Task processing method and computing device

Abstract

The embodiment of the invention provides a task processing method and computing equipment, is applied to the technical field of servers, and is used for reducing fluctuation of electrical load of an intelligent computing center, improving power grid safety of the intelligent computing center and reducing operation cost. The method specifically comprises the steps that a computing device obtains target electrical load information; wherein the target power utilization load comprises target power utilization state information and / or target load state information, the target power utilization state information is used for describing the power utilization state of the first power grid, and the target load state information comprises the load state of the main computing power cluster; and the computing device processes the work task of the computing power cluster according to the target power utilization state information and / or the target load state information. Wherein the work task of the computing power cluster comprises the work task of the main computing power cluster and / or the work task of the standby computing power cluster.

Description

Technical Field

[0001] This application relates to the technical field of servers, and in particular, to a task processing method and a computing device. Background Art

[0002] With the rapid development of Artificial Intelligence (AI), the demands for AI applications and training are increasing day by day. Currently, a computing power cluster with a primary and standby architecture (also known as an intelligent computing center) is often adopted to ensure a large number of AI applications and training. The computing power cluster with a primary and standby architecture includes a primary computing power cluster and a standby computing power cluster. Among them, the primary computing power cluster is responsible for executing work tasks, and the standby computing power cluster is used as a backup to quickly take over tasks when the primary computing power cluster fails or is overloaded.

[0003] In one application, the primary computing power cluster and the standby computing power cluster are powered by different power grids. At this time, the primary computing power cluster and the standby computing power cluster are computing power clusters across regions. For computing power clusters across regions, there will be problems of sudden increase or decrease in the power consumption load of the computing power cluster, which leads to poor power grid security and high operating costs of the computing power cluster. Summary of the Invention

[0004] Embodiments of this application provide a task processing method and a computing device, which are used to reduce the fluctuation of the power consumption load of the intelligent computing center, improve the power grid security of the intelligent computing center, and reduce the operating cost.

[0005] In a first aspect, embodiments of this application provide a task processing method, which is applied to a computing device. The computing device is communicatively connected to a primary computing power cluster, a standby computing power cluster, a first power grid, and a second power grid respectively. Among them, the primary computing power cluster and the standby computing power cluster are located in different regions, the primary computing power cluster is powered by the first power grid, and the standby computing power cluster is powered by the second power grid.

[0006] Specifically, the computing device obtains target power consumption load information. Among them, the target power consumption load includes target power consumption status information and / or target load status information. The target power consumption status information is used to describe the power consumption status of the first power grid, and the target load status information includes the load status of the primary computing power cluster. The computing device processes the work tasks of the computing power cluster according to the target power consumption status information and / or the target load status information. Among them, the work tasks of the computing power cluster include the work tasks of the primary computing power cluster and / or the work tasks of the standby computing power cluster. Embodiments of this application control the work tasks of the computing power cluster through the power consumption status information provided by the power grid or the load status information provided by the computing power cluster, so as to adjust the power consumption load of the computing power cluster. Thereby, the impact on the power grid caused by the fluctuation of the power consumption load of the intelligent computing center is avoided, the power grid security of the intelligent computing center is improved, and the operating cost of the computing power cluster is reduced.

[0007] In one implementation, when the target power consumption status information indicates that the first power grid is in the first power consumption state, that is, the first power grid is in the peak power consumption period, the computing device reduces the work tasks being executed by the main computing power cluster; when the target power consumption status information indicates that the first power grid is in the second power consumption state, that is, the first power grid is in the off-peak power consumption period, the computing device increases the work tasks being executed by the main computing power cluster.

[0008] In another implementation, when the target load status information indicates that the main computing power cluster is in the first load state, that is, the main computing power cluster is in the overloaded state, the computing device reduces the work tasks being executed by the main computing power cluster; when the target load status information indicates that the main computing power cluster is in the second load state, that is, the main computing power cluster is in the low-load state, the computing device increases the work tasks being executed by the main computing power cluster.

[0009] Through the above two methods, the computing device accurately judges the power consumption load of the computing power cluster based on the power consumption status of the first power grid or the load status of the main computing power cluster, so as to accurately control the work tasks of the computing power cluster.

[0010] In yet another implementation, the target power consumption status information includes the first power consumption price and the first power consumption load of the first power grid; when the first power consumption price is greater than the first preset power price, and / or, the first power consumption load is greater than the highest threshold of the first load interval, it indicates that the first power grid is in the first power consumption state; when the first power consumption price is less than or equal to the first preset power price, and / or, the first power consumption load is less than or equal to the lowest threshold of the first load interval, it indicates that the first power grid is in the second power consumption state.

