Computing resource dynamic allocation method and system based on industrial Internet of Things data center

Abstract

The invention provides a computing resource dynamic allocation method and system based on an industrial Internet of Things data center, and the method comprises the steps: receiving a business demand of an enterprise user through an industrial Internet of Things user platform, and transmitting the business demand to a data computing center of an industrial Internet of Things management platform through an industrial Internet of Things service platform; monitoring resource data of the business management sub-platform through a monitoring module; determining resource demand characteristics of the enterprise user based on the business demand; determining resource allocation parameters based on the resource demand characteristics and the resource data of the service management sub-platforms; through the control center, according to the resource allocation parameters, a resource allocation instruction is generated, and the method can dynamically allocate computing resources for different sub-platforms so as to optimize execution of a computing task and ensure stable operation and efficient scheduling of a production line.

Description

Technical Field

[0001] This specification relates to the field of data processing, and particularly to a method and system for dynamically allocating computing resources based on an industrial Internet of Things data center. Background Art

[0002] With the advent of the Industrial 4.0 era, the industrial Internet of Things has become the core driving force for the intelligent transformation of the manufacturing industry. In a production line based on the industrial Internet of Things, numerous sensors, intelligent devices, and control systems are interconnected to build a highly integrated, data-driven production environment. In such an environment, the efficient operation of the production line depends on the real-time processing and analysis of a large amount of data, as well as the precise allocation of computing resources. However, in the face of the huge differences in different business requirements, how to dynamically allocate computing resources for the industrial Internet of Things platform under the limited computing resources of the data center has become an important challenge faced by industrial production.

[0003] Based on this, this specification provides a system and method for dynamically allocating computing resources based on an industrial Internet of Things data center, which dynamically allocates computing resources for different sub-platforms to optimize the execution of computing tasks and ensure the stable operation and efficient scheduling of the production line. Summary of the Invention

[0004] One or more embodiments of this specification provide a system for dynamically allocating computing resources based on an industrial Internet of Things data center. The system includes an industrial Internet of Things user platform, an industrial Internet of Things service platform, an industrial Internet of Things management platform, an industrial Internet of Things sensing network platform, and an industrial Internet of Things perception control platform. The industrial Internet of Things user platform is configured to receive the business requirements of enterprise users and send the business requirements to the data computing center of the industrial Internet of Things management platform through the industrial Internet of Things service platform. The industrial Internet of Things management platform includes the data computing center, a business management sub-platform, and a control center. The business management sub-platform includes a monitoring module, and the monitoring module is configured to monitor the resource data of the business management sub-platform. The control center is configured to: determine the resource requirement characteristics of the enterprise user based on the business requirements; determine the resource allocation parameters based on the resource requirement characteristics and the resource data of the business management sub-platform. The resource allocation parameters include the computing resources corresponding to the computing tasks. The control center is further configured to generate a resource allocation instruction according to the resource allocation parameters. The resource allocation instruction is configured to: establish a process corresponding to the computing task; bind the process corresponding to the computing task to the resource core corresponding to the computing task, and the resource core is determined based on the resource allocation parameters.

[0005] One or more embodiments of this specification provide a method for dynamically allocating computing resources based on an industrial Internet of Things data center. The method is executed by a control center of an industrial Internet of Things management platform of a computing resource dynamic allocation system based on an industrial Internet of Things data center. The system includes an industrial Internet of Things user platform, an industrial Internet of Things service platform, the industrial Internet of Things management platform, an industrial Internet of Things sensing network platform, and an industrial Internet of Things perception control platform. The industrial Internet of Things management platform includes a data computing center, a business management sub-platform, and the control center. The business management sub-platform includes a monitoring module. The method includes: receiving the business requirements of enterprise users through the industrial Internet of Things user platform, and sending the business requirements to the data computing center of the industrial Internet of Things management platform through the industrial Internet of Things service platform; monitoring the resource data of the business management sub-platform through the monitoring module; determining the resource requirement characteristics of the enterprise users based on the business requirements; determining the resource allocation parameters based on the resource requirement characteristics and the resource data of the business management sub-platform. The resource allocation parameters include the computing resources corresponding to the computing tasks; generating a resource allocation instruction according to the resource allocation parameters through the control center. The resource allocation instruction is configured to: establish a process corresponding to the computing task; bind the process corresponding to the computing task to the resource core corresponding to the computing task, and the resource core is determined based on the resource allocation parameters.

[0006] One or more embodiments of this specification provide a computer-readable storage medium that stores computer instructions. When a computer reads the computer instructions in the storage medium, the computer executes a method for dynamically allocating computing resources based on an industrial Internet of Things data center.

[0007] Beneficial effects: Based on the method and system for dynamically allocating computing resources based on an industrial Internet of Things data center of this application, it is possible to dynamically allocate computing resources based on the industrial Internet of Things, effectively improve the efficiency of business processing, and further improve resource utilization. BRIEF DESCRIPTION OF THE DRAWINGS

[0008] This specification will be further described in the form of exemplary embodiments, and these exemplary embodiments will be described in detail through the drawings. These embodiments are not restrictive. In these embodiments, the same numbers represent the same structures, where: Figure 1 is a schematic diagram of the platform structure of a computing resource dynamic allocation system based on an industrial Internet of Things data center shown in some embodiments of this specification; Figure 2 is an exemplary flowchart of a method for dynamically allocating computing resources based on an industrial Internet of Things data center shown in some embodiments of this specification; Figure 3 is an exemplary schematic diagram of a demand prediction model shown in some embodiments of this specification; Figure 4 is an exemplary flowchart for determining resource allocation parameters shown in some embodiments of this specification. Detailed implementation manners

[0009] To more clearly illustrate the technical solutions of the embodiments of this specification, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the drawings in the following description are only some examples or embodiments of this specification. For those of ordinary skill in the art, without creative efforts, this specification can also be applied to other similar scenarios based on these drawings. Unless obvious from the language context or otherwise stated, the same reference numerals in the drawings represent the same structure or operation.

