Computer server computing resource allocation system based on edge collaboration

Abstract

The invention discloses a computer server computing resource allocation system based on edge collaboration, and relates to the technical field of resource allocation, the computer server computing resource allocation system comprises a data acquisition module, an environment evaluation module, a load evaluation module, an edge collaboration module and a visualization module, the emigration priority evaluation unit obtains an emigration priority score of the edge node i, and the immigration priority evaluation unit obtains an immigration priority score of the edge node i, so that the attenuation influence of the environment on the performance is quantified through the environment and load evaluation module, the load analysis accuracy is improved, and the load analysis efficiency is improved. Emigration-out and migration-in priorities of node tasks are evaluated based on service importance, user levels, loads and the like, it is ensured that key tasks are migrated preferentially, the system interruption risk is reduced, and resource allocation is optimized; by quantifying the load deviation degree and the dynamic adjustment coefficient and balancing the light load and the performance priority strategy, dynamic requirements can be met, and the overall operation efficiency is improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of resource allocation, in particular to a computer server computing resource allocation system based on edge collaboration. BACKGROUND

[0002] Edge servers are architectures of distributed computing, used to perform edge collaborative computing tasks, and these servers are usually located at the edge of the network, close to data sources and end users, aiming to deploy computing resources at the edge of the network, process and analyze data close to data generation, thereby reducing data transmission delay and providing lower latency and higher real-time performance for data centers.

[0003] Currently, when edge servers allocate resources, they usually use static allocation strategies, without considering changes in edge server environments and load conditions, making it difficult to respond to changing dynamic demands, and lacking analysis of performance differences between different edge servers, which can easily cause weak performance edge servers to be over-allocated tasks, reducing resource utilization and overall operational efficiency, and when handling node overload, the nearest migration strategy is often used, lacking analysis of task priority, increasing the risk of interruption of critical services, and failing to take advantage of edge collaborative computing. SUMMARY

[0004] The purpose of the present application is to provide a computer server computing resource allocation system based on edge collaboration, which solves the problems raised in the background art.

[0005] To achieve the above purpose, the present application provides the following technical solution: a computer server computing resource allocation system based on edge collaboration, comprising a data acquisition module, an environment evaluation module, a load evaluation module, an edge collaboration module and a visualization module.

[0006] The data acquisition module is used to acquire environment data and running data of the server, and the running data includes CPU information, memory information and network usage information.

[0007] The environment evaluation module analyzes the impact of the environment on the performance based on the environment data of each server, and obtains the environment score F env,i of the edge node i.

[0008] The load evaluation module analyzes the CPU information, memory information, network usage information and environment score F env,i of the edge node i, and obtains the load L i of the edge node i.

[0009] The edge collaboration module includes an overload judgment unit, a migration-out priority evaluation unit and a migration-in priority evaluation unit.

[0010] The overload judgment unit is based on the load L of edge node i. i By analyzing the average load of all edge nodes, the load deviation value LD of each edge node i is obtained. i And based on the load deviation value LD of each edge node i i The analysis identified overloaded and lightly loaded nodes.

[0011] The migration priority assessment unit is based on the importance of the service, user level, and the load L of edge node i. i The importance of overloaded nodes is evaluated to obtain the migration priority score P of edge node i. i ;

[0012] The migration priority evaluation unit is based on CPU performance and the load L of edge node i. i The receiving capability of lightly loaded nodes is evaluated to obtain the migration priority score RP of edge node i. i .

[0013] Optionally, the environmental data includes temperature information, humidity information, and energy consumption information. The environmental assessment module performs a weighted fusion analysis based on the temperature information, humidity information, and energy consumption information of each server to obtain the environmental score F of edge node i. env,i This is used to assess the impact of the environment on server performance;

[0014] The environmental score F for edge node i env,i Set environmental thresholds Y1 and Y2, and when the environmental score F of edge node i... env,i <Environmental threshold Y1, or environmental score F of edge node i env,i >When the environmental threshold Y2 is reached, the server is checked.