[0011] In still another implementation, if the target load status information includes the first utilization rate, and the first utilization rate is the utilization rate of the processor in the main computing power cluster; when the first utilization rate is greater than the highest threshold of the preset load interval, it indicates that the main computing power cluster is in the first load state; if the first utilization rate is less than the lowest threshold of the preset load interval, it indicates that the main computing power cluster is in the second load state. Thus, the computing device can use the first utilization rate to avoid the processor of the computing power cluster from being idle or overloaded, which helps to reduce the overall energy consumption.

[0012] In one implementation, the computing device stops or postpones the execution of the offline work tasks on the main computing power cluster to reduce the power consumption load of the main computing power cluster.

[0013] In yet another implementation, the target power consumption status information includes the first power consumption price and the first power consumption load of the first power grid, and the target load status information includes the first utilization rate. The first utilization rate is the processor utilization rate in the main computing power cluster. If one of the first preset condition, the second preset condition, and the third preset condition is satisfied, the online working tasks of the main computing power cluster are controlled to be switched to the standby computing power cluster for execution; wherein, the first preset condition is that the second power consumption price of the second power grid is less than the first power consumption price; the second preset condition is that the second power consumption price is less than or equal to the second preset price, or the second power consumption load of the second power grid is less than the highest threshold of the second load interval; the third preset condition is that the second utilization rate is less than the highest threshold of the preset load interval; the second utilization rate is the processor utilization rate in the standby computing power cluster. That is, when ensuring the normal operation of the standby computing power cluster, the computing device reduces the power consumption load of the main computing power cluster by transferring the online working tasks of the main computing power cluster to the standby computing power cluster.

[0014] Further, if the fourth preset condition is satisfied, or the fifth preset condition is satisfied, the computing device controls the online working tasks of the standby computing power cluster to be switched to the main computing power cluster for execution; wherein, the fourth preset condition is that the first power consumption price is less than or equal to the first preset price, or the first power consumption load is less than or equal to the highest threshold of the first load interval; the fifth preset condition is that the first utilization rate is less than the highest threshold of the preset load interval. Thus, when the power consumption of the main computing power cluster is small, or it is not in an overloaded state, the main computing power cluster can continue to execute the online working tasks.

[0015] In yet another implementation, when the target power consumption status information indicates that the first power grid is in the first power consumption state, and the target load status information indicates that the main computing power cluster is in the first load state, or when the target power consumption status information indicates that the first power grid is in the second power consumption state, or the first power grid is in the first power consumption state, and the target load status information indicates that the main computing power cluster is in the second load state, the working tasks being executed by the main computing power cluster are increased; wherein, the power consumption of the first power consumption state is higher than that of the second power consumption state, and the load of the first load state is higher than that of the second load state. When the first power grid is in a power consumption underestimation state or a low load state, the computing device can increase the power consumption load by continuing to execute the offline working tasks, avoiding sudden changes in the power consumption load.

[0016] In another implementation, when the target power consumption status information indicates that the first power grid is in the first power consumption state and the target load status information indicates that the main computing power cluster is in the first load state, the computing device stops or postpones the execution of the offline work tasks on the main computing power cluster when the target power consumption status information indicates that the first power grid is in the first power consumption state and the target load status information indicates that the main computing power cluster is in the first load state; when the electricity price of the first power grid is higher than that of the second power grid, it controls the online work tasks of the main computing power cluster to be switched to the standby computing power cluster for execution; when the target power consumption status information indicates that the first power grid is in the first power consumption state and the target load status information indicates that the main computing power cluster is in the third load state, it controls the online work tasks of the standby computing power cluster to be switched to the main computing power cluster for execution; when the target power consumption status information indicates that the first power grid is in the second power consumption state and the target load status information indicates that the main computing power cluster is in the second load state, it starts or resumes the execution of the offline work tasks on the main computing power cluster; where the load of the first load state is higher than that of the third load state, and the load of the third load state is higher than that of the second load state. Thus, by monitoring and dynamically scheduling the load status of the computing power cluster of the computing device, the utilization of power resources is effectively optimized and power grid fluctuations are addressed. When the first power grid is in the peak power consumption period and the load of the computing power cluster is too high, by controlling the execution of offline tasks, resource overloading is avoided and the online work tasks are switched to the standby cluster, thereby maximizing the energy usage efficiency. In the case of low power consumption or low load during the off-peak period, tasks are allowed to continue execution to maintain business continuity and reduce energy waste.

[0017] Furthermore, the computing device can control the online work tasks with a target latency greater than or equal to the preset latency in the online work tasks of the main computing power cluster to be switched to the standby computing power cluster for execution; the target latency is the maximum latency that the user can wait for the response of the online work task, and the preset latency is the switching latency for the online work task to be switched from the main computing power cluster to the standby computing power cluster.

