A resource scheduling system for a prefabricated data center

Abstract

The present invention discloses a resource scheduling system for a prefabricated data center, which relates to the related field of control and regulation systems, and includes: obtaining the configuration topology information of multiple prefabricated units by interacting with the target scenario. Making an initial static scheduling decision according to the real-time task requirements to generate an initial resource scheduling policy set. Analyzing the configuration topology information to obtain the connectivity characteristics of multiple prefabricated units and obtaining the real-time resource status information. Conducting iterative decision optimization of the initial resource scheduling policy set to obtain an optimized scheduling policy set. Selecting the top N optimized scheduling policies with the best performance from the self-optimized scheduling policy set, outputting them as a resource scheduling plan set, and performing resource scheduling control of the target prefabricated data center according to the resource scheduling plan set. It solves the technical problems in the prior art that the resource scheduling of prefabricated data centers has poor scheduling accuracy and it is difficult to obtain the optimal resource scheduling policy in a changing environment, resulting in difficult improvement of resource utilization rate and task execution efficiency.

Description

Technical Field

[0001] The present application relates to the field of control and regulation systems, and particularly to a resource scheduling system for a prefabricated data center. Background Art

[0002] With the rapid development of emerging technologies such as cloud computing, big data, and artificial intelligence, as the basic platform for information processing and storage, the scale and complexity of data centers have been continuously increasing. Traditional data centers are mostly based on fixed facilities and independent computer rooms, with a long deployment cycle, limited scalability, high energy consumption, and difficulty in flexibly adapting to changing computing and storage requirements. As a modular and highly integrated solution, a prefabricated data center modularly encapsulates infrastructure such as servers, network devices, refrigeration, and power into standard units (prefabricated units), and then constructs an overall data center through reasonable topological connections. Compared with traditional data centers, prefabricated data centers can be deployed more quickly, expanded more flexibly, maintained more conveniently, and the resources can be flexibly combined according to application scenarios. However, within a prefabricated data center, the network topology composed of numerous prefabricated units is extremely complex, the task requirements change in real time, and the computing resources and environmental parameters fluctuate dynamically, making it difficult to accurately schedule and control the resources of the prefabricated data center.

[0003] Therefore, in the prior art, there are technical problems in the resource scheduling of prefabricated data centers, such as poor scheduling accuracy, difficulty in obtaining the optimal resource scheduling strategy in a changing environment, resulting in difficulties in improving resource utilization and task execution efficiency. Summary of the Invention

[0004] The present application provides a resource scheduling system for a prefabricated data center, which solves the technical problems in the prior art that the resource scheduling of prefabricated data centers has poor scheduling accuracy, difficulty in obtaining the optimal resource scheduling strategy in a changing environment, resulting in difficulties in improving resource utilization and task execution efficiency. Through multi-round iterative decision-making optimization and sample point merging and simplification strategies, combined with real-time equivalent connectivity characteristics and link pressure assessment, a set of resource scheduling plans that can be efficiently and dynamically invoked is finally obtained, thereby improving the resource utilization, task execution efficiency, and operation stability of the prefabricated data center.

[0005] This application provides a resource scheduling system for a prefabricated data center. The system includes: an information collection module for interacting with a target scenario to obtain the configuration topology information of multiple prefabricated units in the target prefabricated data center; a static scheduling decision acquisition module for making an initial static scheduling decision according to the real-time task requirements of the target prefabricated data center and generating an initial resource scheduling policy set; a resource status acquisition module for parsing the configuration topology information to obtain the connectivity characteristics of multiple prefabricated units in the target prefabricated data center and accessing in real time to obtain the real-time resource status information of multiple prefabricated units, where the real-time resource status information includes computing resource status and environmental resource status; an optimized scheduling policy acquisition module for performing iterative decision optimization on the initial resource scheduling policy set according to the connectivity characteristics and the real-time resource status information, and storing multiple optimization results as an optimized scheduling policy set; a scheduling control module for selecting the top N optimal optimized scheduling policies from the optimized scheduling policy set, outputting them as a resource scheduling plan set, and performing resource scheduling control on the target prefabricated data center according to the resource scheduling plan set.

