Dynamic allocation management system for monitoring computing power resources in real time

Abstract

The invention discloses a dynamic allocation management system for monitoring computing power resources in real time, relates to the technical field of computing power resource management, and is used for solving the problem that an existing dynamic allocation management system for the computing power resources cannot efficiently match, optimally utilize, accurately evaluate and dynamically allocate heterogeneous computing power resources. Comprising a central cloud and a plurality of edge nodes in communication connection with the central cloud, and a real-time monitoring prediction unit, a dynamic distribution management unit and a fault detection processing unit are arranged in the central cloud; according to the method, the utilization rate and delay information of the heterogeneous computing power resources are analyzed in detail, so that the resources most suitable for the target processing task can be efficiently matched; by eliminating the computing power resource with a relatively low delay value, the utilization efficiency of the whole resource is improved, and resource waste is avoided; and resource demand prediction is performed in combination with a plurality of parameters, so that the performance and value of the computing power resources to be allocated can be evaluated more accurately. And effective management and optimal utilization of heterogeneous computing power resources are realized.

Description

Technical Field

[0001] The present invention relates to the technical field of computing resource management, and in particular to a dynamic allocation management system for real-time monitoring of computing resources. Background Art

[0002] With the rapid development of domestic cloud computing, big data and other industries, the demand for computing resources is also growing. In order to meet this demand, some large Internet companies, scientific research institutions and cloud computing service providers have begun to invest resources in system research and development. At present, some mature systems have been applied in actual scenarios in China. It can monitor the real-time status of computing resources, such as the utilization rate and load of heterogeneous computing resources such as CPU, GPU, FPGA, NPU, etc., and dynamically adjust the allocation of resources based on this information.

[0003] However, the existing dynamic allocation and management systems for computing resources are unable to efficiently match, optimize utilization, accurately evaluate, and dynamically allocate heterogeneous computing resources, resulting in the inability to better utilize computing resources. Summary of the invention

[0004] The purpose of the present invention is to solve the problem that the existing dynamic allocation management system of computing power resources is unable to efficiently match, optimize utilization, accurately evaluate and dynamically allocate heterogeneous computing power resources, and to propose a dynamic allocation management system for real-time monitoring of computing power resources.

[0005] The object of the present invention can be achieved by the following technical scheme: comprising a central cloud and a plurality of edge nodes connected to the central cloud in communication, wherein the central cloud is provided with a real-time monitoring prediction unit and a dynamic allocation management unit, wherein the real-time monitoring prediction unit is used to collect and analyze the status information of the heterogeneous computing resources corresponding to the edge nodes in real time by using monitoring tools and algorithms, and the specific analysis process is as follows:

[0006] Obtain the target processing task, match the required heterogeneous computing resources corresponding to the target processing task, determine the utilization rate of the heterogeneous computing resources, and when the utilization rate is less than the set utilization rate threshold, mark the heterogeneous computing resources as computing resources to be allocated and number them i; obtain the status information of the computing resources to be allocated, parse all delay durations within the preset time period and analyze the delay value of the computing resources to be allocated, obtain the delay threshold corresponding to the computing resources to be allocated, compare the delay value with the corresponding delay threshold, and when the delay value is less than or equal to the delay threshold, eliminate the computing resources to be allocated; when the delay value is greater than the delay threshold, remove the model of the computing resources to be allocated. Match with several preset models, wherein each of the preset models corresponds to a model constant. If the model of the computing power resources to be allocated matches the corresponding preset model, the model constant that matches the preset model is marked as the model value XHi of the computing power resources to be allocated; the value of the number of floating-point operations of the computing power resources to be allocated is divided by the value of the power to obtain the energy efficiency ratio NXi; the values ​​of the utilization rate, the number of load tasks and the energy consumption ratio corresponding to the computing power resources to be allocated are marked as LYi, FZi and NHi respectively; the resource demand prediction model is substituted to output the resource allocation prediction value of the computing power resources to be allocated; the resource allocation prediction value of the computing power resources to be allocated is sent to the dynamic allocation management unit;

[0007] The dynamic allocation management unit is used to dynamically allocate and manage heterogeneous computing resources.