[0018] In a second aspect, an embodiment of the present application provides a task processing device applied to a computing device. The computing device is communicatively connected to the main computing power cluster, the standby computing power cluster, the first power grid, and the second power grid respectively, where the main computing power cluster and the standby computing power cluster are located in different regions, the main computing power cluster is powered by the first power grid, and the standby computing power cluster is powered by the second power grid.

[0019] The device includes: an acquisition unit, configured to acquire target power consumption load information, where the target power consumption load includes target power consumption status information and / or target load status information, the target power consumption status information is used to describe the power consumption state of the first power grid, and the target load status information includes the load status of the main computing power cluster;

[0020] A processing unit, configured to process the work tasks of the computing power cluster according to the target power consumption status information and / or the target load status information. The work tasks of the computing power cluster include the work tasks of the primary computing power cluster and / or the work tasks of the backup computing power cluster.

[0021] In a third aspect, an embodiment of the present application provides a computing device, including:

[0022] A memory, configured to store programs;

[0023] A processor, configured to execute the programs stored in the memory. When the programs stored in the memory are executed, the processor is configured to execute the method according to any one of the first aspect.

[0024] In a fourth aspect, the present application provides a computer storage medium, configured to store a computer program. When the computer program is executed, it is configured to implement the method provided by any one of the implementation manners in the first aspect of the present application.

[0025] In a fifth aspect, the present application provides a computer program product containing instructions. When it runs on at least one computing device, it enables at least one computing device to implement the method provided by any one of the implementation manners in the first aspect of the present application.

[0026] Any of the above-provided task processing methods, corresponding computing devices, computer-readable storage media, computer program products, etc. are all used to execute the corresponding methods provided above. Therefore, the beneficial effects that can be achieved can refer to the beneficial effects in the corresponding methods, which will not be elaborated here. Description of the Drawings

[0027] Figure 1 It is a schematic diagram of an application scenario provided by an embodiment of the present application;

[0028] Figure 2 It is a schematic diagram of another application scenario provided by an embodiment of the present application;

[0029] Figure 3 It is a flowchart of a task processing method provided by an embodiment of the present application;

[0030] Figure 4A It is a flowchart of a task processing method provided by an embodiment of the present application;

[0031] Figure 4B It is a flowchart of an implementation for controlling the work tasks of the computing power cluster provided by an embodiment of the present application;

[0032] Figure 5 It is a flowchart of yet another task processing method provided by an embodiment of the present application;

[0033] Figure 6Schematic structural diagram of a task processing device provided by an embodiment of the present application;

[0034] Figure 7 Schematic structural diagram of a server provided by an embodiment of the present application. Detailed implementation manners

[0035] The technical solutions in the embodiments of the present application will be clearly and completely described below. Apparently, the described embodiments are some, but not all, of the embodiments of the present application. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present application without creative efforts shall fall within the protection scope of the present application.

[0036] The embodiment of the present application provides a task processing method. By using the power consumption status information provided by the power grid or the load status information provided by the computing power cluster, the work tasks of the computing power cluster are controlled, the impact on the power grid caused by the power consumption load fluctuation of the intelligent computing center is avoided, the power grid security of the intelligent computing center is improved, and the operation cost of the computing power cluster is reduced.

[0037] Among them, the intelligent computing center (also known as the intelligent computing center or the computing power cluster) is an infrastructure that provides powerful computing capabilities, data storage, and algorithm services for AI applications or training. The computing power cluster consists of a large number of high-performance computing devices and is used to execute specific computing tasks (or work tasks).

[0038] The application scenarios of the embodiments of the present application will be introduced below with reference to the accompanying drawings.

[0039] Exemplarily, Figure 1 Schematic diagram of an application scenario provided by an embodiment of the present application.

[0040] As Figure 1 shown, it specifically involves a computing device 101, a first computing power cluster 102, and a second computing power cluster 103. The first computing power cluster 102 is powered by the power grid 104 in the first region, and the second computing power cluster 103 is powered by the power grid 105 in the second region. Among them, the first computing power cluster is located in the first region, and the second computing power cluster is located in the second region. Among them, different power grids are adopted in the first region and the second region.

[0041] The first computing power cluster 102 and the second computing power cluster 103 are backup to each other. That is, when the first computing power cluster 102 fails or needs to be maintained, the second computing power cluster 103 can take over the work tasks of the first computing power cluster 102. When the second computing power cluster 103 fails or needs to be maintained, the first computing power cluster 102 can take over the work tasks of the second computing power cluster 103. This two-way backup relationship can achieve high availability and avoid service interruption caused by a single point of failure.