[0006] In a possible implementation, the static scheduling decision acquisition module is further configured to: access the task management end of the target prefabricated data center to extract a real-time task list; parse the real-time task list to obtain the resource consumption data of each real-time task, and output it as a real-time resource expectation; perform iterative initial scheduling decisions based on the real-time resource expectation and a preset static resource scheduling policy to obtain the initial resource scheduling policy set.

[0007] In a possible implementation, the optimized scheduling policy acquisition module is further configured to: according to the configuration topology information, obtain the configuration parameter information of multiple prefabricated units in the target prefabricated data center, and define an optimization constraint set according to the configuration parameter information, where the optimization constraint set at least includes unit resource capacity constraint, allocation consistency constraint, cross-module allocation constraint, module temperature constraint, and link load constraint; based on the optimization constraint set, perform repair crossover on the initial resource scheduling policy set to obtain a repaired resource scheduling policy set; combine the initial resource scheduling policy set and the repaired resource scheduling policy set to construct an optimization sample set, and perform iterative decision optimization based on the optimization sample set and the optimization constraint set in combination with a preset objective function.

[0008] In a possible implementation manner, the optimization scheduling strategy obtaining module is further configured to: in each iterative decision-making optimization, traverse the optimization sample set to extract a co-target cluster, where the co-target cluster is a sample subset with a common intersection point in the optimization direction, where the sample subset includes at least N sample points, and N is greater than or equal to 3; calculate the objective function value of the common intersection point, and compare it with the objective function values of the N sample points in the co-target cluster; if the objective function value of the common intersection point is the optimal objective function value, merge the N sample points in the co-target cluster to the common intersection point to generate a new sample point.

[0009] In a possible implementation manner, the optimization scheduling strategy obtaining module is further configured to: if the objective function value of the common intersection point is the optimal objective function value, traverse the co-target cluster, calculate the distances between the N sample points and the common intersection point, and calculate the variance of the distances between the N sample points and the common intersection point; if the distances between the N sample points and the common intersection point are all less than a preset merging radius, and the variance of the distances between the N sample points and the common intersection point is less than or equal to a preset variance, use the common intersection point as the new sample point, and calculate the vector sum of the optimization directions and optimization step lengths of the N sample points in the co-target cluster as the new optimization direction and optimization step length; remove the N sample points in the co-target cluster, and merge the new sample point into the optimization sample set.

[0010] In a possible implementation manner, the optimization scheduling strategy obtaining module is further configured to: check the optimization sample set after generating a new sample point, and count the number of sample points; if the number of sample points is less than a first preset sample scale, abort the extraction of the co-target cluster and the merging of sample points; if the number of sample points is less than a second preset sample scale, perform random mutation on the new sample point until the number of sample points is greater than or equal to the second preset sample scale and less than the first preset sample scale, where the second preset sample scale is less than the first preset sample scale.

[0011] In a possible implementation manner, the scheduling control module is further configured to: fit the connectivity feature and the real-time resource status information to obtain a real-time equivalent connectivity feature; based on the real-time equivalent connectivity feature, evaluate the link pressure coefficient of the optimization results in the optimization scheduling strategy set, and correspondingly generate a link pressure sequence; select the first N optimization results in the link pressure sequence as a resource scheduling pre-plan set, and perform resource scheduling control on the target prefabricated data center according to the resource scheduling pre-plan set.

[0012] In a possible implementation, the scheduling control module is further configured to: serialize the resource scheduling plan set to generate a plan call sequence; perform resource scheduling control with the first resource scheduling plan at the top of the plan call sequence, and monitor the resource usage of the target prefabricated data center in real time; if the real-time monitoring result shows that the resource usage exceeds the limit, call the next resource scheduling plan of the first resource scheduling plan in the plan call sequence to perform resource scheduling control until the real-time monitoring result shows that the resource usage does not exceed the limit; based on the resource scheduling plan called when the resource usage does not exceed the limit, perform feedback adjustment of iterative decision-making optimization.