[0008] As a preferred embodiment of the present invention, heterogeneous computing resources include CPU, GPU, FPGA, NPU and ASIC; status information includes performance data, utilization, number of load tasks and energy consumption data; performance data includes number of floating-point operations, delay time, model and temperature; energy consumption data includes power and energy consumption ratio.

[0009] As a preferred implementation of the present invention, the specific process of obtaining the extended value of the computing power resources to be allocated is:

[0010] The delay durations are sorted according to the order of collection, and two adjacent delay durations are compared. If the delay duration in the previous order is greater than the delay duration in the next order, the delay duration in the previous order is subtracted from the delay duration in the next order to obtain the time drop difference value, and the values ​​of all the time drop difference values ​​are summed up to obtain the total time drop value; the average value of all delay duration values ​​is calculated to obtain the delay average value, and the maximum and minimum values ​​of all delay durations are obtained and marked as the maximum delay duration and the minimum delay duration respectively. The total time drop value, the delay average value, the maximum delay duration and the minimum delay duration are normalized and the values ​​of the four are taken, marked as TZ, TP, TMAX and TMIN respectively, and substituted into the preset delay analysis model:

[0011] Output the delay value TYi of the computing power resources to be allocated, where f1, f2, f3 and f4 are the weight factors corresponding to the total time drop value, the average delay value, the maximum delay duration and the minimum delay duration respectively.

[0012] As a preferred implementation of the present invention, the resource demand prediction model is specifically: Output the estimated value ZFi of the computing power resources to be allocated; where hy1, hy2 and hy3 are model thresholds; is the extension weight factor, is the energy efficiency ratio weight factor; is the utilization weight factor; is the energy consumption ratio weight factor; is the weight factor of the load task quantity, i=1, 2, ..., n, where n is the maximum value of the number.

[0013] As a preferred implementation of the present invention, the specific process of performing dynamic allocation management is as follows:

[0014] The heterogeneous computing resources are sorted from large to small according to the resource allocation estimate. If the number of target processing tasks is less than or equal to the number of heterogeneous computing resources, a preset number of heterogeneous computing resources are selected in sequence from front to back according to the sorting, and then the target processing tasks are equally divided and sent to the edge nodes corresponding to the heterogeneous computing resources in turn, and processed by the corresponding heterogeneous computing resources inside the edge nodes; if the number of target processing tasks is greater than the number of heterogeneous computing resources, the target processing tasks are sent to the edge node corresponding to the heterogeneous computing resource with the highest sorting until its utilization rate is equal to the maximum utilization rate threshold, and then the sending is stopped; then the target processing tasks are sent to the edge node corresponding to the second-ranked heterogeneous computing resource; until all target processing tasks are distributed.

[0015] As a preferred embodiment of the present invention, a fault detection processing unit is further provided in the central cloud; the fault detection processing unit is used to perform fault detection and analysis on heterogeneous computing resources, and the specific process is as follows:

[0016] Obtain all temperatures corresponding to heterogeneous computing resources, compare the temperature with the corresponding preset maximum temperature, count the number of temperatures equal to or higher than the maximum temperature to obtain the number of heterogeneous temperatures; calculate the average value of all temperatures to obtain the average temperature. When the average temperature is higher than the set average threshold, obtain the voltage Uj and current Ij of the heterogeneous computing resources, j = 1, 2, ..., m, m represents the acquisition time number, m represents the maximum value of the number, substitute into the formula Output voltage fluctuation value SU and current fluctuation value SI; normalize the number of different temperatures, voltage fluctuation value, average temperature value and current fluctuation value and take the values ​​of the four and substitute them into the formula Obtain the fault detection value GYi of the heterogeneous computing power resources; where β1, β2, β3 and β4 are the weight factors corresponding to the number of different temperatures YW, the average temperature value PJ, the voltage fluctuation value SU and the current fluctuation value SI respectively;