[0010] It should be understood that the "system", "device", "unit" and / or "module" used herein is a method for distinguishing different components, elements, parts, portions or assemblies at different levels. However, if other words can achieve the same purpose, the said words can be replaced by other expressions.

[0011] As shown in this specification and the claims, unless the context clearly indicates an exception, words such as "a", "an", "one" and / or "the" are not specifically singular and may also include plural. Generally speaking, the terms "including" and "comprising" only indicate the inclusion of the clearly identified steps and elements, and these steps and elements do not constitute an exclusive list. The method or device may also include other steps or elements.

[0012] Flowcharts are used in this specification to illustrate the operations performed by the systems according to the embodiments of this specification. It should be understood that the operations before or after do not necessarily need to be executed precisely in sequence. On the contrary, the steps can be processed in reverse order or simultaneously. At the same time, other operations can also be added to these processes, or one or several steps can be removed from these processes.

[0013] Since enterprise users have different business types and different business volumes, and different businesses (such as data statistics, data screening, data analysis, data storage, etc.) involve different data types and processing requirements, and the proportions of resources occupied by the industrial Internet of Things system are also different. Therefore, there is a need for dynamic allocation of computing resources. Some embodiments of this specification provide a dynamic computing resource allocation system and method based on an industrial Internet of Things data center, which can dynamically allocate resources for the processing of different business requirements of enterprise users through the industrial Internet of Things.

[0014] Figure 1It is a schematic diagram of the platform structure of the computing resource dynamic allocation system based on the industrial Internet of Things data center shown in some embodiments of this specification.

[0015] In some embodiments, as Figure 1 shown, the computing resource dynamic allocation system 100 based on the industrial Internet of Things data center may include an industrial Internet of Things user platform 110, an industrial Internet of Things service platform 120, an industrial Internet of Things management platform 130, an industrial Internet of Things sensing network platform 140, and an industrial Internet of Things perception control platform 150.

[0016] The industrial Internet of Things user platform 110 is a platform for interacting with enterprise users. In some embodiments, the industrial Internet of Things user platform 110 is configured to receive the business requirements of enterprise users and send the business requirements to the industrial Internet of Things service platform 120.

[0017] In some embodiments, the industrial Internet of Things user platform 110 may be a platform based on a desktop computer, a tablet computer, a laptop computer, a mobile phone, or other electronic devices capable of data processing and data communication.

[0018] The industrial Internet of Things service platform 120 is a platform for processing the received business requirements. In some embodiments, the industrial Internet of Things service platform 120 may be implemented based on a terminal or a server, etc. In some embodiments, the industrial Internet of Things service platform 120 is configured to send the processed business requirements to the data computing center of the industrial Internet of Things management platform 130.

[0019] In some embodiments, the industrial Internet of Things service platform 120 may interact bidirectionally with the industrial Internet of Things management platform 130. For example, after the industrial Internet of Things management platform 130 calculates the resource allocation parameters according to the business requirements and resource requirement characteristics, it may upload the resource allocation result to the industrial Internet of Things service platform 120, and the industrial Internet of Things service platform 120 further uploads it to the enterprise users.

[0020] The industrial Internet of Things management platform 130 is a platform for controlling the computing resource dynamic allocation system 100 based on the industrial Internet of Things data center. In some embodiments, the industrial Internet of Things management platform 130 may be implemented based on a processor or a server, etc.

[0021] In some embodiments, the industrial Internet of Things management platform 130 may include a data computing center, a business management sub-platform, and a control center (not shown in the figure).

[0022] The data computing center can be used to manage resources and process data and / or information from at least one component of itself or external data sources (e.g., cloud data centers). In some embodiments, the data computing center can be configured with a single server or a server group, which can be centralized or distributed.

[0023] The business management sub-platform refers to the hardware device used to execute computing tasks in the computing resource dynamic allocation system 100 based on the industrial Internet of Things data center. In some embodiments, the business management sub-platform can include a monitoring module configured to monitor the resource data of the business management sub-platform. For example, the monitoring module can include server monitoring hardware, resource management software (such as Nagios, Zabbix, etc.), performance counters, etc.

[0024] The control center is a component used to control the target computing terminal to execute data calculation and processing. In some embodiments, the control center can include a computer device relying on a server. In some embodiments, the control center is configured to determine the resource demand characteristics of enterprise users based on business requirements; determine resource allocation parameters based on the resource demand characteristics and the resource data of the business management sub-platform; the resource allocation parameters include the computing resources corresponding to the computing tasks.

[0025] In some embodiments, the control center is further configured to: determine the resource demand characteristics of enterprise users based on the production status data and business requirements of enterprise users.

[0026] In some embodiments, the control center is further configured to: determine the resource demand characteristics of enterprise users based on the production status data and business requirements of enterprise users and based on a demand prediction model, where the demand prediction model is a machine learning model.

[0027] In some embodiments, the control center is further configured to: determine candidate parameters; determine the computing characteristics corresponding to the candidate parameters based on the candidate parameters, resource demand characteristics, and resource data; determine resource allocation parameters based on the computing characteristics.

[0028] In some embodiments, the control center is further configured to: determine the computing characteristics based on the candidate parameters, resource demand characteristics, and resource data through a feature prediction model, where the feature prediction model is a machine learning model; calculate the weighted values of the computing response speed, data congestion probability, and failure probability of the computing characteristics to determine the parameter evaluation value of the computing characteristics; determine resource allocation parameters based on the parameter evaluation value of the computing characteristics.

[0029] In some embodiments, the control center is further configured to generate a resource allocation instruction according to the resource allocation parameters; the resource allocation instruction is configured to: establish a process corresponding to the computing task; bind the process corresponding to the computing task to the resource core corresponding to the computing task, and the resource core is determined based on the resource allocation parameters.