[0013] One or more technical solutions provided in this application have at least the following technical effects or advantages:

[0014] A resource scheduling system for a prefabricated data center provided in this application includes: obtaining the configuration topology information of multiple prefabricated units in the target prefabricated data center by interacting with the target scenario. Making an initial static scheduling decision according to the real-time task requirements of the target prefabricated data center to generate an initial resource scheduling strategy set. Analyzing the configuration topology information to obtain the connectivity characteristics of multiple prefabricated units in the target prefabricated data center, and accessing in real time to obtain the real-time resource status information of multiple prefabricated units. According to the connectivity characteristics and the real-time resource status information, perform iterative decision-making optimization of the initial resource scheduling strategy set, and store multiple optimization results as an optimized scheduling strategy set. Select the top N optimal optimized scheduling strategies from the optimized scheduling strategy set, output them as a resource scheduling plan set, and perform resource scheduling control of the target prefabricated data center according to the resource scheduling plan set. It solves the technical problems in the prior art that the resource scheduling of prefabricated data centers has poor scheduling accuracy and it is difficult to obtain the optimal resource scheduling strategy in a changing environment, resulting in difficult improvement of resource utilization rate and task execution efficiency. Through multiple rounds of iterative decision-making optimization and sample point merging and simplification strategies, combined with real-time equivalent connectivity characteristics and link pressure assessment, a set of resource scheduling plans that can be efficiently and dynamically called is finally obtained, thereby improving the resource utilization rate, task execution efficiency and operation stability of the prefabricated data center. Description of the Drawings

[0015] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings of the embodiments of the present invention will be briefly introduced below. Flowcharts are used in this application to illustrate the operations performed by the system according to the embodiments of the present application. It should be understood that the operations in the front or below do not necessarily need to be executed precisely in sequence. On the contrary, according to needs, they can be executed in reverse order or simultaneously. At the same time, other operations can also be added to these processes, or one or several operations can be removed from these processes.

[0016] Figure 1 This is a schematic structural diagram of a resource scheduling system for a prefabricated data center provided by an embodiment of the present application.

[0017] Figure 2 This is a schematic flowchart of obtaining an initial resource scheduling policy set in a resource scheduling system for a prefabricated data center of the present application.

[0018] Explanation of reference numerals: Information collection module 11, static scheduling decision acquisition module 12, resource status acquisition module 13, optimized scheduling policy acquisition module 14, scheduling control module 15. Detailed implementation manners

[0019] The above description is only an overview of the technical solution of the present application. In order to be able to understand the technical means of the present application more clearly, it can be implemented according to the content of the specification. And in order to make the above and other purposes, features and advantages of the present application more obvious and understandable, the following specifically gives the detailed implementation manners of the present application.

[0020] In order to make the purpose, technical solution and advantages of the present application clearer, the present application will be further described in detail below with reference to the accompanying drawings. The described embodiments should not be regarded as limitations of the present application. All other embodiments obtained by those of ordinary skill in the art without creative efforts fall within the scope of protection of the present application.

[0021] In the following description, reference is made to "some embodiments", which describe a subset of all possible embodiments. However, it can be understood that "some embodiments" can be the same subset or different subsets of all possible embodiments, and can be combined with each other without conflict. The terms "first" and "second" involved are only used to distinguish similar objects and do not represent a specific order for the objects. The terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, system, product or server that includes a series of steps or units does not necessarily have to be limited to those steps or units clearly listed, but may include other steps or modules that are not clearly listed or are inherent to these processes, systems, products or devices. Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those of ordinary skill in the technical field to which the present application belongs. The terms used herein are only for the purpose of describing the embodiments of the present application.

[0022] An embodiment of the present application provides a resource scheduling system for a prefabricated data center, as Figure 1 shown. The system includes:

[0023] An information collection module 11, configured to interact with a target scenario and obtain configuration topology information of multiple prefabricated units in a target prefabricated data center.

[0024] The static scheduling decision acquisition module 12 is used to make an initial static scheduling decision according to the real-time task requirements of the target prefabricated data center, and generate an initial resource scheduling policy set.

[0025] The resource status acquisition module 13 is used to parse the configuration topology information, obtain the connectivity characteristics of multiple prefabricated units of the target prefabricated data center, and access in real time to obtain the real-time resource status information of multiple prefabricated units, where the real-time resource status information includes the computing resource status and the environmental resource status.

[0026] A prefabricated data center is a modular data center structure formed by pre-integrating and encapsulating infrastructure modules such as cabinets, servers, network switching devices, refrigeration systems, and power systems in standardized modules, i.e., prefabricated units, and then connecting several such prefabricated units through a certain topology. By interacting with the target scenario, the target scenario is the application configuration scenario of the target prefabricated data center, such as the prefabricated data center in a server, etc., and record various parameters and internal configurations of the target prefabricated data center in the target scenario. Obtain the configuration topology information of multiple prefabricated units in the target prefabricated data center of the target scenario, where the configuration topology information is the connection relationship and distribution structure of each prefabricated unit in the target prefabricated data center and the configuration information of the corresponding prefabricated unit, etc. Subsequently, make an initial static scheduling decision according to the real-time task requirements of the target prefabricated data center, and generate an initial resource scheduling policy set.