[0042] In the embodiments of the present application, the computing power cluster responsible for executing most of the work tasks is called the primary computing power cluster, and the other computing power cluster as a backup is called the backup computing power cluster. Among them, when the first computing power cluster executes most of the work tasks, the first computing power cluster is the primary computing power cluster; otherwise, the second computing power cluster is called the primary computing power cluster. For the convenience of description, the following takes the first computing power cluster as the primary computing power cluster and the second computing power cluster as the backup computing power cluster as an example for illustration.

[0043] In the embodiments of the present application, the computing device 101 is respectively communicatively connected to the first computing power cluster 102 and the corresponding power grid 104, the second computing power cluster 103 and the corresponding power grid 105. For the first computing power cluster and the second computing power cluster, they are communicatively connected through a network. In one example, the network can be an optical network.

[0044] The computing device 101 is used to obtain the first power consumption load information (i.e., the target power consumption load information). Among them, the first power consumption load includes the first power consumption state information (i.e., the target power consumption state information) and / or the first load state information (the target load state information). The first power consumption state information is used to describe the power consumption state of the first power grid, and the first load state information includes the load state of the first computing power cluster. The computing device 101 is also used to control the work tasks of the computing power cluster according to the first power consumption state information and / or the first load state information. Among them, the work tasks of the computing power cluster include the work tasks of the first computing power cluster and / or the work tasks of the second computing power cluster.

[0045] Thus, in the embodiments of the present application, the work tasks of the computing power cluster are controlled through the target power consumption state information of the first power grid provided by the power grid or the target load state information provided by the primary computing power cluster, so as to adjust the power consumption load of the primary computing power cluster. Thus, the impact on the power grid caused by the power consumption load fluctuation of the intelligent computing center is avoided, the power grid security of the intelligent computing center is improved, and the operation cost of the computing power cluster is reduced.

[0046] Furthermore, a computing power and electricity collaborative control system can also be deployed on the computing device 101, which is used to execute the task processing method provided by the embodiments of the present application.

[0047] Exemplarily, Figure 2 is a schematic diagram of another application scenario provided by the embodiments of the present application.

[0048] Such as Figure 2As shown in the figure, the computing device 101 deploys an arithmetic-electricity collaborative control system. The arithmetic-electricity collaborative control system includes a computing power cluster monitoring module, a power grid information monitoring module, and a control module. Among them, the computing power cluster monitoring module can monitor the first power consumption status information and / or the second power consumption status information in real time, and the power grid information monitoring module can monitor the first load status information and / or the second load status information in real time. Among them, the second power consumption status information is the power consumption status of the power grid corresponding to the second computing power cluster, and the second load status is used to describe the load status of the second computing power cluster. Figure 2 Exemplarily, it is shown that the first power consumption status information and the second power consumption status information are electricity prices and / or electricity loads.

[0049] Furthermore, a first cluster management system and a second cluster management system are respectively set up for the first computing power cluster and the second computing power cluster. The arithmetic-electricity collaborative control system is used to adjust the work tasks of the first computing power cluster through the first cluster management system and adjust the work tasks of the second computing power cluster through the second cluster management system.

[0050] It should be noted that the application scenarios provided in the embodiments of the present application can also be other cross-regional primary and standby computing power cluster application scenarios, and the embodiments of the present application do not specifically limit them.

[0051] It should be noted that in actual use, the computing device 101 can be a computing device deployed in the primary computing power cluster or a computing device deployed in the standby computing power cluster, and the embodiments of the present application do not specifically limit it.

[0052] It should be noted that the computing device 101 provided in the embodiments of the present application can be a server, specifically an independent physical server, or a server cluster or distributed system composed of multiple physical servers, or a cloud server providing basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, cloud communications, middleware services, domain name services, security services, Content Delivery Network (CDN), and large databases and artificial intelligence platforms. When the above server is a server cluster or distributed system composed of multiple physical servers, the multiple physical servers can form a blockchain, and each physical server is a node on the blockchain. The physical type of the server area can be a rack server, a high-density server, a GPU server, a tower server, or a blade server, a whole cabinet server, etc., and the present application does not specifically limit it.

[0053] The following is a detailed and complete description of the task processing method provided in the embodiments of the present application with reference to the accompanying drawings. The task processing method in the following text is applied in Figure 1 or Figure 2In the described application scenario, taking the first computing power cluster as the main computing power cluster, the power grid corresponding to the first computing power cluster is called the first power grid, and the second computing power cluster as the backup computing power cluster, and the power grid corresponding to the second computing power cluster is called the second power grid as an example for illustration. The computing device in the following text refers to computing device 101.

[0054] Appendix Figure 3 The following is a flowchart of a task processing method provided by an embodiment of the present application, and the method includes the following content:

[0055] S310. The computing device obtains first power consumption load information.