[0015] Optionally, the load assessment module first normalizes the CPU information, memory information, and network usage information of each server to obtain the normalized CPU utilization C of edge node i. i Memory usage A after normalization of edge node i i The bandwidth utilization of edge node i is normalized to allow for comparison of servers of different specifications;

[0016] Then, the CPU utilization C after normalizing the edge node i is... i Memory usage A after normalization of edge node i i The normalized bandwidth utilization of edge node i is weighted and fused together, while the environmental score F of edge node i is also introduced. env,i The analysis ultimately yielded the load L of edge node i. act,i .

[0017] Optionally, the overload judging unit first calculates the average load of all edge nodes to obtain the total average load L of edge nodes avg ;

[0018] Then, the deviation of the edge node i is analyzed based on the load L i of the edge node i and the total average load L avg of edge nodes, and the deviation of the edge node i is quantitatively analyzed to obtain the load deviation LD i of the edge node i

[0019] The overload threshold Y3 and the light load threshold Y4 are set, when the load deviation LD i of the edge node i is greater than the overload threshold Y3, it is determined that the edge node i is overloaded, and when the load deviation LD i of the edge node i is less than the light load threshold Y4, it is determined that the edge node i is lightly loaded.

[0020] Optionally, the migration-out priority evaluation unit analyzes the edge node determined to be overloaded to determine the task migration-out order of the overloaded edge node

[0021] First, the importance of each edge node for business is analyzed to obtain the importance score G i of the edge node i

[0022] Then, the user level and the highest level are compared to obtain the user level score

[0023] Finally, the importance score G i of the edge node i, the user level score, and the load L i of the edge node i are fused to analyze to obtain the migration-out priority score P i of the edge node i, and the task migration-out order of the overloaded edge node is determined according to the migration-out priority score.

[0024] Optionally, the migration-in priority evaluation unit analyzes the edge node determined to be lightly loaded to determine the task migration-in order of the lightly loaded edge node, and the evaluation process is as follows

[0025]

[0026] RP i is the migration-in priority score of the edge node i

[0027] is the light load score of the edge node i

[0028] θ1 is the light load score influence coefficient

[0029] CH i is the CPU core number of the edge node i

[0030] CH avg is the average CPU core number of all edge nodes;

[0031] represents the performance deviation score of edge node i;

[0032] θ2 is a performance influence coefficient;

[0033] The light load score influence coefficient θ1 and the performance influence coefficient θ2 are both 0.5 in the initial state.

[0034] Optionally, the edge node total load mean L avg A threshold range is set to dynamically balance the light load priority and the performance priority strategy, and the specific implementation is as follows:

[0035] When the edge node total load mean L avg > 70%, the light load score influence coefficient θ1 is 0.7, and the performance influence coefficient θ2 is 0.3.

[0036] When the edge node total load mean L avg < 50%, the light load score influence coefficient θ1 is 0.3, and the performance influence coefficient θ2 is 0.7.

[0037] When 50% ≤ the edge node total load mean L avg ≤ 70%, the light load score influence coefficient θ1 is 0.5, and the performance influence coefficient θ2 is 0.5, to balance the light load degree and the performance.

[0038] Optionally, the visualization module is used for visualizing each module data, including a load heat map and a priority distribution pie chart.

[0039] Compared with the prior art, the present application has the following advantages:

[0040] Firstly, the environment evaluation module analyzes the environmental data of each server to determine the influence of the server environment on the performance, and quantitatively analyzes the result to obtain the environmental score of the edge node i.

[0041] Secondly, the importance of the overloaded node is evaluated based on the importance of the service, the user level and the load of the edge node i by the migration-out priority evaluation unit, and the task migration is performed according to the migration-out priority score of each edge node, so that the edge node with important data or serious load can be migrated first, the risk of interruption of the key service is reduced, the receiving capacity of the lightly loaded node is evaluated based on the CPU performance and the load of the edge node i by the migration-in priority evaluation unit, so that the servers with different specifications can be compared, the migration-in priority score of the edge node i is obtained, the influence of the actual performance of each edge node on the system resource allocation can be fully considered, the system resource allocation is more uniform, the overall operation efficiency is improved, and the advantage of edge collaborative computing is fully exerted.