[0027] Furthermore, parse the configuration topology information to obtain the connectivity characteristics of multiple prefabricated units, such as the connection relationship, the number of routes passed, the physical connection distance, the link rate, etc. And access in real time to obtain the real-time resource status information of multiple prefabricated units, such as CPU / GPU utilization rate, memory status, network bandwidth status, storage IO throughput. Among them, the real-time resource status information includes the computing resource status and the environmental resource status, and the environmental resource status is environmental data, such as module temperature, power consumption, etc.

[0028] Such as Figure 2 As shown, the static scheduling decision acquisition module 12 is further used to: access the task management end of the target prefabricated data center and extract the real-time task list. Parse the real-time task list to obtain the resource consumption data of each real-time task, and output it as the real-time resource expectation. Based on the real-time resource expectation and the preset static resource scheduling policy, perform iterative initial scheduling decisions to obtain the initial resource scheduling policy set.

[0029] Initial static scheduling decisions are made according to the real-time task requirements of the target prefabricated data center to generate an initial resource scheduling policy set, including: accessing the task management end of the target prefabricated data center. The task management end is a management interface or management module in the data center responsible for receiving, maintaining, and allocating computing tasks. Extract the real-time task list of the task management end, that is, the list of all tasks ready to be executed or being executed at the current moment, including the basic attributes of these tasks and the corresponding resource requirements. By parsing the real-time task list, obtain the basic attributes of each task, such as the data processing volume, required processing time, etc., and the resource requirements of the tasks, such as memory requirements, network bandwidth requirements, and storage IO throughput. Obtain the resource consumption data of the real-time tasks according to the resource requirements, including: memory consumption, network bandwidth consumption, and storage IO throughput. Output the resource consumption data as the real-time resource expectation. Further, based on the real-time resource expectation and the preset static resource scheduling policy, perform iterative initial scheduling decisions to obtain the initial resource scheduling policy set. The preset static resource scheduling policy is a preset fixed resource scheduling policy, and the preset static resource scheduling policy only allocates the corresponding prefabricated units that meet the requirements according to the resource consumption data of the real-time tasks. For example: for CPU-intensive tasks, allocate resources preferentially in prefabricated computing units with strong computing capabilities. For tasks with high network bandwidth requirements, preferentially select modules with more abundant network links.

[0030] The optimized scheduling policy acquisition module 14 is used to perform iterative decision optimization on the initial resource scheduling policy set according to the connectivity characteristics and the real-time resource status information, and obtain multiple optimization results and store them as an optimized scheduling policy set. The scheduling control module 15 is used to select the top N optimal optimized scheduling policies from the optimized scheduling policy set, output them as a resource scheduling plan set, and perform resource scheduling control on the target prefabricated data center according to the resource scheduling plan set.

[0031] According to the connectivity characteristics and the real-time resource status information, perform iterative decision optimization on the initial resource scheduling policy set, obtain multiple iterative optimization results and store them as an optimized scheduling policy set. Finally, sort the optimized scheduling policy set from large to small according to the output result of the objective function, select the top N optimal optimized scheduling policies, output them as a resource scheduling plan set, and perform resource scheduling control on the target prefabricated data center according to the resource scheduling plan set. This solves the technical problems in the prior art that the resource scheduling of prefabricated data centers has poor scheduling accuracy, it is difficult to obtain the optimal resource scheduling policy in a changing environment, resulting in difficult improvement of resource utilization rate and task execution efficiency. Through multiple rounds of iterative decision optimization and sample point merging and simplification strategies, combined with real-time equivalent connectivity characteristics and link pressure assessment, a set of resource scheduling plans that can be efficiently and dynamically called is finally obtained, thereby improving the resource utilization rate, task execution efficiency, and operation stability of prefabricated data centers.

