Self-balancing group control method based on distributed cloud edge cooperation technology

Abstract

The invention relates to the field of Internet of Things. The self-balancing group control method based on the distributed cloud edge cooperation technology comprises the following steps that S1, a system architecture is designed and built, an overall architecture of cloud edge cooperation is designed, and role positioning and function division of a cloud control center and edge computing nodes are defined; the stability of edge end equipment control is guaranteed through the cloud edge coordination technology, the system has good robustness under the extreme condition, the dynamic load balancing algorithm based on the cloud edge coordination characteristic is developed, the algorithm can comprehensively consider multiple factors such as node performance, position, task characteristic and network condition, and the dynamic load balancing algorithm has good robustness. The task allocation and migration strategy can be intelligently determined according to the received edge node state information; and creating a task scheduling module, and distributing the task from the cloud to a proper edge node according to a result of the load balancing algorithm, or migrating the task between the edge nodes to realize load balancing.

Description

Technical Field

[0001] The present invention relates to a cloud-edge collaboration technology, and more specifically to a self-balancing group control method based on distributed cloud-edge collaboration technology. Background Art

[0002] With the development of technologies such as 5G and IoT, edge computing has gradually become an important means to process massive amounts of real-time data and reduce network latency. However, edge node resources are limited and widely distributed, and effective resource scheduling and load balancing strategies are urgently needed. Cloud-edge collaboration is a new computing model that extends cloud computing to the edge. By integrating the powerful computing power of the cloud and the local advantages of edge nodes, data can be processed locally and optimized globally. Existing cloud-edge collaboration technologies rely on efficient network connections, but due to factors such as geographical distribution and network instability, they will affect the processing effect of high-real-time services; the efficiency of cross-domain and heterogeneous cloud-edge resource collaboration is not high, and when the capabilities of edge devices are uneven, it is difficult to achieve seamless integration and collaborative operations. Summary of the invention

[0003] The main technical problem solved by the present invention is to provide a self-balancing group control method based on distributed cloud-edge collaboration technology, which can solve the problem.

[0004] To solve the above technical problems, according to one aspect of the present invention, more specifically, a self-balancing group control method based on distributed cloud-edge collaboration technology comprises the following steps:

[0005] S1. Design and build the system architecture, design the overall architecture of cloud-edge collaboration, and clarify the role positioning and functional division of the cloud control center and edge computing nodes;

[0006] S2, resource monitoring and information reporting, regular or real-time collection and reporting of CPU usage, memory usage, network bandwidth, storage space, etc.;

[0007] S3, development and implementation of dynamic load balancing algorithm, comprehensively considering multiple factors such as node performance, location, task characteristics and network conditions, to determine the task allocation and migration strategy;

[0008] S4, task scheduling and migration, tasks are allocated from the cloud to appropriate edge nodes, or tasks are migrated between edge nodes to achieve load balancing;

[0009] S5. Edge node autonomy and collaborative processing. When the connection with the cloud is lost or the network is unstable, the edge node can temporarily store and process tasks, and then synchronize the status and results after the network is restored.

[0010] Furthermore, in step S1, group scheduling and control of edge devices are achieved through cloud-edge collaboration technology to ensure the rationality of resource allocation.

[0011] Furthermore, in step S2, by deploying the corresponding software stack on the edge node, it is enabled to have the ability to process local tasks and maintain communication with the cloud; the resource monitoring module is installed and run on the edge node to collect and report key resource indicators such as CPU usage, memory usage, network bandwidth, storage space, etc. regularly or in real time.

[0012] Furthermore, in step S3, when the cloud sends a control instruction to the edge server cluster, the probability of a certain edge being selected is p i According to the control response time, edge connection number, control success rate and other indicators, p i The calculation formula is:

[0013]

[0014] Among them, w1, w2, w3 represent weight coefficients, T0 represents the standard response time, T i represents the average control response time of edge i, N0 represents the maximum number of edge connections, and N i It indicates the number of connections of edge i when the control instruction is issued, C0 indicates the number of edge control times, and C i Indicates the number of successful edge control. And w1, w2, w3 satisfy:

[0015] w1+w2+w3=1.