[0042] Thirdly, the deviation degree of the load of a single edge node from the average load of all edge nodes is quantified, and the threshold of the deviation degree is set to judge the overload and light load, so that the judgment based on the load proportion of a single server is avoided, invalid scheduling is prevented, the quality of load optimization is improved, and the light load score influence coefficient and the performance influence coefficient are dynamically adjusted according to the actual situation of the average load of all edge nodes, so that the system can dynamically balance the strategy of light load priority and performance priority, the system can cope with the changing dynamic demand, the resource allocation is more uniform, and the overall operation efficiency is improved. BRIEF DESCRIPTION OF DRAWINGS

[0043] Figure 1 The system block diagram of the present application is shown in the figure;

[0044] Figure 2 The system flowchart of the present application is shown in the figure. DETAILED DESCRIPTION

[0045] The technical solutions in the embodiments of the present application will be described clearly and completely in combination with the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0046] Embodiment one, please refer to Figure 1 and Figure 2 The present embodiment provides a computer server computing resource allocation system based on edge collaboration, which comprises a data acquisition module, an environment evaluation module, a load evaluation module, an edge collaboration module and a visualization module.

[0047] The data acquisition module is used to acquire the environment data and the running data of the server, and the running data comprises CPU information, memory information and network usage information.

[0048] The environment assessment module analyzes the impact of the environment on performance based on the environmental data of each server, and obtains the environment score F of edge node i. env,i ;

[0049] The load assessment module is based on CPU information, memory information, network usage information, and the environmental score F of edge node i. env,i Fusion analysis yields the load L of edge node i. i ;

[0050] The edge collaboration module includes an overload judgment unit, a migration-out priority evaluation unit, and a migration-in priority evaluation unit;

[0051] The overload judgment unit is based on the load L of edge node i. i By analyzing the average load of all edge nodes, the load deviation value LD of each edge node i is obtained. i And based on the load deviation value LD of each edge node i i The analysis identified overloaded and lightly loaded nodes.

[0052] The migration priority assessment unit is based on the importance of the service, user level, and the load L of edge node i. i The importance of overloaded nodes is evaluated to obtain the migration priority score P of edge node i. i ;

[0053] The migration priority evaluation unit is based on CPU performance and the load L of edge node i. i The receiving capability of lightly loaded nodes is evaluated to obtain the migration priority score RP of edge node i. i Score P according to the migration priority of edge node i i The migration priority score RP of edge node i i Determine the order of resource allocation;

[0054] The visualization module is used to visualize the data from various modules, including load heatmaps and priority distribution pie charts. The load heatmap displays the load of each edge node in the form of a heatmap, while the priority distribution pie chart sorts the priorities of overloaded and lightly loaded nodes according to their percentage and displays them in a pie chart. This allows system administrators to quickly understand the load situation and the order of task migration, thereby improving the overall management efficiency of the system.

[0055] Specifically, the environmental assessment module first analyzes the environmental data of each server to determine the impact of the server's environment on its performance, and then performs quantitative analysis on the results to obtain the environmental score F of edge node i. env,i, and the load L of the edge node i is obtained through the load evaluation module and according to the actual running condition of the server, that is, CPU information, memory information, and network use information analysis i , and by introducing environmental factors, the analysis of the load can consider the dynamic changes of the environment, and the accuracy of the load analysis is improved

[0056] Then the overload judgment unit in the edge coordination module analyzes each edge node and the overall average, quantifies the difference between the single edge node and the overall load average, and obtains the load deviation LD of each edge node i i , in order to obtain the load deviation LD of each edge node i i The threshold is set to divide all nodes into overload nodes, light load nodes and normal nodes, and the normal nodes do not need to be adjusted. The overall resource shortage degree is judged based on the overall load average, which avoids the local optimization trap caused by isolated evaluation of single node state, and avoids frequent invalid scheduling, and improves the quality of load optimization

[0057] After dividing the overload nodes and the light load nodes, the tasks in the overload nodes need to be migrated to the light load nodes to realize load balancing. At this time, the migration priority evaluation unit is used to evaluate the importance of the business, the user level and the load L of the edge node i i The importance of the overload node is evaluated, and the result is quantified to obtain the migration priority score P of the edge node i i According to the migration priority score of each edge node, the tasks are migrated, so as to ensure that the edge nodes with important data or serious load are migrated first, and the risk of interruption of key services is reduced

[0058] Then, the migration priority evaluation unit is used to evaluate the CPU performance and the load L of the edge node i i The receiving capacity of the light load node is evaluated to obtain the migration priority score RP of the edge node i i The actual performance of each edge node can fully consider the influence of the system resource allocation, and not only according to the load ratio to analyze, so that different specifications of servers can be compared, avoid the performance of weak servers being over allocated tasks, make the system resource allocation more uniform, improve the overall operation efficiency, and fully play the advantages of edge collaborative computing.