[0032] The optimization scheduling policy acquisition module 14 is further configured to: according to the configured topology information, obtain the configuration parameter information of multiple prefabricated units in the target prefabricated data center, and define an optimization constraint set according to the configuration parameter information, where the optimization constraint set at least includes unit resource capacity constraint, allocation consistency constraint, cross-module allocation constraint, module temperature constraint, and link load constraint. Based on the optimization constraint set, perform repair crossover on the initial resource scheduling policy set to obtain a repaired resource scheduling policy set. Combine the initial resource scheduling policy set and the repaired resource scheduling policy set to construct an optimization sample set, and based on the optimization sample set and the optimization constraint set, perform iterative decision-making optimization in combination with a preset objective function.

[0033] Perform iterative decision-making optimization on the initial resource scheduling policy set according to the connectivity characteristics and the real-time resource status information, and obtain multiple optimization results stored as an optimization scheduling policy set, including: according to the configured topology information, obtain the configuration parameter information of multiple prefabricated units in the target prefabricated data center, where the configuration parameter information is the initial configuration parameters of the prefabricated units, and define an optimization constraint set according to the configuration parameter information, where the optimization constraint set at least includes unit resource capacity constraint, allocation consistency constraint, cross-module allocation constraint, module temperature constraint, and link load constraint. The unit resource capacity constraint means that the total amount of resources such as CPU, memory, and bandwidth allocated to a unit cannot exceed its maximum carrying capacity. The allocation consistency constraint means that each task must be fully allocated to at least one module, and when it can be allocated within a unit, the selection of multiple allocation units for a single task is not performed. The cross-module allocation constraint is the maximum limit of the number of allocated units, and when the limit is exceeded, the corresponding allocation policy will not be adopted. The module temperature constraint is the highest temperature constraint for task allocation, and tasks will not be allocated for temperatures higher than this constraint. The link load constraint means that data transmission on the network link cannot exceed the designed bandwidth upper limit to maintain normal data exchange speed and latency performance. In the initial resource scheduling policy set, there may be some policies that do not meet the above constraint conditions. Based on the optimization constraint set, perform repair crossover on the initial resource scheduling policy set, that is, perform constraint screening on each scheduling policy in the initial resource scheduling policy set according to the optimization constraint set to obtain a repaired resource scheduling policy set. Combine the initial resource scheduling policy set and the repaired resource scheduling policy set to construct an optimization sample set, and after obtaining the optimization sample set, label the key feature indicators of each policy: such as CPU utilization rate, memory utilization rate, number of units, and current device temperature data. And based on the optimization sample set and the optimization constraint set, perform iterative decision-making optimization in combination with a preset objective function. The preset objective function is: :

[0034] a represents the quality of the optimized samples. The larger a is, the higher the corresponding strategy adaptation degree. b is the current temperature data of the units in the optimized samples. When there are multiple units, b is the average value of the current temperature data of multiple units. y is the resource utilization rate, and the resource utilization rate is the average calculation result of the CPU utilization rate and the memory utilization rate. m and n are positive integers greater than 1, and m is greater than n.

[0035] The optimization scheduling strategy acquisition module 14 is further configured to: in each iterative decision-making optimization, traverse the optimized sample set to extract a common target cluster, where the common target cluster is a subset of samples with a common intersection point in the optimization direction. The subset of samples includes at least N sample points, and N is greater than or equal to 3. Calculate the objective function value of the common intersection point and compare it with the objective function values of the N sample points in the common target cluster. If the objective function value of the common intersection point is the optimal objective function value, then merge the N sample points in the common target cluster to the common intersection point to generate a new sample point.

[0036] Based on the connectivity characteristics and the real-time resource status information, perform iterative decision-making optimization on the initial resource scheduling strategy set, and obtain multiple optimization results stored as the optimization scheduling strategy set. It also includes: in each iterative optimization, search, evaluate, and improve on the sample set to find a better resource scheduling strategy. During this process, due to the large sample size, some sample points may be close in the multi-dimensional resource allocation space and point to similar optimization directions. These similar sample points may represent a relatively stable optimal region or a potential local optimal solution. A sample point refers to the coordinate point of a specific resource allocation scheme in the multi-dimensional variable space (for example, multiple dimensions such as CPU allocation, memory allocation, network bandwidth allocation, etc.). To accelerate convergence and reduce redundant calculations, traverse the optimized sample set to extract common target clusters, classify, merge, and streamline these similar sample points. When extracting the common target cluster, obtain N sample points with a common intersection point, and the common intersection point is the center of the circular area containing the N sample points. The sample subset is composed of N sample points, and the subset of samples includes at least N sample points, and N is greater than or equal to 3. Calculate the objective function value of the common intersection point and compare it with the objective function values of the N sample points in the common target cluster. If the objective function value of the common intersection point is the optimal objective function value, at this time, the common intersection point can represent the N sample points, then merge the N sample points in the common target cluster to the common intersection point to generate a new sample point.

