Edge computing resource allocation management system and control method thereof

Abstract

The invention relates to the technical field of resource allocation management, in particular to an edge computing resource allocation management system and a control method thereof, and the method comprises the steps: S1, task receiving and priority division: receiving a task request of a calculation amount, a data amount, a time delay constraint and a resource demand uploaded by an edge device; according to the method, tasks are divided in batches according to task attributes, the tasks are sorted according to priorities, and the tasks which are urgent to resource demands and have a small number of feasible modes are preferentially processed. In the method, the future load state is evaluated in real time through a time sequence prediction model, and the task priority sorting mechanism of dynamic adjustment is combined; the task allocation strategy is optimized, and the scheduling flexibility is further improved by introducing the task delay relaxation factor, so that the system can reasonably allocate high-priority tasks under the condition of resource shortage, and the resource utilization rate and the scheduling efficiency of the system are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of resource allocation management, and specifically to an edge computing resource allocation management system and a control method thereof. Background Art

[0002] Edge computing resource allocation management system refers to a management system that allocates computing and storage resources at edge nodes close to the user end (such as edge servers or mobile devices);

[0003] The system is designed to effectively utilize the computing, storage and communication resources of edge nodes to handle delay-sensitive tasks while reducing the burden on the central cloud computing platform. Compared with traditional cloud computing, edge computing can reduce transmission delays and improve task completion efficiency by completing task calculations on nodes close to the data source. It is also suitable for scenarios with high real-time requirements, such as smart transportation, industrial Internet of Things and virtual reality applications.

[0004] The existing edge computing resource allocation and management system lacks intelligent and real-time optimization capabilities in resource scheduling when facing complex scenarios. It is difficult to efficiently allocate edge node resources according to dynamically changing load states and task attributes, resulting in increased task delays and excessive energy consumption. Therefore, an edge computing resource allocation and management system and its control method are proposed to address the above problems. Summary of the invention

[0005] The purpose of the present invention is to provide an edge computing resource allocation management system and a control method thereof, so as to solve the problem that the existing edge computing resource allocation management system lacks intelligent and real-time optimization capabilities in resource scheduling when facing complex scenarios, and it is difficult to efficiently allocate edge node resources according to dynamically changing load states and task attributes, resulting in increased task delays and excessive energy consumption.

[0006] To achieve the above object, the present invention provides the following technical solutions:

[0007] An edge computing resource allocation management system and a control method thereof, comprising:

[0008] S1: Task reception and priority division: Receive task requests uploaded by edge devices with computing volume, data volume, latency constraints, and resource requirements;

[0009] Divide tasks into batches according to their attributes and sort them by priority, giving priority to tasks with urgent resource requirements and a small number of feasible modes;

[0010] The objective function of task allocation is:

[0011]

[0012] In the formula, x ij∈{0, 1} indicates whether task i is assigned to node j, C ij =w 1 T ij +w 2 E ij represents the completion time T of the task on node j ij and energy consumption E ij The weighted cost, w 1 and w 2 It is a dynamically adjusted weight factor used to balance latency and energy consumption;

[0013] S2: Resource status monitoring and allocation:

[0014] S21: Real-time monitoring of the available computing resources, storage resources, and communication channel status of edge nodes, and obtaining the current load status L(t), remaining computing capacity R comp and channel bandwidth state R comm data;

[0015] S22: Based on the acquired load state L(t) and remaining computing capacity R comp and channel bandwidth state R comm Data, to assess and predict future resource needs;

[0016] S23: dynamically adjust the task allocation strategy according to the resource status and task priority to provide input for subsequent offloading optimization;

[0017] S3: Dynamic task offloading strategy optimization: Combined with real-time monitoring data, deep reinforcement learning algorithm is used to optimize the task offloading strategy;

[0018] The optimization goal is to minimize delays and energy consumption by maximizing task completion rate;

[0019] S4: Load balancing adjustment and task migration: After completing the initial task allocation, regularly evaluate the load status of edge nodes and then adjust the allocation strategy;

[0020] Through task migration, tasks on high-load nodes are reallocated to low-load nodes to balance the overall resource usage of the system;

[0021] S5: Result feedback and iterative optimization: Real-time feedback of resource allocation and offloading optimization results based on task completion;

[0022] Re-prioritize the tasks that do not meet the allocation conditions and assign them to the next batch for processing;

[0023] Update the weight factor of the task allocation objective function to achieve a dynamic balance between latency and energy consumption. The specific optimization goal is:

[0024]

[0025] In the formula, w 1 and w 2 Dynamically adjusted weight factor, T ij is the task completion time, E ij Energy consumption for task execution;

[0026] Through multiple iterative optimizations, the task completion rate and resource utilization efficiency are gradually improved.