[0059] In addition, the environmental evaluation module analyzes the influence of the environment on the performance of the server based on the environmental data to obtain the environmental influence coefficient. The analysis process of the environmental evaluation module is as follows:

[0060]

[0061] Where F env,iThe environment score for edge node i is given. The closer the score is to 1, the closer the environment is to the ideal state and the smaller the server performance degradation. The lower the score is to 1, the worse the environment is and the more severe the performance degradation. When the score is slightly greater than 1, the range is 1 to 1.3, which means that the environment is better than the ideal state. The number of edge nodes i depends on the number of servers.

[0062] T act,i The actual temperature of edge node i;

[0063] T sta Standard temperature;

[0064] For temperature rating, T act,i -T sta For temperature difference;

[0065] W1 is the temperature influence coefficient;

[0066] Because server hardware is sensitive to temperature, if the actual temperature T of edge node i... act,i Exceeding the standard temperature T sta This can lead to performance degradation, such as the CPU triggering frequency reduction protection due to overheating, and an increase in chip leakage current. Therefore, when T... act,i -T sta Temperature score when temperature difference > 0 By setting a minus sign, When it becomes a negative contribution, the environment score F of edge node i is reduced. env,i It will change with temperature difference T act,i -T sta The value decreases as the temperature difference T increases, reflecting how rising temperatures cause a decrease in server performance. act,i -T sta Temperature score when <0 make When it becomes a positive contributor, the environment score F of edge node i is... env,i Will follow T act,i -T sta The decrease in temperature difference is increased to reflect the improved performance potential of the server under low temperature conditions. However, in practical applications, lower temperature is not always better. The minimum and maximum temperature thresholds can be set according to the actual situation of the server.

[0067] H act,i The actual humidity of edge node i;

[0068] H sta Standard humidity;

[0069] For humidity rating, H sta -H act,i Humidity difference;

[0070] W2 is a humidity influence coefficient;

[0071] Unlike temperature, both too high or too low humidity will affect the server, such as too high humidity will increase the risk of short circuit of the equipment, and the heat dissipation efficiency will decrease, and too low humidity will easily generate static electricity and damage the chip, therefore, the absolute value is set to reflect the humidity difference H act,i H sta on the server performance;

[0072] E act,i is the actual energy consumption of the edge node i;

[0073] E sta is the standard energy consumption;

[0074] E is the energy consumption score, E act,i E sta is the energy consumption difference;

[0075] W3 is an energy consumption influence coefficient;

[0076] The temperature influence coefficient W1 + the humidity influence coefficient W2 + the energy consumption influence coefficient W3 = 1, and the temperature influence coefficient W1 > the energy consumption influence coefficient W3 > the humidity influence coefficient W2;

[0077] If E act,i E sta The energy consumption difference < 0 indicates that the system is in a high-efficiency low-load state, and the server performance attenuation is reduced, at this time The environment score F of the edge node i is increased by using the plus sign env,i If E act,i E sta The energy consumption difference > 0 indicates that the server performance attenuation is more, at this time The environment score F of the edge node i is reduced by adding a value less than 0 env,i , which indicates the influence of the actual energy consumption on the environment score F of the edge node i. env,i

[0078] F is the environment score of the edge node i env,i The environment threshold Y1 and the environment threshold Y2 are set, the environment threshold Y1 takes the value of 0.6, and the environment threshold Y2 takes the value of 1.3, when the environment score F of the edge node i env,i < the environment threshold Y1, or the environment score F of the edge node i env,i > the environment threshold Y2, it indicates that the environmental conditions are deteriorating, and the server performance is seriously damaged, and there is a risk of larger system failure or long-term instability, at this time the computer server is checked.