[0037] The optimization scheduling strategy acquisition module 14 is further configured to: if the objective function value of the common intersection point is the optimal objective function value, traverse the common target cluster, calculate the distances between N sample points and the common intersection point, and calculate the variance of the distances between the N sample points and the common intersection point. If the distances between the N sample points and the common intersection point are all less than a preset merging radius, and the variance of the distances between the N sample points and the common intersection point is less than or equal to a preset variance, then use the common intersection point as a new sample point, and calculate the vector sum of the optimization directions and optimization step lengths of the N sample points in the common target cluster as the new optimization direction and optimization step length. Remove the N sample points in the common target cluster, and merge the new sample point into the optimization sample set.

[0038] When the objective function value of the common intersection point is the optimal objective function value, traverse the common target cluster, calculate the distances between N sample points and the common intersection point, and calculate the variance of the distances between the N sample points and the common intersection point. If the distances between the N sample points and the common intersection point are all less than a preset merging radius, and the variance of the distances between the N sample points and the common intersection point is less than or equal to a preset variance, then use the common intersection point as a new sample point, and calculate the vector sum of the optimization directions and optimization step lengths of the N sample points in the common target cluster as the new optimization direction and optimization step length. The preset variance is a variance parameter set in advance. The greater the variance, the more obvious the corresponding distance fluctuation. When it is less than or equal to the preset variance, the higher the concentration of the corresponding sample points. The preset merging radius is the maximum radius for merging multiple sample points with the common intersection point. Remove the N sample points in the common target cluster, and merge the new sample point into the optimization sample set.

[0039] The optimization scheduling strategy acquisition module 14 is further configured to: check the optimization sample set after generating a new sample point, and count the number of sample points. If the number of sample points is less than a first preset sample size, abort the extraction of the common target cluster and the merging of sample points. If the number of sample points is less than a second preset sample size, randomly mutate the new sample point until the number of sample points is greater than or equal to the second preset sample size and less than the first preset sample size, where the second preset sample size is less than the first preset sample size.

[0040] Merge the N sample points in the co-target cluster to the common intersection point to generate new sample points. After that, it further includes: checking the optimized sample set after generating the new sample points and counting the number of sample points. If the number of sample points is less than the first preset sample scale, at this time, the sample quantity can meet the requirements, then abort the extraction of the co-target cluster and the merging of sample points. The first preset sample scale is a preset sample quantity parameter. If the number of sample points is less than the second preset sample scale, randomly mutate the new sample points until the number of sample points is greater than or equal to the second preset sample scale and less than the first preset sample scale, where the second preset sample scale is less than the first preset sample scale.

[0041] The scheduling control module 15 is further configured to: fit the connectivity feature and the real-time resource status information to obtain the real-time equivalent connectivity feature. Based on the real-time equivalent connectivity feature, evaluate the link pressure coefficient of the optimization results in the optimized scheduling policy set and correspondingly generate a link pressure sequence. Select the first N optimization results in the link pressure sequence as the resource scheduling plan set and perform resource scheduling control for the target prefabricated data center according to the resource scheduling plan set.

[0042] Fitting the connectivity feature and the real-time resource status information means using the real-time resource status information to correct the theoretical connectivity feature. Specifically, it can be based on the load of the connection, such as the load rate of the network, etc., and perform weighted adjustment for different connected links. For example, for a link with a higher network load, multiply its connection distance by a coefficient greater than 1 to obtain a real-time equivalent connectivity feature closer to the current real environment. Based on the real-time equivalent connectivity feature, evaluate the link pressure coefficient of the optimization results in the optimized scheduling policy set and correspondingly generate a link pressure sequence. The higher the pressure coefficient, the more backward the ranking. Select the first N optimization results in the link pressure sequence as the resource scheduling plan set and perform resource scheduling control for the target prefabricated data center according to the resource scheduling plan set.