[0027] As a further optimized content of the present invention, the formula for evaluating and predicting future resource demand is:

[0028]

[0029] In the formula, It represents the predicted load state at the next moment, and the function f is constructed by the time series prediction model.

[0030] As a further optimized content of the present invention, wherein: the unloading strategy is determined by the following formula:

[0031]

[0032] In the formula, s t represents the current system status, including resource status, task attributes and historical uninstallation records, a is the candidate action for uninstallation mode, Q(s t ,a) means in state s t The long-term benefit obtained by choosing action a.

[0033] As a further optimization of the present invention, the adjustment allocation strategy in S4 is:

[0034]

[0035] Where U i (x i ,x -i ) is the utility function of task i, R i (x i ) is the benefit of task i, including the improvement of delay and energy consumption, C i (x i ) is the allocation cost of task i, x -i Indicates the allocation strategy for other tasks.

[0036] As further optimized content of the present invention, among others: during the task receiving and dividing process, after the delay relaxation factor is introduced, the system introduces a dynamic adjustment mechanism for different batches of tasks, and adjusts the range of the relaxation factor by real-time evaluation of the actual occupancy of resources by the tasks, so that high-priority tasks can still obtain priority scheduling when resources are tight, while reducing the impact of low-priority tasks on the system.

[0037] As further optimized content of the present invention, among them: in the result feedback and iterative optimization steps, the system establishes a task execution record library according to the allocation and execution of historical tasks, and performs multi-dimensional evaluation based on the task completion rate, actual delay and energy consumption, to provide data support for the priority sorting and resource allocation strategy of subsequent tasks, thereby optimizing the decision-making accuracy in the iterative process.

[0038] As a further optimized content of the present invention, wherein: the system comprises:

[0039] Task management module: used to receive edge computing task requests, batch and prioritize tasks based on task uncertainty (task release time, task quantity, and resource requirements);

[0040] Resource monitoring module: monitors the computing resources, storage resources and channel status of edge computing nodes in real time, providing real-time data support for task allocation;

[0041] Task offloading module: Determines the task offloading mode based on task priority and optimization algorithm, including local execution, collaborative device execution, and offloading to MEC server;

[0042] Optimization control module: optimizes task allocation and resource allocation through adaptive strategy selection algorithm based on load balancing to avoid resource overload or channel congestion;

[0043] Load balancing module: used for dynamic task migration between edge nodes and load balancing of MEC servers;

[0044] Communication interface module: realizes efficient communication between edge computing nodes, supports task offloading, resource status synchronization and feedback information exchange.

[0045] As a further optimized content of the present invention, the resource monitoring module obtains the computing resources, storage resources and channel status of the edge node in real time, and dynamically adjusts the task allocation plan according to the status difference between the nodes.

[0046] Compared with the prior art, the present invention has the following beneficial effects:

[0047] 1. In the present invention, the future load state is evaluated in real time through the time series prediction model, and the task allocation strategy is optimized in combination with the dynamically adjusted task priority sorting mechanism. The introduction of the task delay relaxation factor further improves the scheduling flexibility, so that the system can reasonably allocate high-priority tasks under the condition of resource constraints, thereby improving the resource utilization and scheduling efficiency of the system;

[0048] 2. In the present invention, a deep reinforcement learning algorithm is used to dynamically decide the task offloading mode based on comprehensive consideration of task attributes, resource status and historical records. By maximizing the optimization goal of long-term benefits, the task completion rate is improved, the delay and energy consumption are significantly reduced, and the intelligence level and reliability of offloading decisions are improved;

[0049] 3. In the present invention, by adjusting the load balancing strategy, the bottleneck problem of high-load nodes can be effectively identified and solved, and tasks can be migrated from high-load nodes to low-load nodes to achieve balanced utilization of global resources. At the same time, combined with the optimization of inter-node communication costs, the transmission delay that may occur during task migration is greatly reduced, ensuring the stability and efficiency of the overall system performance;