[0079] In addition, the load evaluation module evaluation process is as follows:​

[0080]

[0081] wherein L i is the load of the edge node i, the greater the result, the higher the load, and vice versa, the smaller the load, the value range is 0 to 1, even if the result is greater than 1, it is taken as 1;

[0082] C i is the normalized CPU usage rate of the edge node i;

[0083] β1is the CPU usage rate influence coefficient;

[0084] A i is the normalized memory usage rate of the edge node i;

[0085] β2is the memory usage rate influence coefficient;

[0086] D i is the normalized bandwidth usage rate of the edge node i, by normalizing the data, the interference of hardware performance difference on load evaluation can be eliminated, the original load of low-configuration edge nodes such as four-core CPU may be higher than that of high-configuration edge nodes such as eight-core CPU for a long time, so that server hardware of different specifications such as CPU can be compared, and the accuracy of load evaluation is improved;

[0087] β3is the bandwidth usage rate influence coefficient;

[0088] CPU usage rate influence coefficient β1+ memory usage rate influence coefficient β2+ bandwidth usage rate influence coefficient β3=1;

[0089] represents the environment correction factor, when the environment score F env,i of the edge node i is less than 1, the load L act,i of the edge node i is higher, the environment score F env,i of the edge node i is closer to 1, and the load L act,i of the edge node i is lower.

[0090] By weightedly fusing the normalized CPU usage rate, memory usage rate and bandwidth usage rate, different specifications of servers can be compared, and several core data of the server are weightedly fused, which can reduce misjudgment such as CPU idle but memory bottleneck, and by introducing the environment correction factor to comprehensively evaluate the server load condition, the degree of performance decay of the environment can be dynamically evaluated, and the accuracy of the load L act,i of the edge node i is improved.

[0091] In addition, the overload judgment unit judges as follows:

[0092]

[0093] LD i is the load deviation value of the edge node i;

[0094] L avg is the total load average of the edge nodes;

[0095] N is the number of edge nodes;

[0096] The overload threshold is set as Y3, and the light load threshold is set as Y4. When the load deviation value LD i of the edge node i is greater than the overload threshold Y3, it is determined that the edge node i is overloaded. When the load deviation value LD i of the edge node i is less than the light load threshold Y4, it is determined that the edge node i is lightly loaded. Subsequently, the tasks in the overloaded edge node are transferred to the lightly loaded edge node. By quantifying the deviation degree of the load of a single edge node from the total load average of the edge nodes and setting a threshold for the deviation degree, the overload and light load are determined, which avoids frequent invalid scheduling and improves the system operation efficiency.

[0097] Further, the migration priority evaluation unit evaluates all edge nodes determined to be overloaded, and the evaluation process is as follows:

[0098]

[0099] CP i is the migration priority score of the edge node i. The higher the evaluation, the stronger the importance of the edge node and the more serious the load situation;

[0100] G i is the importance score of the edge node i, which is obtained according to the system preset, and the value range is 0 to 1. The closer to 1, the more important the edge node is. For example, if the edge node i is used for financial transaction business, G i is 1

[0101] γ1 is the importance score influence coefficient;

[0102] LV i is the user level of the edge node i, which is obtained according to the information of the user in the system. The user level can be determined according to the system use time or according to the user identity. For example, the system administrator can be defaulted as the highest user level to ensure the management efficiency of the administrator to the system;

[0103] LV max is the highest user level;

[0104] is the user level score of the edge node i;

[0105] γ2 is a user level influence coefficient;

[0106] L i is the load of edge node i;

[0107] L i,max is the maximum load of edge node i, taking the value of 1;

[0108] is the load score of edge node i, when the load L i of edge node i is higher, the load score P of edge node i is higher, thereby increasing the migration-out priority score P i of edge node i;

[0109] γ3 is a load influence coefficient;

[0110] The importance score influence coefficient γ1+ the user level influence coefficient γ2+ the load influence coefficient γ3=1;

[0111] Specifically, the migration-out priority score CP i of each edge node i is calculated, and the edge nodes are sorted according to the scores, with a higher score indicating a higher priority of the edge node i. In the subsequent migration-out of edge node tasks, the edge nodes are migrated in order of the migration-out priority score, ensuring that edge nodes with important data and high loads are given priority, reducing the risk of interruption of critical services and ensuring system stability.

