CN-121979668-A - Dynamic resource scheduling method for edge network cloud

CN121979668ACN 121979668 ACN121979668 ACN 121979668ACN-121979668-A

Abstract

The invention discloses a dynamic resource scheduling method for an edge network cloud, and relates to the field of edge computing. The dynamic resource scheduling method for the edge network cloud comprises the following steps of S1, data acquisition, S2, multi-dimensional evaluation model construction, S3, layered scheduling judgment, S4, scheduling execution and dynamic adjustment. According to the edge network cloud-oriented dynamic resource scheduling method, the edge nodes, the tasks and the cloud center multidimensional data are collected in real time, the dynamic weight resource supply and demand matching degree evaluation model is constructed, and scheduling is judged and judged according to the layers of edge local, edge cooperative and cloud center, so that the problem that the existing scheme only uses the one-sided performance of CPU/memory evaluation, the resource supply and demand matching degree evaluation error is reduced, the time delay standard reaching rate of a time delay sensitive task is improved, and the core scheduling requirement of low time delay and high accuracy of an edge network cloud architecture is met.

Inventors

JING FENG
ZHENG KOUQUAN
ZHANG YIJUN
LI MO
ZHAO LE
SHI YOUWEI
XUE YAO

Assignees

中国人民解放军国防科技大学

Dates

Publication Date: 20260505
Application Date: 20251224

Claims (10)

1. The dynamic resource scheduling method for the edge network cloud is characterized by comprising the following steps of: S1, data acquisition, namely acquiring resource state data of an edge node in an edge network cloud architecture, demand data of a task to be scheduled and resource redundancy data of a cloud center in real time; S2, constructing a multi-dimensional evaluation model, namely constructing a resource supply and demand matching degree evaluation model based on the data acquired in the step S1, wherein the model takes the edge node load rate, the task time delay requirement and the cloud edge link bandwidth as core parameters, and dynamically adjusts the weight of each parameter; S3, performing hierarchical scheduling judgment, namely executing three-layer scheduling judgment logic according to the evaluation result of the step S2; 3.1, judging in a first layer, if the resource supply and demand matching degree of the edge node is more than or equal to a first preset threshold value, scheduling a task to be executed locally by the edge node; 3.2, judging in a second layer, if the resource supply and demand matching degree of the edge nodes is less than a first preset threshold value, and the resource supply and demand matching degree of the peripheral edge node cluster is more than or equal to a second preset threshold value, starting cooperative scheduling among the edge nodes, and distributing tasks to the peripheral edge nodes; 3.3, judging a third layer, namely if the resource supply and demand matching degree of the edge node and the peripheral clusters is smaller than the corresponding preset threshold value and the resource supply and demand matching degree of the cloud center is larger than or equal to the third preset threshold value, scheduling the task to the cloud center for execution; s4, scheduling execution and dynamic adjustment, namely executing the scheduling scheme determined in the step S3, monitoring the resource state and time delay data in the task execution process in real time, and returning to the step S2 to execute the evaluation and scheduling again if the monitored data exceeds a preset range.
2. The method for dynamic resource scheduling for edge network cloud according to claim 1, wherein in step S1: The resource state data of the edge node comprises CPU utilization rate, memory utilization rate, storage residual capacity and real-time network bandwidth; the demand data of the task to be scheduled comprises a task time delay threshold value, a calculation force demand value and a data transmission quantity; The resource redundancy data of the cloud center comprises CPU redundancy rate, memory redundancy rate and cloud side link available bandwidth.
3. The method for dynamic resource scheduling for edge network cloud according to claim 1, wherein the method for dynamically adjusting the parameter weight in step S2 is as follows: When the task to be scheduled is a time delay sensitive task, the weight of the parameter of 'task time delay requirement' is increased to 30% -50%; when the fluctuation amplitude of the edge node load is more than 15%, the weight of the 'edge node load rate' parameter is improved to 25% -40%, and the weight adjustment is optimized by adopting a random forest algorithm.
4. The method for dynamic resource scheduling for edge network cloud according to claim 1, wherein in step S3: the first preset threshold value is 80% -90%, and the threshold value is dynamically updated every week according to historical load fluctuation data of the edge node for about 72 hours; The first layer of judgment further comprises confirming local scheduling if the estimated time delay of the locally executed task of the edge node is less than or equal to the task time delay threshold value, and entering the second layer of judgment otherwise.
5. The dynamic resource scheduling method for the edge network cloud according to claim 1, wherein the specific process of the second layer judgment in the step S3 is as follows: adopting a K nearest neighbor algorithm to screen peripheral edge nodes which are less than or equal to 5km from the current edge node to form an edge cooperative cluster; And if the second preset threshold value is 65-75%, calculating the overall resource supply and demand matching degree of the cluster, and if the transmission delay between edge nodes and the cooperative execution delay are not less than the second preset threshold value and not more than the task delay threshold value, distributing tasks to 2-3 edge nodes with the highest resource redundancy rate in the cluster.
6. The dynamic resource scheduling method for the edge network cloud according to claim 1, wherein the specific process of the third layer judgment in the step S3 is: the third preset threshold value is 70% -80%, and the available bandwidth of the cloud side link is more than or equal to the task data transmission amount/task allowed transmission duration when judging; And if a plurality of available resource pools exist in the cloud center, preferentially scheduling the resource pools with idle time length of more than 2 hours to execute tasks.
7. The method for dynamic resource scheduling for edge network cloud according to claim 1, wherein the triggering condition dynamically adjusted in step S4 comprises: the real-time load rate of the edge node is more than 90%; The actual execution time delay of the task is greater than 110% of the time delay threshold of the task; the cloud-edge link bandwidth drops by >30% and lasts for >5 minutes.
8. The method for dispatching dynamic resources facing to the edge network cloud according to claim 1, further comprising a security authentication step, wherein before dispatching is executed in step S3, identity authentication is carried out on a task initiating terminal, task data and dispatching instructions are encrypted by adopting an AES-256 algorithm, and dispatching is executed after authentication and decryption are successful.
9. The dynamic resource scheduling method for the edge network cloud according to claim 5, wherein the resource allocation of the cooperative scheduling among the edge nodes adopts a particle swarm optimization algorithm, and takes 'minimizing cooperative execution time delay' as an objective function, wherein the constraint condition is that the load rate of each edge node is less than or equal to 85%.
10. The dynamic resource scheduling method for the edge network cloud according to claim 1, further comprising a scheduling result feedback optimization step of recording the resource utilization rate, the task completion rate and the time achievement rate of each scheduling, and optimizing the parameter weight of the evaluation model in step S2 and each preset threshold in step S3 by adopting a gradient descent algorithm based on historical data, wherein the optimization period is 1 time per day.