[0030] In some embodiments, the industrial Internet of Things (IIoT) management platform 130 can interact with the industrial Internet of Things sensing network platform 140. For example, the industrial Internet of Things management platform 130 can send a data collection instruction to the industrial Internet of Things sensing network platform 140.

[0031] The industrial Internet of Things sensing network platform 140 is a network for data transmission in the computing resource dynamic allocation system 100 based on the industrial Internet of Things data center. In some embodiments, the industrial Internet of Things sensing network platform 140 can be configured with communication devices or servers, etc.

[0032] In some embodiments, the industrial Internet of Things sensing network platform 140 can interact with the industrial Internet of Things perception and control platform 150. For example, the industrial Internet of Things sensing network platform 140 can receive relevant data on the production line collected by the industrial Internet of Things perception and control platform 150 and transmit this data to the industrial Internet of Things management platform 130, which further processes and analyzes the data.

[0033] The industrial Internet of Things perception and control platform 150 is a platform for monitoring and controlling the production process.

[0034] In some embodiments, the industrial Internet of Things perception and control platform 150 includes a production monitoring device. The production monitoring device is a device for monitoring the production line. For example, the production monitoring device can include a current sensor, a temperature sensor, a gas metering device, etc. In some embodiments, the production monitoring device is deployed in enterprise users and is configured to obtain production status data of enterprise users.

[0035] For more descriptions of the above content, reference can be made to Figures 2 - 4 and its related descriptions.

[0036] In some embodiments of this specification, the computing resource dynamic allocation system 100 based on the industrial Internet of Things data center can form an information operation closed loop between the industrial Internet of Things user platform and the industrial Internet of Things perception and control platform, and coordinate and operate regularly under the unified management of the industrial Internet of Things management platform to achieve informatization and wisdom of computing resource management; at the same time, the industrial Internet of Things management platform can allocate computing resource configurations to different computing tasks according to different services of enterprise users, which can improve the overall data quality to improve the service quality.

[0037] It should be noted that the above description of the computing resource dynamic allocation system 100 based on the industrial Internet of Things data center is only for the convenience of description and does not limit this specification within the scope of the examples given. It can be understood that for those skilled in the art, after understanding the principle of the system, various platforms of the system can be arbitrarily combined without departing from this principle, or a subsystem can be formed and connected to other platforms.

[0038] Figure 2 is an exemplary flowchart of a computing resource dynamic allocation method based on an industrial Internet of Things data center according to some embodiments of this specification. As Figure 2 shown, process 200 includes the following steps. In some embodiments, process 200 can be executed by a control center.

[0039] Step 210, receive the business requirements of enterprise users through the industrial Internet of Things user platform, and send the business requirements to the data computing center of the industrial Internet of Things management platform through the industrial Internet of Things service platform.

[0040] Enterprise users refer to organizations that need to use the computing resources of the business management sub-platform for data processing. Among them, the business management sub-platform includes various devices for data processing. For example, the business management sub-platform includes multiple CPUs and / or GPUs. Computing resources refer to the computing resources of the business management sub-platform. For more content about the business management sub-platform, reference can be made to Figure 1 the corresponding description, and for more content about computing resources, reference can be made to the following description.

[0041] Business requirements refer to user requirement data related to business computing and / or business processing. For example, business requirements can include business type, upload time, business volume, etc. Among them, business volume can represent the amount of computing resources required for the current business.

[0042] In some embodiments, enterprise users input business requirements through the industrial Internet of Things user platform. After the industrial Internet of Things user platform receives the business requirements of enterprise users, it sends the business requirements to the industrial Internet of Things service platform for processing and further sends them to the data computing center of the industrial Internet of Things management platform.

[0043] Step 220, monitor the resource data of the business management sub-platform through a monitoring module.

[0044] The monitoring module can obtain the resource data of the business management sub-platform in various ways. For example, when the monitoring module is resource management software, it can remotely monitor various performance indicators of the business management sub-platform through network protocols (such as SNMP); for another example, when the monitoring module is a performance counter, the performance counter can provide real-time data of system resources (such as CPU, memory, disk I / O, network traffic, etc.) and applications, and the control center can obtain the resource data of the business management sub-platform based on the aforementioned real-time data.

[0045] Resource data refers to data related to the computing resources of the business management sub-platform. For example, resource data includes the occupied CPU and / or GPU of the business management sub-platform, the available occupation time of at least one CPU and / or GPU, etc.

[0046] Step 230, based on business requirements, determine the resource requirement characteristics of enterprise users.

[0047] Resource requirement characteristics refer to data related to the resources required when executing a certain business. For example, resource requirement characteristics include average resource requirements, etc. Average resource requirements refer to the average data of the resources required when executing the same business. For example, the average number of occupied CPU and / or GPU, the average occupation time of each CPU and / or GPU, etc.

[0048] In some embodiments, the control center can determine the resource requirement characteristics of enterprise users based on business requirements in various ways. For example, the control center can, according to the business requirements of enterprise users, query all the calculation processes during the historical calculation of such business requirements, and count the average resource requirements in all the above calculation processes as the corresponding resource requirement characteristics.

[0049] In some embodiments, the resource requirement characteristics further include peak characteristics, and the control center can also determine the resource requirement characteristics of enterprise users based on the production status data and business requirements of enterprise users.

[0050] Peak characteristics refer to the maximum value of the computing resources occupied when a certain business is executed. For example, the maximum number of occupied CPU and / or GPU, the maximum occupation duration of CPU and / or GPU, etc.

[0051] Production status data is data representing the characteristics related to the production process. For example, production status data includes the current production output of enterprise users, the number of operating production lines, product quality, etc.

[0052] In some embodiments, the production status data is obtained by a production monitoring device deployed in enterprise users.