[0056] The first power consumption load information indicates the power consumption load status of the first computing power cluster, that is, whether the first computing power cluster is in a high power consumption load state or a low power consumption load state. Among them, the power consumption charge in the high power consumption load state is less than the power consumption load in the low power consumption load state. It can be understood that if the first power consumption load information indicates that the first computing power cluster is in a high power consumption load state, it is necessary to adjust the working tasks of the computing power cluster to avoid sudden increase of the first computing power cluster.

[0057] In the embodiment of the present application, the first power consumption load information includes first power consumption status information and / or first load status information. That is, the first power consumption load information can be first power consumption status information, or first load status information, or first power consumption status information and first load status information.

[0058] The first power consumption status information is used to indicate the power consumption status of the first power grid. For example, the first power grid is in a peak power consumption period (also called the first power consumption status), or the first power grid is in a valley power consumption period (also called the second power consumption status). For the same power grid, the power consumption during the peak power consumption period is higher than that during the valley power consumption period.

[0059] Since the power consumption load of the high-power grid computing power cluster is often less than that of the low-power grid computing power cluster, the first power consumption status information can be used to indicate whether the first computing power cluster is in a high-load operation state. Among them, the high-power grid computing power cluster is the computing power cluster that supplies power to the power grid in the peak power consumption period, and the low-power grid computing power cluster is the computing power cluster that supplies power to the power grid in the valley power consumption period.

[0060] In the embodiment of the present application, the first power consumption status information can be the power consumption price and / or the power consumption load of the first power grid. The power consumption price is used to describe the power grid of the current power grid, and the power consumption load is used to describe the power consumption load of the current power grid.

[0061] The computing device can obtain the electricity price and / or electricity load of the first power grid in real time through the power grounding system (also known as the IT system) of the first power grid. Thus, the computing device can adjust the electricity load of the first computing power cluster in real time based on the first electricity status information obtained in real time, avoiding sudden increases or decreases in the electricity load of the first computing power cluster.

[0062] The first load status information is used to indicate the load status of the first computing power cluster. For example, the first computing power cluster is in an overloaded operation state (referred to as the first load status), or the first computing power cluster is in a low-load operation state (referred to as the second load status). It should be noted that the electricity load of the computing power cluster in the overloaded operation state is less than that of the computing power cluster in the low-load operation state. That is to say, the computing power cluster in the overloaded operation state is often in a high-load operation state. Therefore, based on the first load status information, it can be indicated whether the first computing power cluster is in a high-load operation state.

[0063] In the embodiments of the present application, the first load status information may be the processor utilization rate of the first computing power cluster (referred to as the first utilization rate). Among them, the processor may be a Central Processing Unit (CPU), a Graphics Processing Unit (GPU), and / or a Neural Processing Unit (NPU). Among them, the first utilization rate refers to the current usage rate of the processor in the computing power cluster, usually expressed as a percentage (%), reflecting the load status of the computing power cluster. In the embodiments of the present application, the first utilization rate can be used to avoid the processor of the computing power cluster from being idle or overloaded, so it helps to reduce the overall energy consumption.

[0064] S320. The computing device controls the work tasks of the computing power cluster according to the first electricity load information.

[0065] The computing device can control the work tasks of the computing power cluster according to the first electricity status information. It can also control the work tasks of the computing power cluster according to the first load status information, and can also comprehensively consider the first electricity status information and the first load status information to control the work tasks of the computing power cluster. Among them, the work tasks of the computing power cluster include the work tasks of the first computing power cluster and / or the work tasks of the second computing power cluster.

[0066] Example 1: The computing device determines whether the first power grid is in a peak electricity consumption period according to the first electricity status information. If the computing device determines that the first power grid is in a peak electricity consumption period, it reduces the work tasks being executed by the first computing power cluster. If the computing device determines that the first power grid is in a valley electricity consumption period, it increases the work tasks being executed by the first computing power cluster.

[0067] In an embodiment of the present application, if the first power consumption status information includes the electricity price and the power consumption load of the first power grid, the computing device may determine whether the first power grid is in a peak power consumption period according to the electricity price of the first power grid (i.e., the first electricity price) and / or the power consumption load (i.e., the first power consumption compliance).

[0068] In a specific implementation, the computing device may first obtain the average electricity price during the flat period of the first power grid (i.e., the first preset electricity price). Among them, the flat period of the power grid is between the peak period and the low period, and refers to the time period when the power consumption load of the power grid is within the normal load range. For example, the flat period of the first power grid is from 10:00 pm to 7:00 am the next day.