[0017] Set interval one and interval two; when the fault detection value belongs to interval one, shut down the edge node corresponding to the heterogeneous computing power resource, and then convert the fault detection value into shutdown duration according to a certain ratio, and then restart after executing the shutdown duration, and at the same time, feed back the load tasks not executed by the edge node to the dynamic allocation management unit for reallocation; when the fault detection value belongs to interval two, the edge node corresponding to the heterogeneous computing power resource executes the self-check and repair program, and at the same time, feeds back the load tasks not executed by the edge node to the dynamic allocation management unit for reallocation; after the self-check and repair program is completed, recalculate the fault detection value of the heterogeneous computing power resource. If the fault detection value still belongs to interval two, mark the edge node corresponding to the heterogeneous computing power resource as a faulty node, and send the coordinates of the faulty node, the name of the heterogeneous computing power resource and the fault detection value to the intelligent terminal of the corresponding maintenance personnel, and the maintenance personnel arrive at the location of the faulty node to repair or replace the heterogeneous computing power resource.

[0018] Compared with the prior art, the present invention has the following beneficial effects:

[0019] 1. The present invention can efficiently match the most suitable resources for the target processing tasks by analyzing the utilization and delay information of heterogeneous computing resources in detail; by eliminating computing resources with low delay values, the overall resource utilization efficiency is improved and resource waste is avoided; by combining multiple parameters (such as model, energy efficiency ratio, utilization rate, number of load tasks and energy consumption ratio) to predict resource demand, the performance and value of the computing resources to be allocated can be more accurately evaluated. Effective management and optimal utilization of heterogeneous computing resources are achieved.

[0020] 2. By real-time monitoring of key parameters such as temperature, voltage, and current of heterogeneous computing resources, potential faults can be discovered and handled in a timely manner, improving the stability and reliability of the system; different processing measures are automatically taken according to different ranges of fault detection values, realizing intelligent and automated fault handling, reducing service interruptions and data loss caused by faults, and improving system availability. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] In order to facilitate understanding by those skilled in the art, the present invention is further described below with reference to the accompanying drawings.

[0022] Figure 1 It is the overall principle block diagram of the present invention;

[0023] Figure 2 This is a schematic diagram of the internal principles of the central cloud of the present invention. DETAILED DESCRIPTION

[0024] The technical solution of the present invention will be clearly and completely described below in conjunction with the embodiments. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0025] Example 1

[0026] See also Figure 1 As shown, a dynamic allocation management system for real-time monitoring of computing resources includes a central cloud and several edge nodes (edge ​​nodes are road inspection vehicles or other intelligent terminals with edge computing hosts) connected to the central cloud in communication. The central cloud is provided with a real-time monitoring prediction unit, a dynamic allocation management unit and a fault detection processing unit.

[0027] The real-time monitoring prediction unit uses monitoring tools and algorithms to collect and analyze the status information of the corresponding heterogeneous computing resources of the edge nodes in real time. The heterogeneous computing resources include CPU, GPU, FPGA, NPU, ASIC, etc. GPU (Graphics Processing Unit) is a microprocessor specially used for processing images and graphics. It is similar to the central processing unit (CPU) of a computer, but is specially designed to process image and graphics data. GPU usually contains thousands of streaming processors, which can process large amounts of data in parallel to achieve efficient graphics rendering and calculation. NVIDIA is the representative of GPU. NPU: NPU (Neural Processing Unit) is a chip specially used for deep learning calculation. It is one of the hot technologies in the field of artificial intelligence in recent years and is widely used in various artificial intelligence applications, such as autonomous driving, face recognition, intelligent voice and video surveillance. NPU accelerates complex calculations in deep neural networks through large-scale parallel processing units and efficient interconnection structures, and has the advantages of high efficiency, low power consumption, low latency, stability, and programmability. FPGA: A flexible and programmable hardware platform with high computing performance and customizability, capable of providing acceleration and optimization of AI algorithms; in AI applications, it can be used to implement neural network accelerators, high-performance computing units, etc., providing high-performance and low-latency computing capabilities for compute-intensive AI tasks. Intel Stratix 10NX FPGA is specially designed for AI, with AI tensor blocks, including dense low-precision multiplier arrays, adjusted for matrix and vector multiplication, and can perform INT4, INT8, Block FP12 or Block FP16 operations. In addition, these tensor blocks can be cascaded together to support large matrices. DPU: DPU (Data Processing Unit) is a dedicated processor constructed with data as the center. Due to the slowdown in CPU performance growth, DPU was created in order to seek more efficient computing chips. It uses a specific software-defined technology route to support the virtualization of infrastructure layer resources and supports functions such as storage, security, and quality of service management. The emergence of DPU is designed to efficiently handle large-scale data workloads in data centers, including tasks such as data transmission, protocol, protection, compression, analysis, and encryption.

