Zero-carbon data center resource scheduling method and system using unstable clean energy

Abstract

The invention relates to the technical field of clean energy resource scheduling, and provides a zero-carbon data center resource scheduling method and system using unstable clean energy. A zero-carbon data center resource scheduling method using unstable clean energy comprises the following steps: respectively converting available computing power resources and businesses into a computing power resource large rectangle and a business small rectangle, and converting a matching scheduling problem of the computing power resources to the businesses into a packing problem model of the business small rectangle in the computing power resource large rectangle; the method comprises the following steps of: mapping all resource available restrictions and business self-processing requirements in a zero-carbon data center scene into a restriction condition of whether small business rectangles are allowed to be placed, and calculating clean energy quantity required by the business, so as to distribute the business to a time point when a set number of computing power resources support the business to be processed and start processing; and solving the boxing problem model to obtain a distribution scheme. The utilization rate of clean energy is increased, and dependence on thermal power is reduced.

Description

Technical Field

[0001] The present invention relates to the technical field of clean energy resource scheduling, and in particular to a zero-carbon data center resource scheduling method and system using unstable clean energy. Background Art

[0002] The statements in this section merely provide background information related to the present invention and do not necessarily constitute prior art.

[0003] Since wind power and photovoltaic renewable energy generation are unstable due to meteorological fluctuations, they are difficult to be integrated into the large power grid. They are mainly decentralized and self-generated in a small area, so they are called microgrids or microgrids. Therefore, in the case that large-scale, low-cost, long-term energy storage technology is difficult to break through in the short term, it is still difficult for clustered data centers to adopt a large proportion of wind power and photovoltaic power supply.

[0004] Existing data center resource scheduling strategies usually only consider resource availability at the current point in time, and lack the prediction and utilization of future clean energy supply. This limitation has led to low clean energy utilization, and data centers still rely mainly on thermal power supply, resulting in high carbon emissions. At the same time, with the rapid development of fields such as artificial intelligence and big data analysis, a large number of warm and cold businesses have emerged in the storage and computing business of data centers. These tasks usually have low requirements for real-time computing and can tolerate longer waiting times. However, existing resource scheduling strategies fail to make full use of this feature to match these tasks that can be delayed with clean energy supply forecasts, resulting in inefficient resource utilization and energy waste.

[0005] In addition, current data centers lack refined optimization methods for task allocation and resource utilization. Failure to efficiently match pending computing tasks with available computing resources has led to idle computing resources and energy waste. In particular, during the time window when clean energy supply is sufficient, failure to fully utilize renewable energy to complete computing tasks has missed the opportunity to reduce carbon emissions. Summary of the invention

[0006] In order to solve the technical problems existing in the above-mentioned background technology, the present invention provides a zero-carbon data center resource scheduling method and system using unstable clean energy. The present invention comprehensively considers the unstable supply of clean energy when scheduling resources. It can complete the computing business processing of the zero-carbon data center quickly, efficiently and economically without using thermal power and relying entirely on unstable clean electric power, while improving the wind power and photovoltaic absorption rate and reducing the wind and solar power abandonment rate.

[0007] In order to achieve the above object, the present invention adopts the following technical solution: A first aspect of the present invention provides a zero-carbon data center resource scheduling method using unstable clean energy.

[0008] A zero-carbon data center resource scheduling method using unstable clean energy, comprising: Access available computing resources for business and zero-carbon data centers; The available computing resources and businesses are converted into a large rectangle of computing resources and a small rectangle of business respectively, and the matching and scheduling problem of computing resources for businesses is converted into a packing problem model of small rectangles of business in a large rectangle of computing resources. All available resource restrictions and the business's own processing requirements in the zero-carbon data center scenario are mapped into restrictions on whether small rectangles of business are allowed to be placed, and the amount of clean energy required for the business is calculated to allocate the business to a set number of computing resources to support the time point when the business is completed. Solve the packing problem model to obtain an allocation solution.

[0009] Furthermore, the width of the large rectangle of computing resources represents the total number of available servers in a certain scheduling period, and the length represents the time axis corresponding to the scheduling period.

[0010] Furthermore, the width of the small rectangle of the service represents the number of servers to which the service will be allocated, and the length represents the calculation time required after the service is allocated a set number of servers.