[0069] The computing device compares the electricity price of the first power grid with the average electricity price during the flat period of the first power grid. If the electricity price of the first power grid is greater than the average electricity price during the flat period of the first power grid, it is determined that the first power grid is in a peak power consumption period. If the electricity price of the first power grid is less than or equal to the average electricity price during the flat period of the first power grid, it is determined that the first power grid is in a low power consumption period.

[0070] In another specific implementation, the computing device compares the power consumption load of the first power grid with the highest threshold of the first load interval involved in the first power grid. If the power consumption load of the first power grid is greater than the highest threshold of the first load interval, it is determined that the first power grid is in a peak power consumption period. If the power consumption load of the first power grid is less than or equal to the lowest threshold of the first load interval, it is determined that the first power grid is in a low power consumption period. Among them, the first load interval is the normal load interval set for the first power grid. For example, the first load interval is [a, b], where a < b and a is a positive number. Then the highest threshold of the first load interval is b. If the power consumption load of the first power grid is greater than b, the computing device determines that the first power grid is in a peak power consumption period.

[0071] In yet another specific implementation, if the computing device determines that the electricity price of the first power grid is greater than the average electricity price during the flat period of the first power grid, and the power consumption load of the first power grid is greater than the highest threshold of the first load interval, it is determined that the first power grid is in a peak power consumption period. If the computing device determines that the electricity price of the first power grid is less than or equal to the average electricity price during the flat period of the first power grid, and the electricity price of the first power grid is less than or equal to the lowest threshold of the first load interval, it is determined that the first power grid is in a low power consumption period.

[0072] Example 2, if the computing device determines whether the first computing power cluster is in an overloaded state according to the first load status information. If the computing device determines that the first computing power cluster is in an overloaded state, reduce the work tasks being executed by the first computing power cluster. If the computing device determines that the first computing power cluster is in a low load state, increase the work tasks being executed by the first computing power cluster

[0073] In an embodiment of the present application, if the first utilization rate is greater than the highest threshold of the preset load range, it is determined that the first computing power cluster is in an overloaded state (i.e., the first load state). Wherein, the preset load range is a preset normal load range. For example, the preset load range is [50% - 80%], its highest threshold is 80%, and the lowest threshold is 50%.

[0074] Further, if the first utilization rate is less than the lowest threshold of the preset load range, it is determined that the first computing power cluster is in a low load state (i.e., the second load state). In addition, if the first utilization rate is within the preset load range, it is determined that the first computing power cluster is in a normal load state (i.e., the third load state).

[0075] In an embodiment of the present application, the computing device reducing the work tasks being executed by the first computing power cluster includes but is not limited to: stopping or postponing the execution of the offline work tasks of the first computing power cluster, switching the work tasks of the first computing power cluster to the second computing power cluster for execution, etc.

[0076] It should be noted that the computing device can sort the offline tasks according to the power consumption from large to small and stop executing them one by one from large to small, or it can also sort the offline tasks according to the importance priority and stop executing them one by one from low to high. The embodiments of the present application do not specifically limit this.

[0077] The computing device increasing the work tasks being executed by the first computing power cluster includes but is not limited to: switching the work tasks of the second computing power cluster to the first computing power cluster for execution, starting or restarting the execution of the offline work tasks of the first computing power cluster.

[0078] It should be noted that the computing device can sort the offline tasks according to the power consumption from large to small and resume execution one by one from small to large, or it can also sort the offline tasks according to the importance priority and resume execution one by one from high to low. The embodiments of the present application do not specifically limit this.

[0079] Online work tasks have the characteristics of real-time response to customer requests and being latency-sensitive, such as e-commerce services, live broadcast services, etc. Offline work tasks do not require real-time response to customer requests and have long processing times, such as model training, data preprocessing, log analysis, etc. Different processing is performed for different work tasks to achieve refined control of the work tasks of the computing power cluster.

[0080] Example 3: If the computing device determines that the first power grid is in a peak electricity consumption period and the first computing power cluster is in an overloaded state, reduce the workload of the first computing power cluster; if the computing device determines that the first power grid is in a peak electricity consumption period and the first computing power cluster is in a low-load state, or the first power grid is in a valley electricity consumption period, increase the workload of the first computing power cluster. If the computing device determines that the first power grid is in a peak electricity consumption period and the first computing power cluster is in a normal load state, switch the online workload executed by the second computing power cluster to the first computing power cluster for execution. Further, the online tasks executed by the second computing power cluster are specifically online tasks requested through the first computing power cluster and executed by the second computing power cluster.