[0028] The status information includes performance data, utilization, number of load tasks, energy consumption data, etc.; among them, the performance data includes the number of floating-point operations, delay time, model and temperature; the energy consumption data includes power and energy consumption ratio.

[0029] The specific analysis process of the real-time monitoring prediction unit is as follows:

[0030] Obtain the target processing task, match the required heterogeneous computing resources corresponding to the target processing task, determine the utilization rate of the heterogeneous computing resources, and when the utilization rate is less than the set utilization rate threshold, mark the heterogeneous computing resources as computing resources to be allocated and number them i, i = 1, 2, ..., n, where n is the maximum value of the number; obtain the status information of the computing resources to be allocated, parse all the delay durations within the preset time period, sort the delay durations according to the order of collection, compare the two adjacent delay durations, and if the delay duration in the previous order is greater than the delay duration in the later order, Delay duration, the delay duration in the previous order is subtracted from the delay duration in the next order to obtain the time drop difference value, and the values ​​of all the time drop difference values ​​are summed up to obtain the total time drop value; the average value of all delay duration values ​​is calculated to obtain the delay average value, and the maximum and minimum values ​​of all delay durations are obtained and marked as the maximum delay duration and the minimum delay duration respectively, and the total time drop value, the delay average value, the maximum delay duration and the minimum delay duration are normalized and the values ​​of the four are taken, marked as TZ, TP, TMAX and TMIN respectively, and substituted into the preset delay analysis model:

[0031] Output the delay value TYi of the computing power resources to be allocated, where f1, f2, f3 and f4 are the weight factors corresponding to the total time drop value, the average delay value, the maximum delay duration and the minimum delay duration, respectively, and their sizes are custom settings, such as 1.31, 1.51, 1.67 and 1.75 respectively; obtain the delay threshold corresponding to the computing power resources to be allocated, compare the delay value with the corresponding delay threshold, when the delay value is less than or equal to the delay threshold, the computing power resources to be allocated are eliminated, and when the delay value is greater than the delay threshold, the computing power resources to be allocated are eliminated. The model is matched with several preset models, wherein each of the preset models corresponds to a model constant. If the model of the computing power resource to be allocated matches the corresponding preset model, the model constant that matches the preset model is marked as the model value XHi of the computing power resource to be allocated; the value of the number of floating-point operations of the computing power resource to be allocated is divided by the value of the power to obtain the energy efficiency ratio NXi; the values ​​of the utilization rate, the number of load tasks and the energy consumption ratio corresponding to the computing power resource to be allocated are marked as LYi, FZi and NHi respectively; and the resource demand prediction model is substituted: Output the estimated value ZFi of the computing power resources to be allocated; where hy1, hy2 and hy3 are model thresholds, and hy1<hy2<hy3; is the extension weight factor, and is the energy efficiency ratio weight factor, and is the utilization weight factor, and is the energy consumption ratio weight factor, and is the weight factor of the load task quantity; the estimated value of the computing power resources to be allocated is sent to the dynamic allocation management unit.

[0032] According to the task requirements, select suitable heterogeneous computing resources, check the utilization of these resources, mark the resources below the set threshold as computing resources to be allocated, and number them, collect the delay duration of the computing resources to be allocated, calculate the total time drop value, average delay value, maximum delay duration and minimum delay duration, and perform normalization. Substitute the normalized value into the preset delay analysis model, calculate the delay value of each computing resource to be allocated, compare the delay value with the delay threshold, and eliminate resources that do not meet the requirements; match the preset model for qualified resources, and calculate the energy efficiency The system collects other relevant parameters, substitutes the collected parameters into the resource demand prediction model, and outputs the estimated value of the computing power resources to be allocated; by analyzing the utilization rate and delay information of heterogeneous computing power resources in detail, it can efficiently match the resources that are most suitable for the target processing tasks; by eliminating computing power resources with low delay values, it improves the overall resource utilization efficiency and avoids resource waste; combining multiple parameters (such as model, energy efficiency ratio, utilization rate, number of load tasks and energy consumption ratio) for resource demand prediction can more accurately evaluate the performance and value of computing power resources to be allocated. It realizes the effective management and optimal utilization of heterogeneous computing power resources.