[0112] Meanwhile, the migration-in priority evaluation unit evaluates all edge nodes determined to be lightly loaded, and the evaluation process is as follows:

[0113]

[0114] RP i is the migration-in priority score of edge node i, with a higher score indicating lower load and stronger performance of the node;

[0115] is the light load score of edge node i;

[0116] θ1 is a light load score influence coefficient;

[0117] CH i is the CPU core number of edge node i, indicating the processing capacity of edge node i;

[0118] CH avg is the average CPU core number of all edge nodes;

[0119] represents the performance deviation score of the edge node i, the higher the score, the stronger the performance of the edge node i among all edge nodes, and vice versa, the migration priority score RP of the edge node i i will increase with the increase of , so as to fully consider the influence of the actual performance of the edge node on the system resource allocation;

[0120] θ2 is the performance influence coefficient;

[0121] The light load score influence coefficient θ1 and the performance influence coefficient θ2 are both 0.5 in the initial state.

[0122] Specifically, by introducing the number of CPU cores, it can avoid that the low but weak performance node is over-allocated tasks, or the high performance but close to overload edge node is blindly allocated tasks, balance the resource utilization efficiency and task execution efficiency, and ensure the accuracy of load adjustment.

[0123] Meanwhile, the light load score influence coefficient θ1 and the performance influence coefficient θ2 can be dynamically adjusted according to the overall overload of the system, as follows:

[0124] When the average value of the total load of the edge node L avg > 70%, it indicates that the overall system resource is tight, and the light load score influence coefficient θ1 can be increased to preferentially guarantee that the light load node is not overloaded, at this time the light load score influence coefficient θ1 is 0.7 and the performance influence coefficient θ2 is 0.3.

[0125] When the average value of the total load of the edge node L avg < 50%, it indicates that the overall system resource is sufficient, and tasks are preferentially allocated to high-performance nodes to reduce delay, at this time the light load score influence coefficient θ1 is 0.3 and the performance influence coefficient θ2 is 0.7.

[0126] When 50% ≤ the average value of the total load of the edge node L avg ≤ 70%, it indicates that the system load is balanced, at this time the light load score influence coefficient θ1 is 0.5 and the performance influence coefficient θ2 is 0.5, to balance the light load degree and performance.

[0127] By dynamically adjusting the light load score influence coefficient θ1 and the performance influence coefficient θ2 according to the actual situation of the average value of the total load of the edge node L avg , the system can dynamically balance the light load priority and performance priority strategy, make the system resource allocation more uniform, and improve the system running efficiency.

[0128] Through the cooperation of the migration-out priority evaluation unit and the migration-in priority evaluation unit, the load difference and the allocation weight can be dynamically quantified, manual intervention is reduced, and in abnormal scenarios such as node failure and task burst growth, the system can automatically adjust the task allocation of each edge node, maintain stable operation of the system, reduce the risk of service interruption, and fully exert the advantages of edge collaborative computing.

[0129] While embodiments of the application have been shown and described, it is to be understood that the embodiments described are merely exemplary of the principles and application of the present application. Numerous modifications, changes, variations, substitutions, and equivalents will occur to those of ordinary skill in the art without departing from the spirit and scope of the application as defined by the following claims and their equivalents.

Claims

1. A computer server computing resource allocation system based on edge collaboration, characterized in that: It includes a data acquisition module, an environmental assessment module, a load assessment module, an edge collaboration module, and a visualization module; The data acquisition module is used to collect environmental data and operational data of the server, including extracted CPU information, memory information and network usage information. The environmental assessment module analyzes the impact of the environment on performance based on the environmental data of each server, and obtains the environmental score F of edge node i. env,i ; The load assessment module is based on CPU information, memory information, network usage information, and the environmental score F of edge node i. env,i Fusion analysis yields the load L of edge node i. i ; The edge collaboration module includes an overload judgment unit, a migration-out priority evaluation unit, and a migration-in priority evaluation unit. The overload judgment unit is based on the load L of edge node i. i By analyzing the average load of all edge nodes, the load deviation value LD of each edge node i is obtained. i And based on the load deviation value LD of each edge node i i The analysis identified overloaded and lightly loaded nodes. The migration priority assessment unit is based on the importance of the service, user level, and the load L of edge node i. i The importance of overloaded nodes is evaluated to obtain the migration priority score P of edge node i. i ; The migration priority evaluation unit is based on CPU performance and the load L of edge node i. i The receiving capability of lightly loaded nodes is evaluated to obtain the migration priority score RP of edge node i. i .