[0081] The following takes the XX company's deployment of GPU clusters with 3000 cards (i.e., 3000 independent GPUs) in Area A and Area B respectively to serve the AI applications in Area C as an example to illustrate Example 3. Here, the first power grid is the power grid in Area A, the second power grid is the power grid in Area B, the first computing power cluster is the computing power cluster in Area A, and the second computing power cluster is the computing power cluster in Area B. Among them, the connection requested by the user is the connection requested for the power grid service. The load status information of the computing power cluster is the CPU / GPU / NPU utilization rate.

[0082] If the power grid in Area A is in a peak electricity consumption period and the CPU / GPU / NPU utilization rate of the computing power cluster in Area A is greater than the first utilization rate threshold, control the offline workload of the computing power cluster in Area A to stop execution or postpone execution; switch the e-commerce service requests in Area A to the computing power cluster in Area B for response. Among them, the first utilization rate threshold is the highest threshold of the preset load range. For example, if the preset load range is [50% - 80%], then the first utilization rate threshold is 80%.

[0083] If the power grid in Area A is in a peak electricity consumption period and when the CPU / GPU / NPU utilization rate of the computing power cluster in Area A is less than the first utilization rate threshold and the duration of maintenance is greater than or equal to the preset duration threshold (for example, the preset duration threshold is 30 seconds), control the computing power cluster in Area A to take over the user requests for the e-commerce service in Area A again.

[0084] If the power grid in Area A is in a valley electricity consumption period, or the power grid in Area A is in a peak electricity consumption period but the CPU / GPU / NPU utilization rate of the computing power cluster in Area A is less than the second utilization rate threshold, control the offline workload of the computing power cluster in Area A to start execution or resume execution. Among them, the second utilization rate threshold is the lowest threshold of the preset load range. For example, if the preset load range is [50% - 80%], then the second utilization rate threshold is 50%.

[0085] Thus, the computing device monitors the load status of the computing power clusters in Area A and Area B in real time and performs dynamic scheduling control, effectively optimizing the utilization of power resources and coping with power grid fluctuations. When the power grid in Area A is at the peak electricity consumption period and the load of the computing power cluster is too high, by controlling the execution of offline tasks, resource overloading is avoided, and commercial requests are switched to the standby cluster, thereby maximizing the energy usage efficiency. During the low electricity consumption period or when the load is too low, tasks are allowed to continue execution to maintain business continuity and reduce energy waste.

[0086] In a specific implementation, the computing power and electricity coordination control system can control the work tasks of the first computing power cluster through the first cluster management system, and control the work tasks of the second computing power cluster through the second cluster management system.

[0087] That is, in the embodiments of the present application, the work tasks of the computing power cluster can be controlled through the electricity usage status information provided by the power grid and the load status information provided by the computing power cluster, so as to adjust the electricity load of the computing power cluster. Thus, the impact on the power grid caused by the electricity load fluctuation of the intelligent computing center is avoided, the power grid security of the intelligent computing center is improved, and the operation cost is reduced.

[0088] Furthermore, in the embodiments of the present application, the electricity load information of the computing power cluster is specifically described by taking the electricity usage status information of the computing power cluster as an example. Among them, the second computing power cluster corresponds to the second electricity load information, specifically the second electricity usage status information. The following combines the appendix Figure 4A for specific elaboration.

[0089] Appendix Figure 4A is a flowchart of a task processing method provided by the embodiments of the present application. The method includes the following content:

[0090] S410. The computing device obtains the work task information of the computing power cluster and / or the delay information between the primary and standby clusters.

[0091] The network delay between the primary and standby clusters indicates the network delay between the primary computing power cluster and the standby computing power cluster, reflecting the delay situation of task switching.

[0092] In the embodiments of the present application, the computing device can obtain the work task information and / or network delay of the primary computing power cluster. For example, the computing power and electricity coordination control system can obtain the work task information and / or network delay of the primary computing power cluster in real time through the primary cluster management system.

[0093] S420. The computing device obtains the first electricity usage status information and the second electricity usage status information.

[0094] The second electricity usage status information indicates the electricity usage status of the second power grid. Among them, the electricity usage status of the second power grid includes that the second power grid is at the peak electricity consumption period and the second power grid is at the low electricity consumption period.

[0095] In one implementation, the second power consumption status information may also include the electricity price of the second power grid (i.e., the second electricity price) and / or the power consumption load of the second power grid (i.e., the second power consumption load).

[0096] In one example, the electricity calculation and collaboration control system may obtain the second power consumption status information in real time through the IT system of the second power grid.

[0097] It should be noted that the computing device may obtain the first power consumption status information and the second power consumption status information simultaneously, or first obtain the first power consumption status and then obtain the second power consumption status information, or first obtain the second power consumption status information and then obtain the first power consumption status information. The embodiments of the present application do not specifically limit this.