[0033] The dynamic allocation management unit dynamically allocates and manages heterogeneous computing resources, specifically: sorting the heterogeneous computing resources from large to small according to the resource allocation estimate; if the number of target processing tasks is less than or equal to the number of heterogeneous computing resources, then selecting a preset number of heterogeneous computing resources in sequence from front to back according to the sorting; then sending the target processing tasks in equal parts to the edge nodes corresponding to the heterogeneous computing resources in sequence, and processing them by the corresponding heterogeneous computing resources inside the edge nodes; if the number of target processing tasks is greater than the number of heterogeneous computing resources, then sending the target processing tasks to the edge node corresponding to the heterogeneous computing resource with the highest sorting until its utilization rate is equal to the maximum utilization rate threshold, and then stopping sending; then sending to the edge node corresponding to the second-ranked heterogeneous computing resource; until all target processing tasks are distributed.

[0034] The fault detection processing unit performs fault detection and analysis on heterogeneous computing resources. The specific process is as follows:

[0035] Obtain all temperatures corresponding to heterogeneous computing resources, compare the temperature with the corresponding preset maximum temperature, count the number of temperatures equal to or higher than the maximum temperature to obtain the number of heterogeneous temperatures; calculate the average value of all temperatures to obtain the average temperature. When the average temperature is higher than the set average threshold, obtain the voltage Uj and current Ij of the heterogeneous computing resources, j = 1, 2, ..., m, m represents the acquisition time number, m represents the maximum value of the number, substitute into the formula Output voltage fluctuation value SU and current fluctuation value SI; normalize the number of different temperatures, voltage fluctuation value, average temperature value and current fluctuation value and take the values ​​of the four and substitute them into the formula Obtain the fault detection value GYi of the heterogeneous computing power resources; where β1, β2, β3 and β4 are the weight factors corresponding to the number of different temperatures YW, the average temperature value PJ, the voltage fluctuation value SU and the current fluctuation value SI respectively;

[0036] Set interval one and interval two; when the fault detection value belongs to interval one, shut down the edge node corresponding to the heterogeneous computing power resource, and then convert the fault detection value into shutdown duration according to a certain ratio, and then restart after executing the shutdown duration, and at the same time, feed back the load tasks not executed by the edge node to the dynamic allocation management unit for reallocation; when the fault detection value belongs to interval two, the edge node corresponding to the heterogeneous computing power resource executes the self-check and repair program, and at the same time, feeds back the load tasks not executed by the edge node to the dynamic allocation management unit for reallocation; after the self-check and repair program is completed, recalculate the fault detection value of the heterogeneous computing power resource. If the fault detection value still belongs to interval two, mark the edge node corresponding to the heterogeneous computing power resource as a faulty node, and send the coordinates of the faulty node, the name of the heterogeneous computing power resource and the fault detection value to the intelligent terminal of the corresponding maintenance personnel, and the maintenance personnel arrive at the location of the faulty node to repair or replace the heterogeneous computing power resource.

[0037] By real-time monitoring of key parameters such as temperature, voltage, and current of heterogeneous computing resources, potential faults can be discovered and handled in a timely manner, improving the stability and reliability of the system; different processing measures are automatically taken according to different ranges of fault detection values, realizing intelligent and automated fault handling, reducing service interruptions and data loss caused by faults, and improving system availability.