[0053] For example, when the production monitoring device is a current sensor, the current sensor monitors the current changes on the production line, and the control center can judge the product quality during the production process according to the current changes. For example, the current fluctuation of the welding equipment can reflect the welding quality. For the specific description of the production monitoring device, please refer to Figure 1 the corresponding content of

[0054] In some embodiments, the control center can determine the resource demand characteristics of the enterprise user in various ways based on the production status data and business requirements of the enterprise user. For example, the control center can query all the calculation processes during the historical calculation of this type of business requirement according to the type of the current business requirement of the enterprise user and the production status data, count the peak characteristics and average resource demands of each of the above calculation processes, sort the peak characteristics, and select the largest peak characteristic and the average resource demand among them as the corresponding resource demand characteristics.

[0055] In some embodiments, the control center can also determine the resource demand characteristics of the enterprise user based on the production status data and business requirements of the enterprise user through a demand prediction model. For more content on this part, please refer to Figure 3 the corresponding description of

[0056] In some embodiments of this specification, considering the peak characteristics, it is possible to accurately control the number and time of CPUs and / or GPUs occupied by the calculation of the current business requirement, and reduce unnecessary resource waste.

[0057] Step 240, determine the resource allocation parameters based on the resource demand characteristics and the resource data of the business management sub-platform.

[0058] The resource allocation parameter refers to the data related to the computing resources of the business management sub-platform allocated to at least one computing task.

[0059] A computing task refers to a specific job or operation that needs to be executed in the business management sub-platform, usually involving data processing, calculation, and analysis, etc. The type of the computing task is related to the type corresponding to the enterprise user. Taking the computing tasks corresponding to a gas enterprise as an example, the computing tasks can include tasks such as calculating the future gas reserve volume and transmission volume based on the obtained data.

[0060] In some embodiments, the resource allocation parameter includes the computing resources allocated to the computing task. For example, the allocated CPU and / or GPU resources, including the number and occupation duration of the CPU and / or GPU, etc.

[0061] In some embodiments, the control center may determine resource allocation parameters in various ways based on resource demand characteristics and resource data of the business management sub-platform. For example, the control center may use linear programming to determine resource allocation parameters.

[0062] Exemplarily, the objective function of linear programming is shown as the following formulas (1) and (2): (1) (2) Wherein, represents minimizing the maximum completion time, which is an objective function for improving the computational response speed of all computing tasks; can characterize whether computing task (i) is assigned to CPU (j) or GPU (j), takes a value of 0 or 1. When is 1, it means that computing task (i) is assigned to CPU (j) or GPU (j), otherwise it is not assigned; CPU (j) and GPU (j) refer to the corresponding coding representations after dividing the CPU and GPU into coding areas, that is, the CPU of the j-th coding area and the GPU of the j-th coding area; represents the execution time of computing task (i) (that is, in the resource demand characteristics, the average resource demand of the business type corresponding to the business demand i of the enterprise user); represents maximizing the area utilization rate or balancing the load, that is, when the computing tasks are evenly distributed among the CPUs or GPUs, the minimum time difference of the load tasks among the CPUs or GPUs.

[0063] The constraint conditions corresponding to the objective function include that each computing task must be assigned to at least one coding area, and the load of each coding area cannot exceed its available time. The formula representations of the constraint conditions are shown as the following formulas (3) and (4), wherein the available time of each coding area can be determined based on the resource data of the obtained business management sub-platform: (3) (4) In some embodiments, the solution of the objective function can be achieved by various methods. For example, the Pareto optimal method, the ε-constraint method, etc.

[0064] In some embodiments, the control center may determine resource allocation parameters based on computing characteristics. For more content on this part, reference can be made to Figure 4 the corresponding description.

[0065] Step 250, through the control center, generate a resource allocation instruction according to the resource allocation parameters.

[0066] A resource allocation instruction refers to the strategy and mechanism for allocating resources.

[0067] In some embodiments, the resource allocation instruction is configured to: establish a process corresponding to a computing task according to resource allocation parameters; bind the process corresponding to the computing task to the resource core corresponding to the computing task, where the resource core is determined based on the resource allocation parameters.

[0068] A resource core, that is, a CPU core or a GPU core, is the basic processing unit of a CPU or a GPU. Each CPU core or GPU core can independently execute computing tasks, such as reading instructions, processing data, and executing programs.

[0069] In some embodiments of this specification, dynamically allocating computing resources based on the industrial Internet of Things can effectively improve the efficiency of business processing and further improve resource utilization.

[0070] Figure 3 It is an exemplary schematic diagram of a demand prediction model shown in some embodiments of this specification.

[0071] In some embodiments, the control center can determine the resource demand characteristics 340 of an enterprise user based on the production status data 310 and business requirements 320 of the enterprise user and based on the demand prediction model 330.

[0072] A demand prediction model is a model for determining resource demand characteristics. In some embodiments, the demand prediction model can be a machine learning model. For example, a Recurrent Neural Network (RNN) model, or other trained machine learning models.

[0073] In some embodiments, the input of the demand prediction model includes the production status data and business requirements of an enterprise user, and the output includes the resource demand characteristics of the enterprise user.

[0074] In some embodiments, the demand prediction model can be obtained through various methods. For example, it can be trained and obtained based on multiple labeled training samples through the gradient descent method or other feasible methods.

[0075] In some embodiments, the training samples and their corresponding labels can be determined based on historical data. The training samples can be obtained based on historical data that meets expectations. The training samples can include the production status data and business requirements of at least one historical enterprise user in the first historical period in the historical data that meets expectations, and the label can be the historical resource demand characteristics actually occupied by the corresponding historical enterprise user in the second historical period. Among them, the historical data that meets expectations can be the situation where the computing resources allocated to historical enterprise users meet their computing requirements, the load of the business management sub-platform does not exceed the limit, and the computing efficiency of the entire business management sub-platform is not lower than the preset threshold. Both the first historical period and the second historical period are periods in the past, and the first historical period is earlier than the second historical period.