[0011] Furthermore, the packing problem model of the small business rectangle in the large rectangle of computing power resources is as follows: the width of the large rectangle of computing power resources is used as the vertical coordinate and the length as the horizontal coordinate to establish a two-dimensional coordinate system containing only the first quadrant; the small business rectangle is placed in the large rectangle of computing power resources, and a Boolean variable is used to indicate whether the business is placed, and the start time of the business processing and the first positive sequence value of the server number used by the business are determined. Combined with the size of the small business rectangle, the end time of the business processing and the required number of servers are determined; in the two-dimensional coordinate system, the server number set and the time slot set are used as coordinate units on the coordinate axis to locate the position of the small business rectangle.

[0012] Furthermore, the restrictions on whether to allow the small business rectangle to be placed include: the small business rectangles should not overlap with each other, the small business rectangles should not exceed the scope of the large rectangle of computing resources, and the clean energy supply should meet the business processing needs.

[0013] Furthermore, the small business rectangles should not overlap with each other, that is, different businesses should not repeatedly occupy server resources at the same time; the small business rectangles should not exceed the scope of the large rectangle of computing resources, that is, the computing resources used for business processing do not exceed the number of servers that can be scheduled within the scheduling cycle, and the business processing completion time is no later than the server schedulable time; the clean energy supply should meet the business processing needs, that is, the clean energy of the zero-carbon data center at each time is allocated to different businesses, ensuring that the clean energy required by the business is available at all time points to support the current business processing.

[0014] Furthermore, the packing problem model is solved to obtain an allocation plan; the method includes: taking the maximum number of small business rectangles packed into a large rectangle of computing resources as the goal, or taking the maximum area occupied by a large rectangle of computing resources as the goal, using a solver to solve the packing problem model to obtain an allocation plan.

[0015] Furthermore, the business processing requirements are for processing warm and cold data, that is, the business does not require a timely response, and is transmitted to a remote data center in the form of offline data for processing.

[0016] A second aspect of the present invention provides a zero-carbon data center resource scheduling system using unstable clean energy.

[0017] A zero-carbon data center resource scheduling system using unstable clean energy, comprising: The infrastructure layer includes clean energy and computing resources. The clean energy is used for resource scheduling strategies to match computing resources to business processing; the computing resources are the core hardware resources for processing business; The management and control layer is used to obtain the computing resource status and clean energy supply in the infrastructure layer, monitor the change data, and generate the constraints of the large rectangle of computing resources in the bin packing problem model; The scheduling and orchestration layer is used to obtain the available computing resources of the business and the zero-carbon data center; convert the available computing resources and the business into a large rectangle of computing resources and a small rectangle of business respectively, and convert the matching scheduling problem of available computing resources for the business into a packing problem model of the small rectangle of business in the large rectangle of computing resources; map all resource availability restrictions and the business's own processing requirements in the zero-carbon data center scenario into restrictions on whether the small rectangle of business is allowed to be placed, calculate the amount of clean energy required for the business, and allocate the business to a set number of computing resources to support the time point when the business is completed; solve the packing problem model to obtain an allocation plan.

[0018] Furthermore, the width of the large rectangle of computing resources represents the total number of available servers within a scheduling cycle, and the length represents the timeline corresponding to the scheduling cycle; the width of the small rectangle of the business represents the number of servers to which the business will be allocated, and the length represents the computing time required after the business is allocated a set number of servers.

[0019] Compared with the prior art, the present invention has the following beneficial effects: The present invention improves the utilization rate of clean energy, reduces dependence on thermal power by allocating services to time windows with sufficient clean energy, achieves zero-carbon operation of the data center, and reduces carbon emissions.

[0020] The present invention simplifies the complexity of resource scheduling. It converts the resource allocation problem into a two-dimensional non-overlapping packing problem of a small rectangle of "server × time" corresponding to the business in a large rectangle of "server × time" corresponding to the computing power resource through the rectangular resource expression of "server × time". The problem is solved by a solver, thereby improving the scheduling efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] The accompanying drawings in the specification, which constitute a part of the present invention, are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute improper limitations on the present invention.