[0016] Furthermore, in step S4, the coordination between edge nodes is aimed at minimizing the network loss of the cloud master station, and the control signal of the entire edge node is controlled by the cloud control center. In order to select a scientific weight coefficient, the linear regression algorithm in machine learning is used to solve the optimal weight coefficients w1, w2, and w3. Expressed as feature vectors x1, x2, x3, then:

[0017] p=w1x1+w2x2+w3x3

[0018] In matrix form:

[0019] p=w T x

[0020] Using the least squares method to estimate these weight coefficients, the loss function is:

[0021] y (i) Represents the true output value of the i-th sample. After obtaining the weight coefficients, when there is a new control request, the probability of all edges being selected is calculated, and the one with the largest probability is selected for control.

[0022] Furthermore, in the step S4, the task scheduling allocation is performed by first selecting the dominant node of the edge node and performing a controllability calculation. The controllability calculation indicators may include network response delay, edge node memory, CPU and other indicators, and these attributes are weightedly calculated, and then the control target is determined, mainly with the minimum control delay and the fastest edge response as the control target, and the indicators and of all edge nodes are calculated, and then the constraints for coordinated control between edge nodes, the total control signal volume, the control response time, resource consumption, etc. are determined according to the control target, and finally the control task allocation between edge nodes is determined according to the dynamic programming results.

[0023] Furthermore, in step S5, the edge autonomous control strategy is that the control signal sent from the cloud is temporarily stored in the edge center node. This strategy needs to be enabled when the network connection between the edge and the cloud is unstable or disconnected.

[0024] Furthermore, the edge autonomous control strategy is: first, a weighted calculation is performed based on the memory, CPU, and network conditions of all edge nodes to automatically select the edge center as a temporary task allocation point; then the control objectives and constraints are determined, and the judgment of cloud reconnection or network improvement must be added here to determine when the edge center node returns the task scheduling rights to the cloud center, and then the control tasks are allocated according to the scheduling results. Due to unstable or interrupted cloud connections, they are reported to the cloud after the connection is restored.

[0025] The beneficial effects of the self-balancing group control method based on distributed cloud-edge collaboration technology of the present invention are:

[0026] By setting up cloud-edge collaboration technology to ensure the stability of edge device control, the system has better robustness in extreme cases; build a cloud control center to be responsible for the acquisition of global views, resource scheduling decisions, task allocation and other management tasks, and deploy the corresponding software stack on the edge nodes to enable them to process local tasks and maintain communication with the cloud; install and run resource monitoring modules on edge nodes to collect and report key resource indicators such as CPU usage, memory usage, network bandwidth, and storage space regularly or in real time; develop a dynamic load balancing algorithm based on cloud-edge collaboration characteristics, which should be able to comprehensively consider multiple factors such as node performance, location, task characteristics, and network conditions, so that it can intelligently determine the task allocation and migration strategy based on the received edge node status information; create a task scheduling module to allocate tasks from the cloud to appropriate edge nodes according to the results of the load balancing algorithm, or migrate tasks between edge nodes to achieve load balancing; design an event-driven mechanism for cloud-edge collaboration, which can temporarily store and process tasks when the connection with the cloud is lost or the network is unstable, and synchronize the status and results after the network is restored. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] The present invention is further described in detail below with reference to the accompanying drawings and specific implementation methods.

[0028] Figure 1 It is a schematic diagram of the overall structure of a self-balancing group control method based on distributed cloud-edge collaboration technology of the present invention. DETAILED DESCRIPTION

[0029] The present invention will be described in detail below with reference to the accompanying drawings and in combination with embodiments. It should be noted that the embodiments and features in the embodiments of the present application can be combined with each other without conflict.

