0089] For each scheduled stream data computing task in each time slot, each scheduled stream data computing task can only be distributed on one server and cannot be served by multiple servers:

[0090]

[0091] In the formula, is the set of all time slots, is the set of all users, ε is the set of all edge servers, and t represents the set The elements in u represent the set The elements in the set ε, e represents the elements in the set ε, and x u,e,t is a binary decision variable, indicating whether the stream data computing task of user u is executed by edge server e at time slot t. is the sum of the delays perceived by user u at the beginning of time slot t, α is the threshold of user perceived delay, r u,e,t is the resource value requested by user u from edge server e in time slot t, β e is the available resource value of edge server e.

[0092] S4. Decompose the overall optimization target into time slots, construct the time slot optimization target and time slot constraints, use the convex optimization strategy to solve the time slot optimization target, and solve the real number matrix obtained by solving the time slot optimization target. The result of any row and any column in u on the integer domain The probability of The probability of This process is implemented by random sampling. Integrate to form the solution on the integer domain The solution on the integer domain will be restored As a decision on the redistribution of stream data computing tasks, the user stream data computing tasks are deployed and migrated according to the decision on the redistribution of stream data computing tasks;

[0093] Specifically, during the task redistribution process, the new edge server that provides streaming data computing services to users needs to select the server with the most available bandwidth from all servers that host the target user streaming data computing model and user data, including the central cloud server, to migrate the streaming data computing model and download the user data, so as to ensure that the migration process consumes the shortest time. After each edge and cloud server reach a consensus, the edge deploys the streaming data computing model of the corresponding user and starts to provide streaming data computing services to the corresponding user;

[0094] Before the next time slot begins, each edge server continuously receives stream data calculation requests (control flow information) from end users, and the edge server executes instructions, calculates and transmits them. At the same time, the edge server records the total service delay during the interaction process and transmits the delay information to the central cloud server, providing server reliability information for the scheduler to perform task redistribution in the next time slot;

[0095] At the end of the time slot, the edge server performs a propagation delay sampling with all users. Then the edge server uploads the delay information such as the calculation delay, transmission delay, and propagation delay with the terminal used at the current moment to the central cloud server as the network status information of the user and the edge sampled by the central cloud server at the beginning of the next time slot.

[0096] Wherein, the time slot optimization objective is:

[0097]

[0098] in, is the set of all time slots, is the set of all users, ε is the set of all edge servers, and t represents the set The elements in u represent the set The elements in the set ε, e represents the elements in the set ε, and x u,e,t represents a binary decision variable, indicating whether the stream data computing task of user u is executed by edge server e at time slot t. If x u,e,t =1, then the stream data computing task is executed by the edge server, otherwise the stream data computing task is not executed by the edge server, x u,e,t-1 Indicates the decision result of whether the stream data computing task of user u in the previous time slot is assigned to edge server e. is the sum of the delays perceived by user u at the beginning of time slot t, represents the mean delay of all users in time slot t, is the delay value when user u reconnects with edge server e, o e,t represents the penalty term that varies over time.

[0099] The time-varying penalty term oe,t The determination process includes: for any time slot t, the total delay of all users served by each edge server e before time slot t Calculate the variance and normalize the variance value to get o e,t , o e,t The value is updated according to time as the task migration penalty item in each time slot t;

[0100] Said o e,t The calculation formula is:

[0101]

[0102] Among them, the collection represents the set of users served by edge server e in the historical time slot; represents the set of time slots served by edge server e for user u; is the average delay of all users in time slot t, is the sum of the delays perceived by user u at the beginning of time slot t.

[0103] Time slot constraints:Time slot constraints:

[0104] in, It is the decision matrix composed of all task deployment decision variables at time slot t, and its domain is the continuous interval [0,1] in the real number field; h(.) is the vector composed of the constraints of each time slot t, including the vector composed of the overall constraints.

[0105] S5. At the beginning of the next time slot, a new round of stream data computing task redistribution is executed. The central cloud server reconstructs the optimization problem with the goal of minimizing the overall system delay difference and the user's redistribution loss based on the information of the previous time slot obtained by the scheduler and the service status of each edge upload, and sends out the plan and executes a new round of stream data computing task redistribution.

[0106] Example 2

[0107] A cloud-edge network with 5 edge servers and 5 users is taken as an example to further illustrate the random rounding process. Under this configuration, the resource request and allocation settings are set to each server to provide services to at most 1 user, the network status between the edge and the user is set to a random number that satisfies the normal distribution, and the user's latency threshold is set to 120ms.

[0108] (1) After solving with the convex optimization strategy CVXPY, the results are [[0.1,0.1,0.3,0.4,0.1],[0.4,0.1,0.3,0.1,0.1],[0.1,0.4,0.3,0.1,0.1],[0.1,0.3,0.4,0.1,0.1],[0.1,0.1,0.3,0.1,0.4]], and the result after random rounding is [[0,0,0,1,0],[1,0,0,0,0],[0,1,0,0,0],[0,0,1,0,0],[0,0,0,0,1]].