[0022] Figure 1 is a flow chart of a zero-carbon data center resource scheduling method using unstable clean energy shown in the present invention; Figure 2 It is a flowchart of the present invention showing the completion of processing by predicting the spatiotemporal distribution of unstable clean energy and scheduling resource supply services in a targeted manner; Figure 3 It is a schematic diagram of mapping the packing problem to the zero-carbon data center resource scheduling problem shown in the present invention; Figure 4 is a zero-carbon data center architecture diagram shown in the present invention; In the figure, 01, the first small rectangle corresponding to the business, 02, the second small rectangle corresponding to the business, 03, the third small rectangle corresponding to the business, 04, the fourth small rectangle corresponding to the business, 05, the fifth small rectangle corresponding to the business, 06, the sixth small rectangle corresponding to the business, 14, the predicted clean energy power generation power curve at each time point, 15, the gray coverage area. DETAILED DESCRIPTION

[0023] The present invention will be further described below in conjunction with the accompanying drawings and embodiments.

[0024] It should be noted that the following detailed descriptions are all illustrative and intended to provide further explanation of the present invention. Unless otherwise specified, all technical and scientific terms used herein have the same meanings as those commonly understood by those skilled in the art to which the present invention belongs.

[0025] It should be noted that the terms used herein are only for describing specific embodiments and are not intended to limit exemplary embodiments according to the present invention. As used herein, unless the context clearly indicates otherwise, the singular form is also intended to include the plural form. In addition, it should be understood that when the terms "comprising" and / or "including" are used in this specification, it indicates the presence of features, steps, operations, devices, components and / or combinations thereof.

[0026] It should be noted that the flowcharts and block diagrams in the accompanying drawings illustrate the possible implementation architecture, functions and operations of the methods and systems according to various embodiments of the present disclosure. It should be noted that each box in the flowchart or block diagram can represent a module, a program segment, or a part of a code, and the module, program segment, or a part of a code may include one or more executable instructions for implementing the logical functions specified in each embodiment. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in an order different from that marked in the accompanying drawings. For example, two boxes represented in succession can actually be executed substantially in parallel, or they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the flowchart and / or block diagram, and the combination of boxes in the flowchart and / or block diagram can be implemented using a dedicated hardware-based system that performs a specified function or operation, or can be implemented using a combination of dedicated hardware and computer instructions.

[0027] Embodiment 1 like Figure 1 As shown, this embodiment provides a zero-carbon data center resource scheduling method using unstable clean energy, including: Access available computing resources for business and zero-carbon data centers; The available computing resources and businesses are converted into a large rectangle of computing resources and a small rectangle of business respectively, and the matching and scheduling problem of computing resources for businesses is converted into a packing problem model of small rectangles of business in a large rectangle of computing resources. All available resource restrictions and the business's own processing requirements in the zero-carbon data center scenario are mapped into restrictions on whether small rectangles of business are allowed to be placed, and the amount of clean energy required for the business is calculated to allocate the business to a set number of computing resources to support the time point when the business is completed. Solve the packing problem model to obtain an allocation solution.

[0028] The present invention establishes a resource scheduling strategy for a zero-carbon data center and builds a binning problem resource scheduling model to allocate resources for services that process warm and cold data. First, a "server × time" rectangular resource expression method is proposed, so that on the one hand, the number of available servers in the data center and the number of servers required by the service are set as the server dimension, and on the other hand, a server resource scheduling cycle and the processing time required by the service are set as the time dimension, thereby converting computing power resources and services into a large rectangle of computing power resources and a small rectangle of services. Second, based on the basic modeling idea of ​​the binning problem and combined with the above resource conversion results, the matching scheduling problem of computing power resources to business needs is converted into a two-dimensional non-overlapping binning problem of small business rectangles in a large rectangle of computing power resources, and the matching of data processing services to computing power resources is mapped by placing small business rectangles in the large rectangle of computing power resources, and the resource availability restrictions and business processing requirements in the zero-carbon data center scenario are mapped by the restriction conditions of whether to allow placement, thereby simplifying resource scheduling. Third, on the basis of the above, the two major attributes of the number of servers required by the service and the required processing time are used as the intermediate link, and the restriction relationship between clean energy and service allocation is established by calculating the power consumption required for processing the service, and then the service is allocated to the time point when clean energy is sufficient to support its completion of processing. Finally, the solver solves the packing problem of maximizing the number of small business rectangles placed in the large rectangle of computing resources or maximizing the area occupied by the large rectangle of computing resources, and obtains an efficient allocation solution under this method. The present invention optimizes the utilization efficiency of clean energy by matching computing tasks with the volatility of clean energy supply, helps to improve the absorption rate of clean energy, and reduces the phenomenon of wind and solar power abandonment.