[0038] Embodiment 2:

[0039] Based on the first embodiment, the dynamic allocation management unit further sets a task queuing program, which queues the target processing tasks in sequence according to the sorting rules, so as to ensure that the tasks are executed in sequence and reduce the waiting time. The sorting rule is: select the target processing task within the preset time period, which can be one minute, five minutes, ten minutes or one hour or other time periods; identify the target processing task with an expedited mark, and if the target processing task is identified as having an expedited mark, the target processing task is ranked first. If multiple target processing tasks with expedited marks are identified, they are ranked according to the order in which the target processing tasks are received; then sort other target processing tasks without expedited marks, and match the keywords of the target processing tasks with several preset keywords. Each preset keyword has a priority score. If the keyword of the target processing task is consistent with the preset keyword, the priority score of the corresponding keyword is matched. All priority scores matched to the target processing task are summed to obtain the total priority score, and sorted according to the total priority score. If the total priority scores are consistent, they are sorted according to the chronological order. After the sorting is completed, the target processing tasks that are not matched with the priority score are sorted according to the chronological order. After the sorting is completed; the sorting rule is executed for the next preset time period.

[0040] The preferred embodiments of the present invention disclosed above are only used to help explain the present invention. The preferred embodiments do not describe all the details in detail, nor do they limit the invention to only specific implementation methods. Obviously, many modifications and changes can be made according to the content of this specification. This specification selects and specifically describes these embodiments in order to better explain the principles and practical applications of the present invention, so that those skilled in the art can understand and use the present invention well. The present invention is limited only by the claims and their full scope and equivalents.

Claims

1. A dynamic allocation management system for real-time monitoring of computing resources, comprising a central cloud and a plurality of edge nodes connected to the central cloud in communication, wherein the central cloud is provided with a real-time monitoring prediction unit and a dynamic allocation management unit, characterized in that: The real-time monitoring prediction unit is used to collect and analyze the status information of the heterogeneous computing resources corresponding to the edge nodes in real time using monitoring tools and algorithms. The specific analysis process is as follows: Obtain the target processing task, match the required heterogeneous computing resources corresponding to the target processing task, determine the utilization rate of the heterogeneous computing resources, and when the utilization rate is less than the set utilization rate threshold, mark the heterogeneous computing resources as computing resources to be allocated and number them i; obtain the status information of the computing resources to be allocated, parse all delay durations within the preset time period and analyze the delay value of the computing resources to be allocated, obtain the delay threshold corresponding to the computing resources to be allocated, compare the delay value with the corresponding delay threshold, and when the delay value is less than or equal to the delay threshold, eliminate the computing resources to be allocated; when the delay value is greater than the delay threshold, Then the model of the computing power resources to be allocated is matched with several preset models, wherein each of the preset models corresponds to a model constant. If the model of the computing power resources to be allocated matches the corresponding preset model, the model constant that matches the preset model is marked as the model value XHi of the computing power resources to be allocated; the number of floating-point operations of the computing power resources to be allocated is divided by the power value to obtain the energy efficiency ratio NXi; the utilization rate, the number of load tasks and the energy consumption ratio corresponding to the computing power resources to be allocated are marked as LYi, FZi and NHi respectively; the resource demand prediction model is substituted to output the resource demand prediction value of the computing power resources to be allocated; Sending the estimated value of the computing power resources to be allocated to the dynamic allocation management unit; The dynamic allocation management unit is used to dynamically allocate and manage heterogeneous computing resources.

2. A real-time monitoring and dynamic allocation management system for computing resources according to claim 1, characterized in that: Heterogeneous computing resources include CPU, GPU, FPGA, NPU and ASIC; status information includes performance data, utilization, number of load tasks and energy consumption data; performance data includes the number of floating-point operations, delay time, model and temperature; energy consumption data includes power and energy consumption ratio.

3. A real-time monitoring and dynamic allocation management system for computing resources according to claim 1, characterized in that: The specific process of obtaining the extended value of the computing power resources to be allocated is as follows: The delay durations are sorted according to the order of collection, and two adjacent delay durations are compared. If the delay duration in the previous order is greater than the delay duration in the next order, the delay duration in the previous order is subtracted from the delay duration in the next order to obtain the time drop difference value, and the values ​​of all the time drop difference values ​​are summed up to obtain the total time drop value; the average value of all delay duration values ​​is calculated to obtain the delay average value, and the maximum and minimum values ​​of all delay durations are obtained and marked as the maximum delay duration and the minimum delay duration respectively. The total time drop value, the delay average value, the maximum delay duration and the minimum delay duration are normalized and the values ​​of the four are taken, marked as TZ, TP, TMAX and TMIN respectively, and substituted into the preset delay analysis model: Output the delay value TYi of the computing power resources to be allocated, where f1, f2, f3 and f4 are the weight factors corresponding to the total time drop value, the average delay value, the maximum delay duration and the minimum delay duration respectively.