[0076] Merely by way of example, the control center can input the training samples into the initial demand prediction model to obtain the output of the initial demand prediction model; construct a loss function based on the output of the initial demand prediction model and the label corresponding to the training samples; based on the loss function, iteratively update the parameters of the initial demand prediction model; until the iteration end condition is met, the training is completed, and a trained demand prediction model is obtained. Among them, the iteration end condition includes the convergence of the loss function, the number of iterations reaching the threshold, etc.

[0077] In some embodiments, the demand prediction model is obtained by training based on a training data set including multiple training samples. The training data set includes a training set and a test set.

[0078] The training set is a data set used to train the demand prediction model.

[0079] The test set is a data set used to evaluate the demand prediction model after the training is completed.

[0080] In some embodiments, the training set and the test set are determined based on the business type. Exemplarily, the control center can divide the business requirements of enterprise users and their associated production status data into multiple databases according to the business type, and randomly extract a preset proportion of samples from each of the multiple databases into the test set or the training set. Among them, the preset proportion refers to the preset proportion of extracting samples from the database to form the test set and the training set. For example, the preset proportion can be 3:7, then if the extracted samples are used to form the training set, the extraction proportion in the database is 0.7. The preset proportion can be preset based on experience or set by system default.

[0081] In some embodiments, among multiple training samples, different training samples have different learning rates, and the learning rate of the training sample is related to the computing characteristics corresponding to the training sample.

[0082] The learning rate is data used to control the size of the model weight update step when training the demand prediction model.

[0083] A computational feature refers to a feature related to the processing of training samples by the business management sub-platform. For example, computational features may include computational response speed, etc. The computational response speed can be determined based on the time required for the business management sub-platform to complete the corresponding computational task. The longer the required time, the smaller the computational response speed.

[0084] In some embodiments, the control center can determine the computational response speed of the business management sub-platform when processing various business requirements based on the historical computational data of the business management sub-platform, and determine the average computational response speed corresponding to each type of business requirement as the computational feature of this type of business requirement, and then determine the computational feature corresponding to the training sample according to the business type corresponding to the training sample.

[0085] In some embodiments, the learning rate corresponding to the training sample is positively correlated with its corresponding computational feature.

[0086] In some embodiments of this specification, using a training data set to train and test the model can improve the accuracy of the obtained model; when training the model, the computational response speed is also considered, further taking into account the impact of sample quality on the model. The higher the computational response speed of the training sample, the better the quality of the sample. In order to learn the implicit rules of such samples, the learning rate can be increased, so that the finally obtained model is more in line with the actual rules.

[0087] In some embodiments of this specification, through a machine learning model, the resources required for the current computation can be estimated more accurately; which is beneficial for subsequent resource allocation to improve the allocation efficiency.

[0088] Figure 4 is an exemplary flowchart of determining resource allocation parameters shown in some embodiments of this specification.

[0089] In some embodiments, the control center is further configured to: determine candidate parameters; based on the candidate parameters, resource requirement features, and resource data, determine the computational feature corresponding to the candidate parameters; and based on the computational feature, determine the resource allocation parameters.

[0090] In some embodiments, as Figure 4 shown, process 400 includes the following steps. Process 400 can be executed by the control center.

[0091] Step 410, determine candidate parameters.

[0092] Candidate parameters refer to the data that serve as alternative resource allocation parameters.

[0093] In some embodiments, the resource allocation parameters further include the bandwidth adjustment amount of the network device where the communication channel of the computational task is located.

[0094] A communication channel refers to the medium or path for transmitting information. For example, it can be the acquisition channel for the service requirements corresponding to a computing task or the transmission channel for the computing results of a computing task, etc. The service management sub-platform can include multiple communication channels, and different computing tasks can transmit data based on different communication channels.

[0095] The bandwidth adjustment amount refers to the ability or mechanism to adjust the data transmission rate in network communication. The bandwidth adjustment amount can include the target bandwidth amount to be adjusted, etc. For example, by increasing the bandwidth of the network device where the communication channel is located, the data transmission rate of this communication channel can be increased, which is conducive to improving the execution efficiency of the computing task corresponding to this communication channel.

[0096] In some embodiments, the control center can determine candidate parameters in various ways. For example, the control center can count the usage times of each resource allocation parameter in the historical data, and select the top N resource allocation parameters with the most usage times as candidate parameters. Among them, the value of N can be preset as needed. Another example is that the control center can randomly generate multiple resource allocation parameters as candidate parameters.

[0097] Step 420: Based on the candidate parameters, resource demand characteristics, and resource data, determine the computing characteristics corresponding to the candidate parameters.

[0098] In some embodiments, the control center can determine the computing characteristics corresponding to the candidate parameters in various ways based on the candidate parameters, resource demand characteristics, and resource data. For example, the control center can determine the computing characteristics corresponding to the candidate parameters by means of vector matching.

[0099] Exemplarily, the control center can construct a vector database based on historical data. The vector database can include at least one reference vector and its corresponding label.

[0100] The control center can construct at least one clustering vector based on the executed historical resource allocation parameters in the historical data, as well as the historical resource demand characteristics and historical resource data corresponding to the historical resource allocation parameters; cluster the at least one clustering vector to form a preset number of cluster centers; construct at least one reference vector based on the historical candidate parameters, historical resource demand characteristics, and historical resource data corresponding to the cluster centers; and use the respective historical computing characteristics corresponding to each cluster center as the label of its corresponding reference vector. The preset number can be preset based on requirements. There are various clustering algorithms, such as K-means clustering, hierarchical clustering, etc.

[0101] In some embodiments, the control center may construct a feature vector based on candidate parameters, resource demand characteristics, and resource data; perform a match in the vector database based on the feature vector to determine a reference vector with the highest similarity to the feature vector, and use the label of the reference vector as the calculation feature corresponding to the candidate parameter. Among them, the similarity can be determined based on the vector distance. The smaller the vector distance, the higher the similarity.