[0029] The packing problem, the classic packing problem, requires putting a certain number of items into some boxes so that the sum of the sizes of the items in each box does not exceed the box capacity and the number of boxes used is minimized. The present invention refers to a business obtaining a part of the available resources for calculation and processing by packing the items into the available resource box and occupying a part of the space.

[0030] The business involving processing warm and cold data refers to the business that does not require a timely response and is often transmitted to a remote data center for processing in the form of offline data, such as the training of large artificial intelligence models.

[0031] For warm and cold data services that can tolerate delays, such as large-scale artificial intelligence model training, the present invention optimizes task scheduling and resource allocation to perform computing business processing in time windows with sufficient clean energy. This not only meets the timely response to business needs, but also improves the energy efficiency ratio (computing performance provided per unit of energy consumption), effectively reduces energy waste, and improves the utilization rate of clean energy.

[0032] The "server × time" rectangular resource expression method refers to the use of rectangles to represent data processing services, available computing resources and clean energy. The width of the rectangle (server): represents the number of servers required for data processing services or the number of available servers in the data center. Using "server" as the dimension of width allows different services and resources to be compared and scheduled under the same dimension. The length of the rectangle (time): represents the computing time required for data processing services or a scheduling cycle of a data center. Using "time" as the length dimension ensures that all services and resources are consistent in the time dimension. The area of ​​the rectangle (clean energy consumption): The area is the product of the length and width of the rectangle. Therefore, when the server power is known, the product of the number of servers (width) and the working time (length) can be used to represent the amount of electricity consumed, thereby further converting the area of ​​the rectangle into the amount of clean energy consumed. Therefore, the area of ​​the rectangle can quantify the amount of clean energy required for data processing services.

[0033] The width of the large rectangle of computing power resources represents the total number of servers available in a certain scheduling period. In specific scheduling, the numbers corresponding to different servers can be used to form a set = , the length of the large rectangle represents the time axis of the scheduling cycle. In specific scheduling, the time axis can be set to different lengths of seconds or minutes as the minimum unit according to your own needs. Divided into different numbers of independent time slots, forming a horizontal axis coordinate represented by length, such as the set is = .

[0034] The width of the small rectangle of the business indicates the number of servers to which the business will be assigned. In specific scheduling, the business can be collected and aggregated to form a set = The length of the small rectangle represents the computing time required after the service is assigned a set number of servers. .

[0035] The small business rectangle is placed in the large computing resource rectangle, and the width of the large computing resource rectangle can be used as the vertical coordinate and the length as the horizontal coordinate to form a two-dimensional coordinate system that only includes the first quadrant. In this coordinate system, the server number set and the time slot set are used as coordinate units on the coordinate axis, so that the first quadrant is divided into the total number of available independent time slots. ×Total number of available servers The minimum available resource cells to be allocated, the coordinates of these cells are represented as The small business rectangle is placed into the large computing resource rectangle, that is, the appropriate length, width and coordinates of the lower left corner of the small business rectangle are found. To further describe this position, we use the Boolean variable Indicates business Is it at the lower left corner? Place it at the place where it is placed, it is 1, otherwise it is 0, so as to determine the start time for a certain business process The minimum server number used by this service Determine the end time of business processing based on the processing time represented by the length and width of the small business rectangle and the number of servers required. and the number of servers required .