[0043] The scheduling control module 15 is further configured to: serialize the resource scheduling plan set to generate a plan call sequence. Perform resource scheduling control with the first resource scheduling plan at the top of the plan call sequence and monitor the resource usage of the target prefabricated data center in real time. If the real-time monitoring result shows that the resource usage exceeds the limit, then call the next resource scheduling plan of the first resource scheduling plan in the plan call sequence for resource scheduling control until the real-time monitoring result shows that the resource usage does not exceed the limit. Based on the resource scheduling plan called when the resource usage does not exceed the limit, perform feedback adjustment of iterative decision-making optimization.

[0044] Perform resource scheduling control for the target prefabricated data center according to the resource scheduling plan set, including: serializing the resource scheduling plan set according to the initial sorting to generate a plan invocation sequence. Perform resource scheduling control with the first resource scheduling plan at the top of the plan invocation sequence, perform resource scheduling control with the plan ranked first in the plan invocation sequence, and monitor the resource usage of the target prefabricated data center in real time. If the real-time monitoring result shows that the resource usage exceeds the limit, that is, there is a situation where the resource usage limit is exceeded, then call the next resource scheduling plan of the first resource scheduling plan in the plan invocation sequence to perform resource scheduling control until the real-time monitoring result shows that the resource usage does not exceed the limit. Based on the resource scheduling plan called when the resource usage does not exceed the limit, perform feedback adjustment of iterative decision-making optimization.

[0045] In the embodiment of the present application, the configuration topology information of multiple prefabricated units in the target prefabricated data center is obtained by interacting with the target scenario. An initial static scheduling decision is made according to the real-time task requirements of the target prefabricated data center to generate an initial resource scheduling strategy set. Analyze the configuration topology information to obtain the connectivity characteristics of multiple prefabricated units in the target prefabricated data center, and access in real time to obtain the real-time resource status information of multiple prefabricated units, where the real-time resource status information includes the computing resource status and the environmental resource status. According to the connectivity characteristics and the real-time resource status information, perform iterative decision-making optimization of the initial resource scheduling strategy set, obtain multiple optimization results and store them as an optimized scheduling strategy set. Select the top N optimal optimized scheduling strategies from the optimized scheduling strategy set, output them as a resource scheduling plan set, and perform resource scheduling control for the target prefabricated data center according to the resource scheduling plan set. It solves the technical problems in the prior art that the resource scheduling of prefabricated data centers has poor scheduling accuracy and it is difficult to obtain the optimal resource scheduling strategy in a changing environment, resulting in difficult improvement of resource utilization rate and task execution efficiency. Through multiple rounds of iterative decision-making optimization and sample point merging and simplification strategies, combined with real-time equivalent connectivity characteristics and link pressure assessment, a set of resource scheduling plans that can be efficiently and dynamically called is finally obtained, thereby improving the resource utilization rate, task execution efficiency and operation stability of the prefabricated data center.

[0046] The above specific embodiments do not constitute a limitation on the protection scope of this application. Those skilled in the art should understand that various modifications, combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principle of this application shall be included within the protection scope of this application. In some cases, the actions or steps recorded in this application can be executed in a sequence different from that in the embodiments and still achieve the desired results. Additionally, the processes depicted in the drawings do not necessarily require the specific order or consecutive order shown to achieve the desired results. In certain embodiments, multitasking and parallel processing are also possible or may be advantageous.