[0109] (2) The change of the user's synchronization state within 300 time slots is as follows: Figure 3 As shown in the figure, the results show that the delay difference between users can be controlled to fluctuate within an overall fluctuation of about 10ms, which can better guarantee the streaming data synchronization needs of multiple users; Figure 4 This indicates that the delay of each user is less than the threshold of 110ms, meeting the user's real-time requirements.

[0110] Example 3

[0111] A synchronization device for collaborative stream data processing based on edge assistance, comprising:

[0112] The delay state sampling module obtains the service status of each server at the beginning of each time slot, the network propagation delay and transmission delay between the edge and the end user, and the load of the edge server at the beginning of each time slot, and records the total delay data of all users served by each edge in each time slot;

[0113] Deploy the decision-making computing module to minimize the overall system delay variance and the reconnection loss associated with task migration as the overall optimization goal based on the current delay status and historical edge service conditions, and confirm the overall constraints; decompose the overall optimization goal into time slots, construct time slot optimization goals and time slot constraints, solve the time slot optimization goal using a convex optimization strategy, and randomly round off the solution result as a decision for redistribution of streaming data computing tasks;

[0114] Deploy decision-making application modules to form a consensus on the cloud, edge, and end, execute decisions on the redistribution of stream data computing tasks in each time slot, and migrate user task instances.

[0115] The synchronization device for collaborative stream data processing based on edge assistance also includes a central cloud server, a central scheduler, an edge server node, a terminal device, and a base station and network that undertakes relay, wherein the edge and the central cloud server are connected through a high-speed link, and the user and the edge or cloud are connected through a wireless network. The central cloud server provides direct or indirect services to all users participating in the game, is responsible for running the complete model of stream data calculation and storing the data of all users, and redistributes the stream data calculation tasks through the scheduler. When the user is directly connected to the cloud service, the server processes the uplink control flow data from the client and sends the calculated data to the end user. A scheduler module runs on the cloud server, which periodically samples the delay status between the edge and each user, as well as the edge's calculation status and historical service data, so as to deploy the stream data calculation task according to the current overall system delay status. On the other hand, a large number of users are mainly served by edge servers. The edge server has computing and storage resources, but compared with the central cloud server, the resources of the edge server are limited, and the user's server request cannot exceed the load limit of the edge. When the end user transmits the control flow information to the edge via the wireless network, the edge performs calculations and stream data calculations, and finally transmits the results to the end user through the base station and the wireless network; on the other hand, it makes full use of the idle time of the communication link to forward the user control flow to the cloud server, which synchronizes it through the high-speed link between the cloud and the edge. The user's terminal devices are heterogeneous, and devices that can perform wireless transmission and projection tasks, such as smartphones, personal computers, and head-mounted displays, can all take on the tasks of the terminal.

[0116] During a longer application cycle, multi-party terminal devices based on streaming data interaction periodically interact with the cloud-edge asynchronous transmission streaming data computing device, and provide streaming data computing services to users according to this method. User task instances are migrated and deployed according to the task redistribution strategy to achieve continuous synchronization between user nodes and ensure good real-time performance.

[0117] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device.

[0118] Finally, it should be noted that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art can still modify the technical solutions described in the aforementioned embodiments or replace some of the technical features therein by equivalents. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A synchronization method for collaborative stream data processing based on edge assistance, characterized in that: The method steps include: At the beginning of each time slot, obtain the service status of each server at the beginning of the current time slot, the network propagation delay and transmission delay between the edge and the end user, and the load of the edge server; According to the service status of each server at the start of the current time slot, the network propagation delay and transmission delay between the edge and the end user, and the load of the edge server, the overall delay variance of each user flow data computing task deployment is calculated; The overall optimization goal is to minimize the overall system delay variance and the reconnection loss associated with task migration, and confirm the overall constraints; the overall optimization goal is: In the formula, is the set of all time slots, is the set of all users, ε is the set of all edge servers, and t represents the set The elements in u represent the set The elements in the set ε, e represents the elements in the set ε, and x u,e,t represents a binary decision variable, indicating whether the stream data computing task of user u is executed by edge server e at time slot t. If x u,e,t =1, then the stream data computing task is executed by the edge server, otherwise the stream data computing task is not executed by the edge server. is the sum of the delays perceived by user u at the beginning of time slot t, The value is the network transmission delay Propagation Delay And the computational transmission delay of the edge server The sum of represents the average delay of all users in time slot t, x u,e,t-1 Indicates the decision result of whether the stream data computing task of user u in the previous time slot is assigned to edge server e. is the delay value when user u reconnects with edge server e, ω is the importance of the reconnection loss caused by task migration relative to the synchronization state; The overall optimization goal is decomposed into time slots, the time slot optimization goal and time slot constraints are constructed, the time slot optimization goal is solved using the convex optimization strategy, and the solution result is randomly rounded as the decision for the redistribution of the streaming data computing task. Deploy and migrate user stream data computing tasks based on the decision of stream data computing task redistribution; At the beginning of the next time slot, a new round of stream data computing task redistribution is performed.