[0036] The restriction conditions for whether to allow the inclusion include: small business rectangles should not overlap with each other, small business rectangles should not exceed the scope of large resource rectangles, and clean energy supply should meet business processing needs; The small business rectangles should not overlap each other, which means that different businesses should not repeatedly occupy server resources at the same time, as shown in the following formula: Represented as business i Is it in the time slot? k Processed inside:

[0037] The small business rectangle should not exceed the scope of the large resource rectangle, which means that the computing resources used in business processing do not exceed the number of servers that can be scheduled within the scheduling cycle, and the processing completion time of the business is not later than the server's schedulable time, as shown in the following formula:

[0038]

[0039] The restriction relationship between clean energy and business allocation refers to the server power consumption calculated according to the component power table. and the number of servers required by the business itself Combined with the power required by the business, according to the clean energy spatiotemporal variation curve, the clean energy of the zero-carbon data center at each moment is reasonably allocated to different businesses to ensure that there is sufficient clean energy at all times. Support current business processing, thereby avoiding the use of thermal power from the public power grid, as shown in the following formula:

[0040]

[0041] The efficient allocation scheme under the method is obtained by simply using mathematical programming software such as Gurobi, according to the packing problem scenario and the constraint condition objective function, to obtain the coordinates of the lower left corner of the small business rectangle ( ， ) and length He Kuan .

[0042] The objective function refers to the method under which a specific indicator in the packing scenario can be selected to represent the resource utilization efficiency, thereby improving it in a targeted manner to achieve efficient allocation; Optionally, the resource utilization efficiency is expressed by the number of small rectangles packed, as shown in the following formula:

[0043] Optionally, the resource utilization efficiency is expressed by the occupied large rectangular area, as shown in the following formula:

[0044] The present invention proposes a zero-carbon data center resource scheduling strategy based on the idea of ​​bin packing problem. With the help of the "server × time" rectangular resource expression, the computing resource scheduling problem is abstracted into a two-dimensional non-overlapping bin packing problem, which realizes efficient resource allocation for business processing involving warm and cold data, improves resource scheduling efficiency, and promotes green development.

[0045] The present invention predicts the temporal and spatial distribution of clean energy, allocates data processing services to time periods with sufficient clean energy supply, makes full use of renewable energy sources such as wind power and photovoltaics, reduces dependence on thermal power, reduces carbon emissions, and improves the utilization rate of clean energy.

[0046] The present invention transforms the resource scheduling problem into a packing problem by adopting the "server × time" rectangular resource expression, thus simplifying the complexity of resource scheduling. Optimization using the solver can achieve efficient matching of services and resources, maximize the utilization of computing resources, and reduce idle computing resources and energy waste.

[0047] Figure 2This is a flowchart of the present invention that predicts the spatiotemporal distribution of clean energy and dispatches resource supply services in a targeted manner to complete the processing. In this embodiment, the application of the present invention to process a data processing service specifically includes the following process: In region A, where clean energy is scarce or a large amount of warm and cold data is generated, after the data processing business is generated, it enters the packing decision module. In this module, the existing available resources are allocated according to business needs, and the business scheduling plan is determined.

[0048] The available resources come from the zero-carbon data center in Region B, which relies on a continuous supply of clean energy to power its internal servers. The computing resources of the servers are allocated to specific data processing businesses according to the allocation results of the packing decision. After the resource allocation is completed, the business occupies the computing resources of the server and consumes clean energy to perform business processing.

[0049] During business processing, it is necessary to consider whether there is sufficient clean energy during this period to ensure that data processing business only consumes clean energy.

[0050] After the business processing is completed, the computing resources occupied by the business are released so that they can be matched and used by other businesses again.

[0051] Figure 3 This is a diagram of the mapping of the packing problem to the zero-carbon data center resource scheduling problem. Figure 2 The example description of how a business allocates existing available resources based on the idea of ​​packing decision is given in the figure.

[0052] (1) Coordinate system and axis Based on the idea of ​​constructing a bin packing problem model in the form of "server × time" rectangular resource expression, the computing power resources available in a certain period of time in the future can be abstracted into a two-dimensional rectangle of the number of available servers and time, and further expressed as a coordinate system with only the first quadrant, as shown in the following example: The left vertical axis represents the server number, and there are 5 servers in total (A, B, C, D, E); The horizontal axis represents the schedulable time, from 0 to 9 hours.

[0053] The right vertical axis represents the power (KW) of clean energy, and the predicted clean energy power generation curve 14 at each time point above it is the predicted clean energy power generation at each time point; the gray covered area 15 is the clean energy power generation.