4. A real-time monitoring and dynamic allocation management system for computing resources according to claim 1, characterized in that: The resource demand prediction model is specifically: Output the estimated value ZFi of the computing power resources to be allocated; where hy1, hy2 and hy3 are model thresholds; is the extension weight factor, is the energy efficiency ratio weight factor; is the utilization weight factor; is the energy consumption ratio weight factor; is the weight factor of the load task quantity, i=1, 2, ..., n, where n is the maximum value of the number.

5. A real-time monitoring and dynamic allocation management system for computing resources according to claim 1, characterized in that: The specific process of dynamic allocation management is as follows: The heterogeneous computing resources are sorted from large to small according to the estimated value of the resources. If the number of target processing tasks is less than or equal to the number of heterogeneous computing resources, a preset number of heterogeneous computing resources are selected in sequence from front to back according to the sorting, and then the target processing tasks are equally divided and sent to the edge nodes corresponding to the heterogeneous computing resources in turn, and processed by the corresponding heterogeneous computing resources in the edge nodes; If the number of target processing tasks is greater than the number of heterogeneous computing resources, the target processing tasks are sent to the edge node corresponding to the heterogeneous computing resource with the highest ranking until its utilization rate is equal to the maximum utilization rate threshold, and then the sending is stopped; then the tasks are sent to the edge node corresponding to the second-ranked heterogeneous computing resource until all target processing tasks are distributed.

6. A real-time monitoring and dynamic allocation management system for computing resources according to claim 1, characterized in that: The central cloud is also provided with a fault detection processing unit; the fault detection processing unit is used to perform fault detection and analysis on heterogeneous computing resources, and the specific process is as follows: Obtain all temperatures corresponding to heterogeneous computing resources, compare the temperature with the corresponding preset maximum temperature, count the number of temperatures equal to or higher than the maximum temperature to obtain the number of heterogeneous temperatures; calculate the average value of all temperatures to obtain the average temperature. When the average temperature is higher than the set average threshold, obtain the voltage Uj and current Ij of the heterogeneous computing resources, j = 1, 2, ..., m, m represents the acquisition time number, m represents the maximum value of the number, substitute into the formula Output voltage fluctuation value SU and current fluctuation value SI; normalize the number of different temperatures, voltage fluctuation value, average temperature value and current fluctuation value and take the values ​​of the four and substitute them into the formula Obtain the fault detection value GYi of the heterogeneous computing power resources; where β1, β2, β3 and β4 are the weight factors corresponding to the number of different temperatures YW, the average temperature value PJ, the voltage fluctuation value SU and the current fluctuation value SI respectively; Set interval one and interval two; when the fault detection value belongs to interval one, shut down the edge node corresponding to the heterogeneous computing power resource, and then convert the fault detection value into shutdown duration according to a certain ratio, and then restart after executing the shutdown duration, and at the same time, feed back the load tasks not executed by the edge node to the dynamic allocation management unit for reallocation; when the fault detection value belongs to interval two, the edge node corresponding to the heterogeneous computing power resource executes the self-check and repair program, and at the same time, feeds back the load tasks not executed by the edge node to the dynamic allocation management unit for reallocation; after the self-check and repair program is completed, recalculate the fault detection value of the heterogeneous computing power resource. If the fault detection value still belongs to interval two, mark the edge node corresponding to the heterogeneous computing power resource as a faulty node, and send the coordinates of the faulty node, the name of the heterogeneous computing power resource and the fault detection value to the intelligent terminal of the corresponding maintenance personnel, and the maintenance personnel arrive at the location of the faulty node to repair or replace the heterogeneous computing power resource.

Images