[0102] In some embodiments, the control center is further configured to: determine the calculation feature based on the candidate parameter, resource demand characteristics, and resource data through a feature prediction model.

[0103] In some embodiments, the calculation feature further includes the data congestion probability and the failure probability.

[0104] The data congestion probability refers to the probability that the service demand needs to wait for the CPU and / or GPU resources to be released before being used.

[0105] The failure probability refers to the probability of an accident occurring during the calculation process that causes the service demand not to be executed normally. For example, the failure probability includes the probability of problems such as suspension, insufficient memory, and excessive network transmission time occurring during the calculation process. Suspension refers to the state where a process, program, or system temporarily stops execution or cannot continue to run normally.

[0106] The feature prediction model is a model used to determine the calculation feature. In some embodiments, the feature prediction model can be a machine learning model. For example, a Neural Network (NN) model, or other trained machine learning models.

[0107] In some embodiments, the input of the feature prediction model includes candidate parameters, resource demand characteristics, and resource data, and the output includes the calculation feature.

[0108] In some embodiments, the feature prediction model can be obtained through various methods. For example, it can be trained and obtained based on multiple second training samples with second labels through the gradient descent method or other feasible methods.

[0109] In some embodiments, the second training sample and its corresponding second label can be determined based on historical data. The second training sample may include the historical resource allocation parameters actually executed in the historical data, and the historical resource demand characteristics of at least one historical enterprise user and the historical resource data of at least one historical business management sub - platform corresponding thereto. The second label may include the historical calculation feature corresponding to the actual execution of the historical resource allocation parameter in the historical data. For example, the calculation response speed, data congestion probability, and failure probability when the historical resource allocation parameter is actually executed. Among them, if data congestion occurs, the data congestion probability is 1, otherwise it is 0, and the determination of the failure probability is the same.

[0110] The training method of the feature prediction model is similar to that of the demand prediction model, which can be referred to the previous description.

[0111] Step 430: Determine the resource allocation parameters based on the calculated features.

[0112] In some embodiments, the control center can determine the resource allocation parameters based on the calculated features in various ways. For example, the control center can count multiple calculated features, and select the candidate parameters corresponding to the calculated features whose calculation response speed is greater than the preset speed threshold and the data congestion probability and failure probability are lower than the preset probability threshold as the finally used resource allocation parameters. Among them, the preset speed threshold refers to the maximum value of the preset calculation response speed, and the preset probability threshold refers to the minimum values of the preset data congestion probability and failure probability. The preset speed threshold and the preset probability threshold can be preset based on experience or default settings of the system.

[0113] In some embodiments, the control center can determine the parameter evaluation value of the calculated feature based on the weighted values of the calculation response speed, data congestion probability, and failure probability of the calculated feature; and determine the resource allocation parameters based on the parameter evaluation value of the calculated feature.

[0114] The parameter evaluation value can represent the evaluation result of the comprehensive performance of the resource allocation parameter. The larger the parameter evaluation value is, the more the resource allocation parameter should be preferentially applied. For example, the control center can use the candidate parameter corresponding to the calculated feature with the largest parameter evaluation value as the resource allocation parameter.

[0115] In some embodiments, the control center can obtain the weighted values of the calculation response speed, data congestion probability, and failure probability of the calculated feature in various ways. For example, the control center can obtain the foregoing weighted values through a preset formula. Exemplarily, the preset formula is shown as the following formula (5): (5) Wherein, represents the weighted value, represents the calculation response speed, represents the data congestion probability, represents the failure probability, 、 、 are preset coefficients. Among them, 、 、 can be preset based on experience.

[0116] In some embodiments, the weighted values of the calculation response speed, data congestion probability, and failure probability of the calculated feature are related to the resource demand characteristics of enterprise users, that is 、 、 It can be determined based on the resource requirement characteristics of enterprise users.

[0117] In some embodiments, the greater the average resource requirement in the resource requirement characteristics of enterprise users, the smaller it is; the greater the difference between the total data volume of the corresponding business requirements of enterprise users and the available computing resources of the business management sub-platform, the smaller it is; the greater the total data volume of the business requirements of enterprise users, the smaller it is.

[0118] In some embodiments of this specification, by associating each weight with the resource requirement characteristics of enterprise users, the obtained weighted value can be made more accurate, and thus the resource allocation parameters determined subsequently can be more in line with the actual situation.

[0119] In some embodiments of this specification, through a machine learning model, the calculation response speed, data congestion probability, and failure probability are predicted, which can ensure the accuracy of the obtained calculation characteristics and reduce the time waste in the prediction process; based on the weighted values of each item of the calculation characteristics, a parameter evaluation value is determined, and then the resource allocation parameters are determined, which is beneficial to finding the best solution and thus effectively improving the actual service effect.

[0120] In some embodiments, the resource allocation instruction is configured to: obtain the communication information of the computing task; determine the communication channel of the computing task according to the communication information; and adjust the bandwidth allocation strategy of the network device where the communication channel of the computing task is located according to the bandwidth adjustment amount corresponding to the communication channel in the resource allocation parameters, so as to adjust the bandwidth of the communication channel of the computing task.

[0121] The communication information refers to the data related to the communication of the business requirements in the computing task. In some embodiments, the communication information includes at least one of the communication source address, communication destination address, and communication relay address corresponding to the computing task.

[0122] The communication source address refers to the address of the device or node that sends information in the network. For example, the address of the device or node that sends business requirements in the network.

[0123] The communication destination address refers to the address of the target device or node to which the information is to be sent. For example, the address of the target device or node to which the calculation result is to be sent.

[0124] The communication relay address refers to the address of the intermediate device or node that the data may pass through during the transmission from the source address to the destination address.