[0054] (2) Clean energy constraints According to the idea of ​​building a packing problem model, the “server × time” rectangular resource expression is used to represent data services as small rectangles. The horizontal length represents the required processing time, and the vertical width represents the number of servers required. The area of ​​the small rectangle is proportional to the power consumption of the service.

[0055] By adjusting the unit ratio of the coordinate system, the gray area (clean power) in each time unit is made to have the same order of magnitude as the small rectangular area (business power consumption), so that when allocating resources, it is possible to compare whether the business consumption is less than the clean power in the current time period.

[0056] The sum of the business power consumption areas in each time period should be smaller than the gray coverage area, indicating that the total consumption in this time period does not exceed the available clean energy, ensuring low-carbon operation; when the position and size of the small rectangle are determined, the resource matching of the business can be determined.

[0057] (3) Specific business placement examples According to the restriction conditions of whether to allow the inclusion and the restriction relationship between clean energy and business allocation, the position, length and width of the small business rectangle in the coordinate system are obtained.

[0058] The processing time of the first small rectangle 01 corresponding to the business is outside the available time, and occupies the B server from 1:00 to 4:00, overlapping with the business of the second small rectangle 02 corresponding to the business, and the placement is wrong; The second small rectangle 02 corresponding to the business shares the server numbered B with the first small rectangle 01 corresponding to the business from 1:00 to 4:00, which is wrongly placed; The sixth server does not exist in the fifth small rectangle 05 corresponding to the business, and the total area of ​​the business small rectangles from 0:00 to 1:00 is larger than the gray coverage area representing clean energy, which is wrongly placed; The third small rectangle 03 corresponding to the business uses the server numbered D to process from 4:00 to 9:00, and the clean energy during this time period meets the business processing requirements, so the placement is correct; The fourth small rectangle 04 corresponding to the business uses servers numbered A, B, and C to process from 5:00 to 9:00, and the clean energy during this time period meets the business processing requirements, and the placement is correct; The sixth small rectangle 06 corresponding to the business uses the server numbered E to process from 6:00 to 8:00, and the clean energy meets the business processing requirements during this time period, and the placement is correct.

[0059] The present invention abstracts complex scheduling problems into packing problems, uses the "server × time" rectangular resource expression method to simplify the scheduling model, and automatically generates optimized resource allocation plans through mathematical programming tools, thereby reducing the need for manual configuration and adjustment, reducing the workload of resource schedulers, simplifying the resource scheduling process, improving work efficiency, and realizing intelligent and automated resource scheduling.

[0060] The present invention predicts and monitors the spatiotemporal distribution of unstable clean energy, and prioritizes scheduling computing tasks to time periods with sufficient clean energy supply; maximizes clean energy utilization and significantly reduces carbon emissions by optimizing resource allocation and scheduling strategies; tracks and manages energy usage in real time to reduce energy waste; optimizes the geographical layout of data centers, fully utilizes clean energy resources in certain areas, and promotes the green transformation of the data center industry; and promotes the sustainable development of the digital economy by reducing carbon emissions and energy consumption in data centers, which has positive social and economic value.

[0061] Embodiment 2 This embodiment provides a zero-carbon data center resource scheduling system using unstable clean energy, including: The infrastructure layer includes clean energy and computing resources. The clean energy is used for resource scheduling strategies to match computing resources to business processing. The computing resources refer to the core hardware resources used by the data center to process business, including but not limited to CPU, GPU, memory, storage and network resources. Clean energy supply supports the operation of these computing resources, reduces dependence on traditional thermal power, and improves the zero-carbon operation and maintenance capabilities of data centers. The management and control layer is used to obtain the computing resource status and clean energy supply in the infrastructure layer, and monitor the change data to provide support for the constraints of the large rectangle of computing resources in the bin packing problem model; The scheduling and orchestration layer is used to obtain computing resource information from the management and control layer, and in combination with business needs, convert computing resources and business needs into a large rectangle of computing resources and a small rectangle of business needs. By converting the matching scheduling problem of computing resources and business needs into a packing problem model of small business rectangles in a large rectangle of computing resources, the scheduling and orchestration layer can optimize resource allocation, map the available resource restrictions and business processing requirements in the zero-carbon data center scenario into constraints on whether small business rectangles are allowed to be placed, calculate the amount of clean energy required for the business, and then decide which computing resources should be allocated to the business to ensure that the business starts processing and completes business needs at a specific time point; finally, the scheduling and orchestration layer solves the packing problem model to obtain the final resource allocation plan.