Claims

1. A resource scheduling system for a prefabricated data center, characterized in that: The system comprises: An information collection module, used for interacting with a target scenario and obtaining configuration topology information of multiple prefabricated units in a target prefabricated data center; A static scheduling decision acquisition module is used to make initial static scheduling decisions based on the real-time task requirements of the target prefabricated data center and generate an initial resource scheduling strategy set; A resource status acquisition module, used to parse the configuration topology information, obtain connectivity characteristics of the plurality of prefabricated units of the target prefabricated data center, and access and obtain the plurality of prefabricated units in real time to obtain real-time resource status information, wherein the real-time resource status information includes computing resource status and environmental resource status; An optimized scheduling strategy acquisition module is used to perform iterative decision optimization of the initial resource scheduling strategy set according to the connectivity characteristics and the real-time resource status information, and obtain multiple optimization results and store them as an optimized scheduling strategy set; A scheduling control module, used to select the best top N optimization scheduling strategies from the optimization scheduling strategy set, output them as a resource scheduling plan set, and perform resource scheduling control of the target prefabricated data center according to the resource scheduling plan set; The optimization scheduling strategy acquisition module is also used for: According to the configuration topology information, obtain configuration parameter information of a plurality of the prefabricated units in the target prefabricated data center, and define an optimization constraint set according to the configuration parameter information, wherein the optimization constraint set includes at least a unit resource capacity constraint, an allocation consistency constraint, a cross-module allocation constraint, a module temperature constraint, and a link load constraint; Based on the optimization constraint set, performing repair cross-talk on the initial resource scheduling strategy set to obtain a repair resource scheduling strategy set; Combining the initial resource scheduling strategy set with the repair resource scheduling strategy set to construct an optimization sample set, and performing iterative decision optimization based on the optimization sample set and the optimization constraint set in combination with a preset objective function; Wherein, the dispatch control module is also used for: Serializing the resource scheduling plan set to generate a plan calling sequence; Using the first resource scheduling plan at the top of the plan call sequence to perform resource scheduling control, and monitor the resource usage of the target prefabricated data center in real time; If the real-time monitoring result shows that the resource usage exceeds the limit, calling the next resource scheduling plan of the first resource scheduling plan in the plan calling sequence to perform resource scheduling control until the real-time monitoring result shows that the resource usage does not exceed the limit; Based on the resource scheduling plan called when the resource usage does not exceed the limit, feedback adjustment of iterative decision-making optimization is performed; The dispatch control module is also used for: Fitting the connectivity feature with the real-time resource status information to obtain a real-time equivalent connectivity feature; Based on the real-time equivalent connectivity characteristics, evaluating the link pressure coefficient of the optimization result of the optimization scheduling strategy concentration, and generating a link pressure sequence accordingly; The first N optimization results in the link pressure sequence are selected as a resource scheduling plan set, and resource scheduling control of the target prefabricated data center is performed according to the resource scheduling plan set.

2. A resource scheduling system for a prefabricated data center as claimed in claim 1, characterized in that: The static scheduling decision acquisition module is also used for: Access the task management terminal of the target prefabricated data center and extract the real-time task list; Parsing the real-time task list to obtain resource consumption data of each real-time task, and outputting the data as real-time resource expectation; An iterative initial scheduling decision is performed based on the real-time resource expectation and a preset static resource scheduling strategy to obtain the initial resource scheduling strategy set.

3. A resource scheduling system for a prefabricated data center as claimed in claim 2, characterized in that: The optimized scheduling strategy acquisition module is also used for: In each iterative decision optimization, the optimization sample set is traversed to extract a common target cluster, wherein the common target cluster is a sample subset having a common intersection in the optimization direction, wherein the sample subset includes at least N sample points, and N is greater than or equal to 3; Calculate the objective function value of the common intersection point and compare it with the objective function values ​​of the N sample points in the common target cluster; If the objective function value of the common intersection is the optimal objective function value, the N sample points in the common target cluster are merged into the common intersection to generate a new sample point.

4. A resource scheduling system for a prefabricated data center as claimed in claim 3, characterized in that: The optimized scheduling strategy acquisition module is also used for: If the objective function value of the common intersection is the optimal objective function value, traverse the common target cluster, calculate the distances between N sample points and the common intersection, and calculate the variance of the distances between the N sample points and the common intersection; If the distances between the N sample points and the common intersection are all less than the preset merging radius, and the variance of the distances between the N sample points and the common intersection is less than or equal to the preset variance, the common intersection is taken as a new sample point, and the vector sum of the optimization direction and the optimization step of the N sample points in the common target cluster is calculated as the new optimization direction and optimization step; The N sample points in the common target cluster are removed, and the new sample points are merged into the optimized sample set.

5. The resource scheduling system for a prefabricated data center according to claim 3, characterized in that: The optimized scheduling strategy acquisition module is also used for: Checking the optimized sample set after generating new sample points, and counting the number of sample points; If the number of sample points is less than the first preset sample size, terminating the extraction of the common target cluster and the merging of the sample points; If the number of sample points is less than the second preset sample size, random mutation is performed on the new sample points until the number of sample points is greater than or equal to the second preset sample size and less than the first preset sample size, wherein the second preset sample size is smaller than the first preset sample size.

Images