2. The computer server computing resource allocation system based on edge collaboration according to claim 1, characterized in that: The environmental data includes temperature, humidity, and energy consumption information. The environmental assessment module performs a weighted fusion analysis based on the temperature, humidity, and energy consumption information of each server to obtain the environmental score F of edge node i. env,i This is used to assess the impact of the environment on server performance; The environmental score F for edge node i env,i Set environmental thresholds Y1 and Y2, and when the environmental score F of edge node i... env,i <Environmental threshold Y1, or environmental score F of edge node i env,i >When the environmental threshold Y2 is reached, the server is checked.

3. The computer server computing resource allocation system based on edge collaboration according to claim 2, characterized in that: The load assessment module first normalizes the CPU, memory, and network usage information of each server to obtain the normalized CPU utilization C of edge node i. i Memory usage A after normalization of edge node i i The bandwidth utilization of edge node i is normalized to allow for comparison of servers of different specifications; Then, the CPU utilization C after normalizing the edge node i is... i Memory usage A after normalization of edge node i i The normalized bandwidth utilization of edge node i is weighted and fused together, while the environmental score F of edge node i is also introduced. env,i The analysis ultimately yielded the load L of edge node i. act,i .

4. The edge-collaboration-based computer server computing resource allocation system according to claim 3, characterized in that: The overload judgment unit first calculates the average load of all edge nodes to obtain the average total load L of the edge nodes. avg ; Then, based on the load L of edge node i i and the average total load L of edge nodes avg We analyze the deviation of edge node i and quantify the deviation of edge node i to obtain the load deviation value LD of edge node i. i ; Set the overload threshold to Y3 and the light load threshold to Y4. When the load of edge node i deviates from the value LD... i When the overload threshold Y3 is reached, the edge node i is determined to be overloaded. When the load of edge node i deviates from the value LD... i When the light load threshold Y4 is less than 4, the edge node i is determined to be lightly loaded.

5. The edge-collaboration-based computer server computing resource allocation system according to claim 4, characterized in that: The migration priority evaluation unit analyzes the edge nodes that are determined to be overloaded and determines the task migration order of the overloaded edge nodes. First, the importance of each edge node is analyzed based on its different business applications, resulting in an importance score G for edge node i. i ; Then, compare the user's level with the highest level to obtain the user level score; Finally, based on the importance score G of edge node i... i User rating and load L of edge node i i The fusion analysis yields the migration priority score P for edge node i. i The order in which tasks migrate out of overloaded edge nodes is determined by their migration priority scores.

6. The computer server computing resource allocation system based on edge collaboration according to claim 5, characterized in that: The migration priority evaluation unit analyzes the edge nodes that are determined to be lightly loaded and determines the migration order of tasks for the lightly loaded edge nodes. The evaluation process is as follows: Where RP i Score the migration priority of edge node i; Give the light load score to edge node i; θ1 is the influence coefficient of the light load score; CH i Let i be the number of CPU cores in edge node i; CH avg The average number of CPU cores across all edge nodes; This indicates the performance deviation score of edge node i; θ2 is the performance impact coefficient; The light load rating influence coefficient θ1 + performance influence coefficient θ2 = 1. Under the initial condition, both the light load rating influence coefficient θ1 and the performance influence coefficient θ2 are 0.

5.

7. The computer server computing resource allocation system based on edge collaboration according to claim 6, characterized in that: The average total load L of the edge nodes avg A threshold range is set to dynamically balance the strategies of prioritizing light loads and prioritizing performance, as follows: When the average total load of edge nodes is L avg When the load is greater than 70%, the light load rating influence coefficient θ1 is 0.7, and the performance influence coefficient θ2 is 0.

3. When the average total load of edge nodes is L avg When the load is less than 50%, the light load rating influence coefficient θ1 is 0.3 and the performance influence coefficient θ2 is 0.

7. When 50% ≤ the average total load L of edge nodes avg For loads ≤70%, the light load rating influence coefficient θ1 is 0.5, and the performance influence coefficient θ2 is 0.5, in order to balance the light load level and performance.

8. The computer server computing resource allocation system based on edge collaboration according to claim 1, characterized in that: The visualization module is used to visualize the data from each module, including load heatmaps and priority distribution pie charts.

Images