[0062] Figure 4 This is a zero-carbon data center architecture diagram that the present invention relies on for specific implementation. Figure 2 Examples of the functional features that a zero-carbon data center in region B should have.

[0063] In order to support the normal operation of the above resource scheduling strategy, the zero-carbon data center should include at least three layers of structure: infrastructure layer, management and control layer, and scheduling and orchestration layer.

[0064] The infrastructure layer includes two main resources: computing power resources and clean energy; Computing resources: including core hardware for processing data services, including GPUs, CPUs, etc., for processing general services and intelligent computing services, and providing high-performance servers for supercomputing services, as well as memory and hard disks required for data processing and communication networks required for data transmission. The above computing resources are summarized in the form of servers for resource scheduling strategies to match to services to complete processing; Clean energy: provided by the clean energy power generation microgrid, which converts the abundant light and wind energy in area B into electrical energy to generate clean energy; chemical energy storage devices are used for the storage and dispatch of clean energy. Meteorological observation stations and satellite radars do not need to be self-provided, but can obtain light and wind energy forecast data from existing related facilities such as the Meteorological Bureau through interfaces, providing a basis for the temporal and spatial distribution of clean energy. Clean energy is used for resource scheduling strategies to match computing resources for business processing.

[0065] The management and control layer aggregates information from the infrastructure layer to assist resource scheduling strategies. Its functions mainly include computing resource pooling, clean energy prediction, and resource perception.

[0066] Computing resource pooling: Integrate the underlying computing resources into a resource pool to facilitate dynamic resource allocation for different businesses. This function supports the position and size adjustment of small business rectangles in the box packing model.

[0067] Clean energy forecast: forecast the supply of clean energy based on meteorological data and provide power change trends for upper-level dispatch.

[0068] Resource perception function: Computing resource perception and clean energy perception monitor the server status and clean energy supply respectively, and feed back the change data to the scheduling layer in real time to generate resource rectangles and clean energy constraints in the packing model.

[0069] The scheduling and orchestration layer is the core layer of the zero-carbon data center. On the one hand, it collects data processing business information, and on the other hand, it solves the packing problem in the resource scheduling strategy.

[0070] Business aggregation and intelligent decision-making: Collect the business information to be processed, abstract each business into a small rectangle based on the packing model strategy, and find a suitable placement based on resource and power constraints.

[0071] Resource coordination and matching: Based on the lower-level computing power energy perception change data, clean energy perception change data and the summarized business information, through calculations based on the two-dimensional packing model, the appropriate computing power resources and time windows are matched for the business, ensuring that the small rectangles do not overlap when placed and meet the constraints of clean energy, achieving resource coordination and releasing occupied resources after completing business processing.

[0072] The present invention provides a detailed data center architecture, intuitively displays the overall structure and key components of a zero-carbon data center, supports the implementation and optimization of resource scheduling strategies, facilitates understanding of the collaborative working relationship of various parts of the system, and improves the visualization level of design and management.

[0073] The present invention establishes a resource scheduling strategy for a zero-carbon data center, which includes a resource scheduling model, a data center architecture design, and server time management, effectively allocating resources for data processing services involving warm and cold data, and achieving efficient matching of resources with business needs.

[0074] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A zero-carbon data center resource scheduling method using unstable clean energy, characterized in that: include: Access available computing resources for business and zero-carbon data centers; The available computing resources and businesses are converted into a large rectangle of computing resources and a small rectangle of business respectively, and the matching and scheduling problem of computing resources for businesses is converted into a packing problem model of small rectangles of business in a large rectangle of computing resources. All available resource restrictions and the business's own processing requirements in the zero-carbon data center scenario are mapped into restrictions on whether small rectangles of business are allowed to be placed, and the amount of clean energy required for the business is calculated to allocate the business to a set number of computing resources to support the time point when the business is completed. Solve the packing problem model to obtain an allocation solution.