[0050] 4. In the present invention, a task execution record library and a multi-dimensional evaluation mechanism are introduced to record and analyze key indicators such as task completion rate, delay and energy consumption. These data provide a decision-making basis for subsequent task scheduling. Through iterative optimization strategies, the task allocation efficiency is continuously improved, and the adaptability and stability of the system are gradually improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] Figure 1 This is a system block diagram of an edge computing resource allocation management system of the present invention;

[0052] Figure 2 This is a control flow chart of an edge computing resource allocation management system of the present invention. DETAILED DESCRIPTION

[0053] See also Figure 1-2 , the present invention provides a technical solution:

[0054] An edge computing resource allocation management system and control method thereof, comprising: S1: task reception and priority division: receiving a task request with computing amount, data amount, delay constraint and resource requirement uploaded by an edge device;

[0055] Divide tasks into batches according to their attributes and sort them by priority, giving priority to tasks with urgent resource requirements and a small number of feasible modes;

[0056] The objective function of task allocation is:

[0057]

[0058] In the formula, x ij ∈{0, 1} indicates whether task i is assigned to node j, C ij =w 1 T ij +w 2 E ij represents the completion time T of the task on node j ij and energy consumption E ij The weighted cost, w 1 and w 2 It is a dynamically adjusted weight factor used to balance latency and energy consumption;

[0059] S2: Resource status monitoring and allocation:

[0060] S21: Real-time monitoring of the available computing resources, storage resources, and communication channel status of edge nodes, and obtaining the current load status L(t), remaining computing capacity R comp and channel bandwidth state R comm data;

[0061] S22: Based on the acquired load state L(t) and remaining computing capacity R comp and channel bandwidth state R comm Data, to assess and predict future resource needs;

[0062] S23: dynamically adjust the task allocation strategy according to the resource status and task priority to provide input for subsequent offloading optimization;

[0063] S3: Dynamic task offloading strategy optimization: Combined with real-time monitoring data, deep reinforcement learning algorithm is used to optimize the task offloading strategy;

[0064] The optimization goal is to minimize delays and energy consumption by maximizing task completion rate;

[0065] S4: Load balancing adjustment and task migration: After completing the initial task allocation, regularly evaluate the load status of edge nodes and then adjust the allocation strategy;

[0066] Through task migration, tasks on high-load nodes are reallocated to low-load nodes to balance the overall resource usage of the system;

[0067] S5: Result feedback and iterative optimization: Real-time feedback of resource allocation and offloading optimization results based on task completion;

[0068] Re-prioritize the tasks that do not meet the allocation conditions and assign them to the next batch for processing;

[0069] Update the weight factor of the task allocation objective function to achieve a dynamic balance between latency and energy consumption. The specific optimization goal is:

[0070]

[0071] In the formula, w 1 and w 2 Dynamically adjusted weight factor, T ij is the task completion time, E ij Energy consumption for task execution;

[0072] Through multiple iterative optimizations, the task completion rate and resource utilization efficiency are gradually improved.

[0073] As a technical solution for further implementation of this plan, the formula for evaluating and predicting future resource demand is:

[0074]

[0075] In the formula, It represents the predicted load state at the next moment. Function f is constructed by the time series prediction model. The load state is predicted by the time series prediction model, which can identify future changes in resource demand in advance, provide data support for task allocation and resource scheduling, and reduce delays caused by resource competition and load fluctuations. At the same time, the prediction mechanism improves the adaptability of the system in complex task scenarios.

[0076] As a technical solution for further implementing this solution, the uninstallation strategy is determined by the following formula:

[0077]

[0078] In the formula, s t represents the current system status, including resource status, task attributes and historical uninstallation records, a is the candidate action for uninstallation mode, Q(s t ,a) means in state s t The long-term benefits obtained by selecting action a are determined through the deep reinforcement learning algorithm, which combines the system status, task attributes and historical records to dynamically select the best offloading mode, maximize long-term benefits, reduce latency and energy consumption, and optimize the accuracy and reliability of offloading.

[0079] As a technical solution for further implementing this solution, the adjustment allocation strategy in S4 is:

[0080]

[0081] Where U i (x i ,x -i ) is the utility function of task i, Ri (x i ) is the benefit of task i, including the improvement of delay and energy consumption, C i (x i ) is the allocation cost of task i, x -i It represents the allocation strategy of other tasks. By evaluating the utility function of tasks, it realizes the global optimization of resource allocation. This strategy can balance the relationship between delay and energy consumption, avoid high-load nodes from becoming performance bottlenecks, balance resource usage, and improve the overall efficiency of the system.