2. The synchronization method for edge-assisted collaborative stream data processing according to claim 1, characterized in that: The overall constraints include: The perceived delay of each user in each time slot must meet the requirements of interactive applications: The streaming data computing tasks undertaken by each server in each time slot should be within the load range: For each scheduled stream data computing task in each time slot, each scheduled stream data computing task can only be distributed on one server and cannot be served by multiple servers: In the formula, is the set of all time slots, is the set of all users, ε is the set of all edge servers, and t represents the set The elements in u represent the set The elements in the set ε, e represents the elements in the set ε, and x u,e,t is a binary decision variable, indicating whether the stream data computing task of user u is executed by edge server e at time slot t. is the sum of the delays perceived by user u at the beginning of time slot t, α is the threshold of user perceived delay, r u,e,t is the resource value requested by user u from edge server e in time slot t, β e is the available resource value of edge server e.

3. The synchronization method for edge-assisted collaborative stream data processing according to claim 1, characterized in that: The time slot optimization goal is: in, is the set of all time slots, is the set of all users, ε is the set of all edge servers, and t represents the set The elements in u represent the set The elements in the set ε, e represents the elements in the set ε, and x u,e,t represents a binary decision variable, indicating whether the stream data computing task of user u is executed by edge server e at time slot t, x u,e,t-1 Indicates the decision result of whether the stream data computing task of user u in the previous time slot is assigned to edge server e. is the sum of the delays perceived by user u at the beginning of time slot t, represents the mean delay of all users in time slot t, is the delay value when user u reconnects with edge server e, o e,t is a time-varying penalty term.

4. The synchronization method for edge-assisted collaborative stream data processing according to claim 1, characterized in that: Time slot constraints: in, It is the decision matrix composed of all task deployment decision variables at time slot t, and its domain is the continuous interval [0,1] on the real number field; h(.) is the vector composed of the constraints of each time slot t.

5. The synchronization method for edge-assisted collaborative stream data processing according to claim 1, characterized in that: Nonlinear terms The equivalent conversion is performed using an equivalent function, wherein the equivalent function g(x u,e,t ,x u,e,t-1 ) The conditions to be met include: When x u,e,t =1,x u,e,t-1 =1, g(1,1)=0; When x u,e,t =1,x u,e,t-1 =0, g(1,0)=1; When x u,e,t =0,x u,e,t-1 =1, g(0,1)=0; When x u,e,t =0,x u,e,t-1 =0, g(0,0)=0; In the formula, x u,e,t-1 Indicates the decision result of whether the stream data computing task of user u in the previous time slot is assigned to edge server e. is the delay value when user u reconnects with edge server e, x u,e,t is a binary decision variable, indicating whether the stream data computing task of user u is executed by edge server e at time slot t.

6. The synchronization method for edge-assisted collaborative stream data processing according to claim 3, characterized in that: The time-varying penalty term o e,t The determination process includes: for any time slot t, the total delay of all users served by each edge server e before time slot t Calculate the variance and normalize the variance value to get o e,t , o e,t The value is updated according to time as the task migration penalty item in each time slot t; Said o e,t The calculation formula is: Among them, the collection represents the set of users served by edge server e in the historical time slot; represents the set of time slots served by edge server e for user u; is the average delay of all users in time slot t, is the sum of the delays perceived by user u at the beginning of time slot t.

7. The synchronization method for edge-assisted collaborative stream data processing according to claim 1, characterized in that: The random rounding includes: For real matrices The result of any row and any column in u on the integer domain The probability of The probability of The decision variable Integrate to form the solution on the integer domain 8. A synchronization device for collaborative stream data processing based on edge assistance, characterized in that: include: The delay status sampling module obtains the service status of each server at the beginning of each time slot, the network propagation delay and transmission delay between the edge and the end user, and the load of the edge server at the beginning of each time slot, and records the total delay data of all users served by each edge in each time slot; Deploy the decision-making computing module, and according to the current delay status and the historical service status of the edge, minimize the overall system delay variance and the reconnection loss associated with task migration as the overall optimization goal, and confirm the overall constraints; decompose the overall optimization goal into time slots, construct time slot optimization goals and time slot constraints, use the convex optimization strategy to solve the time slot optimization goal, and randomly round the solution result as the decision for the redistribution of the streaming data computing task; the overall optimization goal is: In the formula, is the set of all time slots, is the set of all users, ε is the set of all edge servers, and t represents the set The elements in u represent the set The elements in the set ε, e represents the elements in the set ε, and x u,e,t represents a binary decision variable, indicating whether the stream data computing task of user u is executed by edge server e at time slot t. If x u,e,t =1, then the stream data computing task is executed by the edge server, otherwise the stream data computing task is not executed by the edge server. is the sum of the delays perceived by user u at the beginning of time slot t, The value is the network transmission delay Propagation Delay And the computational transmission delay of the edge server The sum of represents the average delay of all users in time slot t, x u,e,t-1 Indicates the decision result of whether the stream data computing task of user u in the previous time slot is assigned to edge server e. is the delay value when user u reconnects with edge server e, ω is the importance of the reconnection loss caused by task migration relative to the synchronization state; Deploy decision-making application modules to form a consensus on the cloud, edge, and end, execute decisions on the redistribution of stream data computing tasks in each time slot, and migrate user task instances.

Images