2. The zero-carbon data center resource scheduling method using unstable clean energy according to claim 1, characterized in that: The width of the large rectangle of computing power resources represents the total number of available servers in a certain scheduling period, and the length represents the time axis corresponding to the scheduling period.

3. The zero-carbon data center resource scheduling method using unstable clean energy according to claim 1, characterized in that: The width of the small rectangle of the service indicates the number of servers to which the service will be allocated, and the length indicates the calculation time required after the service is allocated a set number of servers.

4. The zero-carbon data center resource scheduling method using unstable clean energy according to claim 1, characterized in that: The packing problem model of the small business rectangle in the large rectangle of computing power resources is as follows: the width of the large rectangle of computing power resources is used as the vertical coordinate and the length as the horizontal coordinate to establish a two-dimensional coordinate system containing only the first quadrant; the small business rectangle is placed in the large rectangle of computing power resources, a Boolean variable is used to indicate whether the business is placed, the start time of the business processing and the first positive sequence value of the server number used by the business are determined, and the end time of the business processing and the required number of servers are determined in combination with the size of the small business rectangle; in the two-dimensional coordinate system, the server number set and the time slot set are used as coordinate units on the coordinate axis to locate the position of the small business rectangle.

5. The zero-carbon data center resource scheduling method using unstable clean energy according to claim 1, characterized in that: The restrictive conditions for whether to allow the small business rectangle to be placed include: the small business rectangles should not overlap with each other, the small business rectangles should not exceed the scope of the large rectangle of computing resources, and the clean energy supply should meet the business processing needs.

6. The zero-carbon data center resource scheduling method using unstable clean energy according to claim 5, characterized in that: The small business rectangles should not overlap with each other, meaning that different businesses should not occupy server resources repeatedly at the same time; the small business rectangles should not exceed the scope of the large rectangle of computing resources, meaning that the computing resources used for business processing do not exceed the number of servers that can be scheduled within the scheduling cycle, and the business processing completion time is no later than the server's schedulable time; the clean energy supply should meet business processing needs, meaning that the clean energy of the zero-carbon data center at each moment is allocated to different businesses to ensure that the clean energy required by the business is available at all time points to support current business processing.

7. The zero-carbon data center resource scheduling method using unstable clean energy according to claim 1, characterized in that: The packing problem model is solved to obtain an allocation plan; the method includes: taking the maximum number of small business rectangles packed into a large rectangle of computing resources as the goal, or taking the maximum area occupied by a large rectangle of computing resources as the goal, using a solver to solve the packing problem model to obtain an allocation plan.

8. The zero-carbon data center resource scheduling method using unstable clean energy according to claim 1, characterized in that: The business processing requirements are for processing warm and cold data, which are transmitted to a remote data center in the form of offline data for processing.

9. A zero-carbon data center resource scheduling system using unstable clean energy, characterized in that: include: The infrastructure layer includes clean energy and computing resources. The clean energy is used for resource scheduling strategies to match computing resources to business processing; the computing resources are the core hardware resources for processing business; The management and control layer is used to obtain the computing resource status and clean energy supply in the infrastructure layer, monitor the change data, and generate the constraints of the large rectangle of computing resources in the bin packing problem model; The scheduling and orchestration layer is used to obtain the available computing resources of the business and the zero-carbon data center; convert the available computing resources and the business into a large rectangle of computing resources and a small rectangle of business respectively, and convert the matching scheduling problem of available computing resources for the business into a packing problem model of the small rectangle of business in the large rectangle of computing resources; map all resource availability restrictions and the business's own processing requirements in the zero-carbon data center scenario into restrictions on whether the small rectangle of business is allowed to be placed, calculate the amount of clean energy required for the business, and allocate the business to a set number of computing resources to support the time point when the business is completed; solve the packing problem model to obtain an allocation plan.

10. The zero-carbon data center resource scheduling system using unstable clean energy according to claim 9, characterized in that: The width of the large rectangle of computing resources represents the total number of available servers within a scheduling cycle, and the length represents the timeline corresponding to the scheduling cycle; the width of the small rectangle of the business represents the number of servers to which the business will be allocated, and the length represents the computing time required after the business is allocated a set number of servers.

Images