[0082] As a technical solution for further implementation of this solution, in the process of task reception and division, after introducing the delay relaxation factor, the system introduces a dynamic adjustment mechanism for different batches of tasks, and adjusts the range of the relaxation factor by real-time evaluation of the actual occupation of resources by tasks, so that high-priority tasks can still be prioritized when resources are tight, while reducing the impact of low-priority tasks on the system. By introducing the delay relaxation factor, the task priority is dynamically adjusted, effectively alleviating task conflicts under resource constraints. High-priority tasks are given priority, and the impact of low-priority tasks is minimized, thereby improving the flexibility and fairness of resource scheduling in the system;

[0083] As a technical solution for further implementation of this solution, in the result feedback and iterative optimization steps, the system establishes a task execution record library based on the allocation and execution of historical tasks, and conducts multi-dimensional evaluation based on the completion rate, actual delay and energy consumption of the tasks, providing data support for the priority sorting and resource allocation strategy of subsequent tasks, thereby optimizing the decision-making accuracy in the iterative process, establishing a task execution record library, and evaluating the task allocation effect through multi-dimensional indicators such as completion rate, delay and energy consumption, providing an accurate basis for subsequent priority sorting and resource allocation. This mechanism improves the system's iterative optimization capabilities and decision-making accuracy, and continuously improves task scheduling efficiency;

[0084] As a technical solution for further implementing this solution, the system includes:

[0085] Task management module: used to receive edge computing task requests, batch and prioritize tasks based on task uncertainty (task release time, task quantity, and resource requirements);

[0086] Resource monitoring module: monitors the computing resources, storage resources and channel status of edge computing nodes in real time, providing real-time data support for task allocation;

[0087] Task offloading module: Determines the task offloading mode based on task priority and optimization algorithm, including local execution, collaborative device execution, and offloading to MEC server;

[0088] Optimization control module: optimizes task allocation and resource allocation through adaptive strategy selection algorithm based on load balancing to avoid resource overload or channel congestion;

[0089] Load balancing module: used for dynamic task migration between edge nodes and load balancing of MEC servers;

[0090] Communication interface module: realizes efficient communication between edge computing nodes, supports task offloading, resource status synchronization and feedback information exchange. Through modular design, it clarifies the division of labor among task management, resource monitoring, task offloading, optimization control, load balancing and communication interface, making the system structure clear and the functions coordinated and efficient. The cooperation between modules improves the scalability and stability of the system, and facilitates upgrade and maintenance.

[0091] As a technical solution for further implementing the present scheme, the resource monitoring module obtains the computing resources, storage resources and channel status of the edge nodes in real time, and dynamically adjusts the task allocation scheme according to the status differences between the nodes. The resource monitoring module can obtain the computing resources, storage resources and channel status in real time, and dynamically adjust the task allocation according to the resource differences between the nodes, thereby ensuring the balance of resource utilization and the efficiency of task processing, reducing resource waste and improving the overall performance of the system.

[0092] This article uses specific examples to illustrate the principles and implementation methods of the present invention. The above examples are only used to help understand the method and core ideas of the present invention. The above is only a preferred implementation of the present invention. It should be pointed out that due to the limitations of textual expression and the objective existence of infinite specific structures, ordinary technicians in this technical field can make several improvements, modifications or changes without departing from the principles of the present invention, and can also combine the above technical features in an appropriate manner; these improvements, modifications, changes or combinations, or the direct application of the inventive concept and technical solution to other occasions without improvement, should be regarded as the protection scope of the present invention.