[0125] In some embodiments, when a process is created, the control center can directly obtain the communication information based on the business requirements.

[0126] In some embodiments, the control center determines the communication channel for each computing task by querying the routing table on the network device based on the communication source address, communication destination address, and communication relay address of each computing task. Herein, the network device is an electronic device used to connect and manage computers and other devices in the network to achieve data transmission, communication, and network services. The routing table is a data structure on the network device used to store path information. The routing table records the destination address of data transmission, the address of the next network device, etc. The routing table can be set manually by the staff or automatically updated by a dynamic routing protocol. The dynamic routing protocol is a mechanism used to automatically discover and maintain routing information in a computer network.

[0127] The bandwidth allocation policy refers to the methods and rules for allocating available bandwidth to different users, devices, or applications in the network. For example, the bandwidth allocation policy can include the bandwidth allocation amount and allocation priority of each network device where the communication channels are located. The bandwidth allocation policy can be determined by any feasible method.

[0128] In some embodiments, the control center adjusts the bandwidth of the communication channel for each task by adjusting the bandwidth allocation policy of the network device where the communication channel for each task is located according to the bandwidth adjustment amount corresponding to the communication channel for each task in the resource allocation parameters. For example, in the bandwidth allocation policy of the network device where the communication channel is located, the corresponding bandwidth allocation amount is increased to the bandwidth adjustment amount corresponding to the resource allocation parameters.

[0129] In some embodiments of this specification, by generating multiple candidate parameters and determining the best solution by evaluating the actual computing characteristics of each candidate parameter, the actual service effect can be effectively improved.

[0130] The basic concepts have been described above. Obviously, for those skilled in the art, the above detailed disclosure is only an example and does not constitute a limitation to this specification. Although not explicitly stated here, those skilled in the art may make various modifications, improvements, and corrections to this specification. Such modifications, improvements, and corrections are proposed in this specification, so such modifications, improvements, and corrections still fall within the spirit and scope of the exemplary embodiments of this specification.

[0131] Meanwhile, this specification uses specific terms to describe the embodiments of this specification. Such as "one embodiment", "an embodiment", and / or "some embodiments" mean a certain feature, structure, or characteristic related to at least one embodiment of this specification. Therefore, it should be emphasized and noted that "an embodiment" or "one embodiment" or "an alternative embodiment" mentioned twice or more at different positions in this specification does not necessarily refer to the same embodiment. In addition, certain features, structures, or characteristics in one or more embodiments of this specification can be combined appropriately.

[0132] In addition, unless otherwise specified in the claims, the order of the processing elements and sequences, the use of numerical and alphabetical characters, or the use of other names described in this specification are not used to limit the order of the processes and methods in this specification. Although some currently useful embodiments of the invention are discussed through various examples in the above disclosure, it should be understood that such details are for illustrative purposes only. The appended claims are not limited to the disclosed embodiments. On the contrary, the claims are intended to cover all modifications and equivalent combinations that conform to the essence and scope of the embodiments of this specification. For example, although the system components described above can be implemented by hardware devices, they can also be implemented only through software solutions, such as installing the described system on existing servers or mobile devices.

[0133] Similarly, it should be noted that, in order to simplify the presentation of the disclosure in this specification and thus help the understanding of one or more embodiments of the invention, in the foregoing description of the embodiments of this specification, multiple features are sometimes grouped into one embodiment, drawing, or description thereof. However, this method of disclosure does not mean that the features required by the subject matter of this specification are more than those mentioned in the claims. In fact, the features of the embodiments are fewer than all the features of the individual embodiments disclosed above.

[0134] In some embodiments, numbers are used to describe the components and the quantity of attributes. It should be understood that such numbers used to describe the embodiments are modified by the modifiers "about", "approximate" or "substantially" in some examples. Unless otherwise specified, "about", "approximate" or "substantially" indicate that the stated number allows a variation of ±20%. Accordingly, in some embodiments, the numerical parameters used in the specification and claims are approximate values, and such approximate values may vary according to the characteristics required by individual embodiments. In some embodiments, the numerical parameters should consider the specified significant digits and adopt the method of retaining the general number of digits. Although the numerical ranges and parameters used in some embodiments of this specification to confirm the breadth of their scope are approximate values, in specific embodiments, such numerical settings are as precise as possible within the feasible range.

[0135] For each patent, patent application, patent application publication, and other materials cited in this specification, such as articles, books, specifications, publications, documents, etc., their entire contents are hereby incorporated into this specification by reference. This excludes the application history documents that are inconsistent with or conflict with the content of this specification, and also excludes the documents that limit the broadest scope of the claims of this specification (currently or subsequently appended to this specification). It should be noted that if there are any inconsistencies or conflicts between the descriptions, definitions, and / or uses of terms in the supplementary materials of this specification and the content described in this specification, the descriptions, definitions, and / or uses of terms in this specification shall prevail.

[0136] Finally, it should be understood that the embodiments described in this specification are only used to illustrate the principles of the embodiments of this specification. Other variations may also fall within the scope of this specification. Therefore, by way of example and not limitation, alternative configurations of the embodiments of this specification may be regarded as consistent with the teachings of this specification. Accordingly, the embodiments of this specification are not limited to the embodiments explicitly presented and described in this specification.

Claims

1. A computing resource dynamic allocation system based on an industrial Internet of Things data center, characterized in that: The system includes an industrial Internet of Things user platform, an industrial Internet of Things service platform, an industrial Internet of Things management platform, an industrial Internet of Things sensor network platform and an industrial Internet of Things perception control platform; The industrial Internet of Things user platform is configured to receive business requirements of enterprise users and send the business requirements to the data computing center of the industrial Internet of Things management platform through the industrial Internet of Things service platform; The industrial Internet of Things management platform includes the data computing center, the business management sub-platform and the control center; the business management sub-platform includes a monitoring module, and the monitoring module is configured to monitor the resource data of the business management sub-platform; The control center is configured to: Based on the business needs, determining resource demand characteristics of the enterprise user; Determining resource allocation parameters based on the resource demand characteristics and the resource data of the business management sub-platform; the resource allocation parameters include computing resources corresponding to the computing tasks; The control center is further configured to generate a resource allocation instruction according to the resource allocation parameter; the resource allocation instruction is configured to: Establishing a process corresponding to the computing task; The process corresponding to the computing task is bound to a resource core corresponding to the computing task, where the resource core is determined based on the resource allocation parameter.