Description

Dynamic resource scheduling method for edge network cloud Technical Field The invention relates to the field of edge computing, in particular to a dynamic resource scheduling method for an edge network cloud. Background The existing edge network cloud resource scheduling method has the following technical defects: most schemes only take the CPU/memory utilization rate of the edge node as a scheduling basis, and ignore key parameters such as task delay requirement, cloud edge link bandwidth and the like, so that 'resource matching deviation' is caused, for example, a delay sensitive task is scheduled to an edge node with low load but long distance, and execution delay exceeding is caused. The threshold value and the weight are static, namely, the threshold value (such as the upper limit of the load of the edge node) of the scheduling judgment and the parameter weight (such as the importance duty ratio of the time delay requirement) are fixed values, so that the method can not adapt to the dynamic switching of the time delay sensitive/computation intensive task caused by the fluctuation of the load of the edge node (such as the difference of 40% of the peak load in the morning and evening), and the scheduling flexibility is insufficient. The collaborative scheduling mechanism is imperfect, the existing edge collaborative multi-consideration only considers a single node, clustered collaborative logic is not formed, the resource allocation lacks the support of an optimization algorithm, imbalance problems of partial node overload and partial node idle are easy to occur, and meanwhile, the cloud center scheduling does not combine the idle duration of a resource pool and the constraint of a link bandwidth, so that the resource utilization rate is low. The safety and dynamic adjustment are lacking, most schemes do not integrate identity authentication and data encryption, the scheduling instruction is tampered and task data leakage risks exist, a real-time monitoring and dynamic adjustment mechanism is lacking after scheduling, and when the load of an edge node is suddenly increased and the bandwidth of a link is suddenly reduced, the task cannot be timely rescheduled, so that the task failure rate is increased. Disclosure of Invention The invention aims to provide a dynamic resource scheduling method for an edge network cloud, which aims to solve the problems in the background technology. In order to achieve the purpose, the invention provides the following technical scheme that the dynamic resource scheduling method for the edge network cloud comprises the following steps: S1, data acquisition, namely acquiring resource state data of an edge node in an edge network cloud architecture, demand data of a task to be scheduled and resource redundancy data of a cloud center in real time; S2, constructing a multi-dimensional evaluation model, namely constructing a resource supply and demand matching degree evaluation model based on the data acquired in the step S1, wherein the model takes the edge node load rate, the task time delay requirement and the cloud edge link bandwidth as core parameters, and dynamically adjusts the weight of each parameter; S3, performing hierarchical scheduling judgment, namely executing three-layer scheduling judgment logic according to the evaluation result of the step S2; 3.1, judging in a first layer, if the resource supply and demand matching degree of the edge node is more than or equal to a first preset threshold value, scheduling a task to be executed locally by the edge node; 3.2, judging in a second layer, if the resource supply and demand matching degree of the edge nodes is less than a first preset threshold value, and the resource supply and demand matching degree of the peripheral edge node cluster is more than or equal to a second preset threshold value, starting cooperative scheduling among the edge nodes, and distributing tasks to the peripheral edge nodes; 3.3, judging a third layer, namely if the resource supply and demand matching degree of the edge node and the peripheral clusters is smaller than the corresponding preset threshold value and the resource supply and demand matching degree of the cloud center is larger than or equal to the third preset threshold value, scheduling the task to the cloud center for execution; s4, scheduling execution and dynamic adjustment, namely executing the scheduling scheme determined in the step S3, monitoring the resource state and time delay data in the task execution process in real time, and returning to the step S2 to execute the evaluation and scheduling again if the monitored data exceeds a preset range. Preferably, in the step S1: The resource state data of the edge node comprises CPU utilization rate, memory utilization rate, storage residual capacity and real-time network bandwidth; the demand data of the task to be scheduled comprises a task time delay threshold value, a calculation force demand value and a data transmission q