[0030] According to one aspect of the present invention, Figure 1 As shown, a self-balancing group control method based on distributed cloud-edge collaboration technology is provided, which includes the following steps:

[0031] S1. Design and build system architecture, design the overall architecture of cloud-edge collaboration, clarify the role positioning and functional division of the cloud control center and edge computing nodes; realize group scheduling and control of edge devices through cloud-edge collaboration technology to ensure the rationality of resource allocation.

[0032] S2. Resource monitoring and information reporting: regularly or in real time collect and report CPU usage, memory usage, network bandwidth, storage space, etc.; deploy corresponding software stacks on edge nodes to enable them to handle local tasks and maintain communication with the cloud; install and run resource monitoring modules on edge nodes to regularly or in real time collect and report key resource indicators such as CPU usage, memory usage, network bandwidth, storage space, etc.

[0033] S3. Research and development and implementation of dynamic load balancing algorithms, comprehensively considering multiple factors such as node performance, location, task characteristics, and network conditions, to determine task allocation and migration strategies; calculate the probability of all edges being selected, and select the one with the highest probability for control, and determine the control target, mainly with the minimum control delay and the fastest edge response as the control target.

[0034] S4, task scheduling and migration, tasks are allocated from the cloud to appropriate edge nodes, or tasks are migrated between edge nodes to achieve load balancing; the main control objectives are to minimize control delay and fastest edge response, and calculate the indicators and of all edge nodes, and then determine the constraints for coordinated control between edge nodes based on the control objectives, the total control signal volume, the control response time, resource consumption, etc., and finally determine the control task allocation between edge nodes based on the dynamic planning results.

[0035] S5, edge node autonomy and collaborative processing, when the connection with the cloud is lost or the network is unstable, it can temporarily store and process tasks, and synchronize the status and results after the network is restored; edge autonomous control strategy, the control signal sent by the cloud is temporarily stored in the edge center node, this strategy needs to be enabled when the network connection between the edge and the cloud is unstable or disconnected.

[0036] The working principle of this device is: when the cloud sends a control instruction to the edge server cluster, the probability of a certain edge being selected is p i , according to indicators such as control response time, number of edge connections, and control success rate, it can be concluded pi The calculation formula is:

[0037]

[0038] Among them, w1, w2, w3 represent weight coefficients, T0 represents the standard response time, T i represents the average control response time of edge i, N0 represents the maximum number of edge connections, and N i It indicates the number of connections of edge i when the control instruction is issued, C0 indicates the number of edge control times, and C i Indicates the number of successful edge control. And w1, w2, w3 satisfy:

[0039] w1+w2+w3=1

[0040] In order to select scientific weight coefficients, the linear regression algorithm in machine learning is used to solve the optimal weight coefficients w1, w2, and w3. Expressed as feature vectors x1, x2, x3, then:

[0041] p=w1x1+w2x2+w3x3

[0042] Abbreviated to matrix form p = w T x, use the least squares method to estimate these weight coefficients, then the loss function

[0043]

[0044] Among them, m represents the number of training samples, h w (x (i) ) represents the model’s predicted value for the i-th sample, y (i) Represents the true output value of the i-th sample. After obtaining the weight coefficients, when there is a new control request, the probability of all edges being selected is calculated, and the one with the largest probability is selected for control.

[0045] The electrical components that appear in this article are all electrical components that exist in reality.

[0046] Of course, the above description is not a limitation of the present invention, and the present invention is not limited to the above examples. Changes, modifications, additions or substitutions made by ordinary technicians in this technical field within the essential scope of the present invention also fall within the protection scope of the present invention.

Claims

1. A self-balancing group control method based on distributed cloud-edge collaboration technology, characterized by: The following steps are involved: S1. Design and build the system architecture, design the overall architecture of cloud-edge collaboration, and clarify the role positioning and functional division of the cloud control center and edge computing nodes; S2, resource monitoring and information reporting, regular or real-time collection and reporting of CPU usage, memory usage, network bandwidth, storage space, etc.; S3, development and implementation of dynamic load balancing algorithm, comprehensively considering multiple factors such as node performance, location, task characteristics and network conditions, to determine the task allocation and migration strategy; S4, task scheduling and migration, tasks are allocated from the cloud to appropriate edge nodes, or tasks are migrated between edge nodes to achieve load balancing; S5. Edge node autonomy and collaborative processing. When the connection with the cloud is lost or the network is unstable, it can temporarily store and process tasks, and then synchronize the status and results after the network is restored.