2. The system according to claim 1, characterized in that The industrial Internet of Things perception control platform also includes a production monitoring device, which is deployed in the enterprise user and is configured to obtain production status data of the enterprise user; The resource demand characteristics also include peak characteristics; The control center is further configured to: The resource demand characteristics of the enterprise user are determined based on the production status data and the business needs of the enterprise user.

3. The system according to claim 2, characterized in that The control center is further configured to determine the resource demand characteristics of the enterprise user based on the production status data and the business needs of the enterprise user and based on a demand prediction model, and the demand prediction model is a machine learning model.

4. The system according to claim 1, characterized in that The resource allocation parameters also include a bandwidth adjustment amount of the network device where the communication channel of the computing task is located; the control center is further configured to: Determine candidate parameters; Determine, based on the candidate parameters, the resource demand characteristics, and the resource data, a calculation characteristic corresponding to the candidate parameters; Based on the calculation characteristics, determining the resource allocation parameter; The resource allocation instructions are configured to: Acquire communication information of the computing task, the communication information including at least one of a communication source address, a communication destination address, and a communication transit address; Determining a communication channel for the computing task according to the communication information; According to the bandwidth adjustment amount corresponding to the communication channel in the resource allocation parameter, the bandwidth allocation strategy of the network device where the communication channel of the computing task is located is adjusted to adjust the bandwidth of the communication channel of the computing task.

5. The system according to claim 4, characterized in that The computing characteristics include computing response speed, data congestion probability, and failure probability; The control center is further configured to: Based on the candidate parameters, the resource demand characteristics, and the resource data, the calculation characteristics are determined by a characteristic estimation model, wherein the characteristic estimation model is a machine learning model; Determining a parameter evaluation value of the calculation feature based on a weighted value of the calculation response speed, the data congestion probability, and the failure probability of the calculation feature; The resource allocation parameter is determined based on the parameter evaluation value of the calculation characteristic.

6. A method for dynamically allocating computing resources based on an industrial Internet of Things data center, characterized in that: The method is executed by a control center of an industrial Internet of Things management platform of a computing resource dynamic allocation system based on an industrial Internet of Things data center, wherein the system includes an industrial Internet of Things user platform, an industrial Internet of Things service platform, the industrial Internet of Things management platform, an industrial Internet of Things sensor network platform and an industrial Internet of Things perception control platform; The industrial Internet of Things management platform includes a data computing center, a business management sub-platform and the control center; The business management sub-platform includes a monitoring module; The method comprises: Receiving business needs of enterprise users through the industrial Internet of Things user platform, and sending the business needs to the data computing center of the industrial Internet of Things management platform through the industrial Internet of Things service platform; Monitoring resource data of the business management sub-platform through the monitoring module; Based on the business needs, determining resource demand characteristics of the enterprise user; Determining the resource allocation parameters based on the resource demand characteristics and the resource data of the business management sub-platform; the resource allocation parameters include computing resources corresponding to the computing tasks; Generate a resource allocation instruction through the control center according to the resource allocation parameter; the resource allocation instruction is configured as follows: Establishing a process corresponding to the computing task; The process corresponding to the computing task is bound to a resource core corresponding to the computing task, where the resource core is determined based on the resource allocation parameter.

7. The method according to claim 6, characterized in that The resource demand characteristics also include peak characteristics. The determining of the resource demand characteristics of the enterprise user based on the business demand includes: Based on the production status data of the enterprise user and the business demand, the resource demand characteristics of the enterprise user are determined; wherein the production status data is acquired by a production monitoring device deployed in the enterprise user.

8. The method according to claim 7, characterized in that The determining the resource demand characteristics of the enterprise user based on the production status data of the enterprise user and the business demand includes: Based on the production status data and the business needs of the enterprise user and based on a demand prediction model, the resource demand characteristics of the enterprise user are determined, and the demand prediction model is a machine learning model.

9. The method according to claim 6, characterized in that The resource allocation parameter also includes a bandwidth adjustment amount of a network device where the communication channel of the computing task is located. The resource allocation parameter is determined based on the resource demand characteristics and the resource data of the service management sub-platform, including: Determine candidate parameters; Determine, based on the candidate parameters, the resource demand characteristics, and the resource data, a calculation characteristic corresponding to the candidate parameters; Based on the calculation characteristics, determining the resource allocation parameter; The resource allocation instructions are configured to: Acquire communication information of the computing task, the communication information including at least one of a communication source address, a communication destination address, and a communication transit address; Determining a communication channel for the computing task according to the communication information; According to the bandwidth adjustment amount corresponding to the communication channel in the resource allocation parameter, the bandwidth allocation strategy of the network device where the communication channel of the computing task is located is adjusted to adjust the bandwidth of the communication channel of the computing task.

10. The method according to claim 9, characterized in that The calculation characteristics include calculation response speed, data congestion probability, and failure probability; and determining the resource allocation parameters based on the calculation characteristics includes: Based on the candidate parameters, the resource demand characteristics, and the resource data, the calculation characteristics are determined by a characteristic estimation model, wherein the characteristic estimation model is a machine learning model; Determining a parameter evaluation value of the calculation feature based on a weighted value of the calculation response speed, the data congestion probability, and the failure probability of the calculation feature; The resource allocation parameter is determined based on the parameter evaluation value of the calculation characteristic.

Images