Claims

1. A control method for an edge computing resource allocation management system, characterized in that: include: S1: Task reception and priority division: Receive task requests uploaded by edge devices with computing volume, data volume, latency constraints, and resource requirements; Divide tasks into batches according to their attributes and sort them by priority, giving priority to tasks with urgent resource requirements and a small number of feasible modes; The objective function of task allocation is: In the formula, x ij ∈{0, 1} indicates whether task i is assigned to node j, C ij =w1T ij +w2E ij represents the completion time T of the task on node j ij and energy consumption E ij The weighted cost of, w1 and w2 are dynamically adjusted weight factors used to balance delay and energy consumption; S2: Resource status monitoring and allocation: S21: Real-time monitoring of the available computing resources, storage resources, and communication channel status of edge nodes, and obtaining the current load status L(t), remaining computing capacity R comp and channel bandwidth state R comm data; S22: Based on the acquired load state L(t) and remaining computing capacity R comp and channel bandwidth state R comm Data, to assess and predict future resource needs; S23: dynamically adjust the task allocation strategy according to the resource status and task priority to provide input for subsequent offloading optimization; S3: Dynamic task offloading strategy optimization: Combined with real-time monitoring data, deep reinforcement learning algorithm is used to optimize the task offloading strategy; The optimization goal is to minimize delays and energy consumption by maximizing task completion rate; S4: Load balancing adjustment and task migration: After completing the initial task allocation, regularly evaluate the load status of edge nodes and then adjust the allocation strategy; Through task migration, tasks on high-load nodes are reallocated to low-load nodes to balance the overall resource usage of the system; S5: Result feedback and iterative optimization: Real-time feedback of resource allocation and offloading optimization results based on task completion; Re-prioritize the tasks that do not meet the allocation conditions and assign them to the next batch for processing; Update the weight factor of the task allocation objective function to achieve a dynamic balance between latency and energy consumption. The specific optimization goal is: Where w1 and w2 are dynamically adjusted weight factors, T ij is the task completion time, E ij Energy consumption for task execution; Through multiple iterative optimizations, the task completion rate and resource utilization efficiency are gradually improved.

2. The control method of the edge computing resource allocation management system according to claim 1, characterized in that: The formula for evaluating and predicting future resource demand is: In the formula, It represents the predicted load state at the next moment, and the function f is constructed by the time series prediction model.

3. The control method of the edge computing resource allocation management system according to claim 1, characterized in that: The uninstallation strategy is determined by the following formula: In the formula, s t represents the current system status, including resource status, task attributes and historical uninstallation records, a is the candidate action for uninstallation mode, Q(s t ,a) means in state s t The long-term benefit obtained by choosing action a.

4. The control method of the edge computing resource allocation management system according to claim 1, characterized in that: The adjustment allocation strategy in S4 is: Where U i (x i ,x -i ) is the utility function of task i, R i (x i ) is the benefit of task i, including the improvement of delay and energy consumption, C i (x i ) is the allocation cost of task i, x -i Indicates the allocation strategy for other tasks.

5. The control method of the edge computing resource allocation management system according to claim 1, characterized in that: During the task receiving and dividing process, after the delay relaxation factor is introduced, the system introduces a dynamic adjustment mechanism for tasks in different batches. By real-time evaluation of the actual resource occupancy of the tasks, the range of the relaxation factor is adjusted, so that high-priority tasks can still be given priority scheduling when resources are tight, while reducing the impact of low-priority tasks on the system.

6. The control method of the edge computing resource allocation management system according to claim 1, characterized in that: In the result feedback and iterative optimization steps, the system establishes a task execution record library based on the allocation and execution of historical tasks, and conducts a multi-dimensional evaluation based on the task completion rate, actual delay and energy consumption, providing data support for the priority sorting and resource allocation strategy of subsequent tasks, thereby optimizing the decision-making accuracy in the iterative process.

7. An edge computing resource allocation management system according to any one of claims 1 to 6, characterized in that: The system comprises: Task management module: used to receive edge computing task requests, batch and prioritize tasks based on task uncertainty (task release time, task quantity, and resource requirements); Resource monitoring module: monitors the computing resources, storage resources and channel status of edge computing nodes in real time, providing real-time data support for task allocation; Task offloading module: Determines the task offloading mode based on task priority and optimization algorithm, including local execution, collaborative device execution, and offloading to MEC server; Optimization control module: optimizes task allocation and resource allocation through adaptive strategy selection algorithm based on load balancing to avoid resource overload or channel congestion; Load balancing module: used for dynamic task migration between edge nodes and load balancing of MEC servers; Communication interface module: realizes efficient communication between edge computing nodes, supports task offloading, resource status synchronization and feedback information exchange.

8. An edge computing resource allocation management system according to claim 7, characterized in that: The resource monitoring module obtains the computing resources, storage resources and channel status of the edge nodes in real time, and dynamically adjusts the task allocation scheme according to the status differences between the nodes.

Images