2. According to claim 1, a self-balancing group control method based on distributed cloud-edge collaboration technology is characterized by: In step S1, group scheduling and control of edge devices are implemented through cloud-edge collaboration technology to ensure the rationality of resource allocation.

3. According to claim 1, a self-balancing group control method based on distributed cloud-edge collaboration technology is characterized by: In step S2, the corresponding software stack is deployed on the edge node to enable it to have the ability to process local tasks and maintain communication with the cloud; the resource monitoring module is installed and run on the edge node to collect and report key resource indicators such as CPU usage, memory usage, network bandwidth, storage space, etc. regularly or in real time.

4. According to claim 1, a self-balancing group control method based on distributed cloud-edge collaboration technology is characterized by: In step S3, when the cloud sends a control instruction to the edge server cluster, the probability that a certain edge is selected is p i According to the control response time, edge connection number, control success rate and other indicators, p i The calculation formula is: Among them, w1, w2, w3 represent weight coefficients, T0 represents the standard response time, T i represents the average control response time of edge i, N0 represents the maximum number of edge connections, and N i It indicates the number of connections of edge i when the control instruction is issued, C0 indicates the number of edge control times, and C i Indicates the number of successful edge control. And w1, w2, w3 satisfy: w1+w2+w3=1.

5. According to claim 1, a self-balancing group control method based on distributed cloud-edge collaboration technology is characterized by: In step S3, the coordination between edge nodes aims to minimize the network loss of the cloud master station, and the control signal of the entire edge node is controlled by the cloud control center. In order to select a scientific weight coefficient, the linear regression algorithm in machine learning is used to solve the optimal weight coefficients w1, w2, and w3. Expressed as feature vectors x1, x2, x3, then: p=w1x1+w2x2+w3x3 In matrix form: p=w T x Using the least squares method to estimate these weight coefficients, the loss function is: Among them, m represents the number of training samples, h w (x (i) ) represents the model’s predicted value for the i-th sample, y (i) Represents the true output value of the i-th sample. After obtaining the weight coefficients, when there is a new control request, the probability of all edges being selected is calculated, and the one with the largest probability is selected for control.

6. According to claim 1, a self-balancing group control method based on distributed cloud-edge collaboration technology is characterized by: In the step S4, the task scheduling allocation is performed by first selecting the dominant node of the edge node and performing controllability calculation. The controllability calculation indicators may include network response delay, edge node memory, CPU and other indicators, and these attributes are weightedly calculated, and then the control target is determined, mainly with the minimum control delay and the fastest edge response as the control target, and the indicators and of all edge nodes are calculated, and then the constraints of coordinated control between edge nodes, the total control signal quantity, the control response time, resource consumption, etc. are determined according to the control target, and finally the control task allocation between edge nodes is determined according to the dynamic programming results.

7. According to claim 1, a self-balancing group control method based on distributed cloud-edge collaboration technology is characterized by: In step S5, the edge autonomous control strategy is used, and the control signal sent from the cloud is temporarily stored in the edge center node. This strategy needs to be enabled when the network connection between the edge and the cloud is unstable or disconnected.

8. The self-balancing group control method based on distributed cloud-edge collaboration technology according to claim 7 is characterized by: The edge autonomous control strategy is: first, weighted calculation is performed based on the memory, CPU, and network conditions of all edge nodes to automatically select the edge center as the temporary task allocation point; then the control objectives and constraints are determined. Here, the judgment of cloud reconnection or network improvement must be added to determine when the edge center node returns the task scheduling right to the cloud center, and then the control task is allocated according to the scheduling result. Due to unstable or interrupted cloud connection, it will be reported to the cloud after the connection is restored.

Images