[0098] It should be noted that steps S410 and S420 may be performed simultaneously, or S410 may be executed first and then S420, or S420 may be executed first and then S410. The embodiments of the present application do not specifically limit this.

[0099] S430. The computing device controls the work tasks of the computing power cluster according to the first power consumption status information and the second power consumption status information.

[0100] Appendix Figure 4B FIG. is a flowchart of an implementation for controlling the work tasks of a computing power cluster provided by an embodiment of the present application. Specifically, it includes the following content:

[0101] S4310. The computing device determines the power consumption status of the first power grid according to the first power consumption status information.

[0102] S4320. If the first power grid is in a peak power consumption period, the computing device controls the work tasks of the computing power cluster in combination with the second power consumption status information.

[0103] Specifically, the computing device first controls the offline tasks on the first computing power cluster to stop execution or postpone execution through the first cluster management system, thereby reducing the power consumption load of the first computing power cluster.

[0104] Then, the computing device determines the power consumption status of the second power grid according to the second power consumption status information. If the second power grid is in a valley power consumption period, the computing device controls the connection switching of user requests in the online work tasks of the first computing power cluster to the second computing power cluster through the first cluster management system, and the second computing power cluster responds to the user requests.

[0105] To improve the user experience, after determining that the second power grid is in the off-peak period, the computing device further calculates that the device, through the main cluster management system, switches the online work tasks on the first computing power cluster, in which the maximum tolerance delay of the user request is greater than or equal to the network delay between the primary and standby computing power clusters, to the second computing power cluster. For example, if the delay information between the primary and standby computing power clusters is 30 ms, the user request connections with a maximum tolerance delay of the user request greater than 30 ms are switched from the first computing power cluster to the second computing power cluster.

[0106] Exemplarily, the switch from the first computing power cluster to the second computing power cluster can be achieved by modifying the Internet Protocol (IP) address corresponding to the Domain Name System (DNS) domain name in the online work task to the IP address of the second computing power cluster.

[0107] It should be noted that if the electricity price of the second power grid is less than or equal to the average electricity price during the flat period of the second power grid (i.e., the second preset electricity price), and / or the electricity load of the second power grid is less than the highest threshold of the second load interval set for the second power grid, it is determined that the second power grid is in the off-peak period.

[0108] Alternatively, the computing device can compare the second electricity price with the first electricity price. If the second electricity price is less than the first electricity price, the computing device controls the user request connections in the online work tasks of the first computing power cluster to be switched to the second computing power cluster through the first cluster management system, and the second computing power cluster responds to the user requests.

[0109] Then, after a period of time, when the first electricity status information obtained in real time by the computing device indicates that the first power grid is not in the peak period, the computing device controls the first computing power cluster to take over the online work tasks at the location of the first computing power cluster through the first cluster management system. At the same time, the computing device controls the second computing power cluster to stop receiving and responding to the online work tasks at the location of the first computing power cluster through the second cluster management system.

[0110] Exemplarily, the computing device controls the first computing power cluster to take over the user requests at the location of the first computing power cluster through the first cluster management system, which can be achieved by modifying the IP address corresponding to the DNS domain name in the user request connection to the IP address of the first computing power cluster.

[0111] Thus, the computing device can, according to the first electricity status information and the second electricity status information, combined with the classification of the work tasks, finely adjust the work tasks of the computing power clusters, further avoid large fluctuations in the electricity load of the computing power clusters, and avoid affecting the processing of the online work tasks and improve the user experience.

[0112] S4330. If the first power grid is in a low electricity consumption period, the computing device adjusts the offline work tasks in the first computing power cluster to start execution or resume execution.

[0113] In summary, through the delay information between the primary and standby clusters and the second power consumption state in the embodiments of the present application, the refined adjustment of the computing power cluster tasks is realized, avoiding large fluctuations in the electricity load during the adjustment process and the impact on the processing of online work tasks, and improving the user experience.

[0114] Continuing to take the electricity load information of the computing power cluster as an example of the load state information of the computing power cluster for discussion. Attached Figure 5 FIG. is a flowchart of another task processing method provided by an embodiment of the present application. The method includes the following:

[0115] S510. The computing device obtains the work task information of the computing power cluster and / or the network delay between the primary and standby clusters.

[0116] S520. The computing device obtains the first load state information and the second load state information.

[0117] The second load state information is used to indicate the load state of the second computing power cluster.

[0118] It should be noted that the computing device can obtain the first load state information and the second load state information simultaneously, or obtain the first power consumption state first and then the second load state information, or obtain the second load state information first and then the first load state information. The embodiments of the present application do not specifically limit this.

[0119] It should be noted that steps S510 and S520 can be performed simultaneously, or S510 can be executed first and then S520, or S520 can be executed first and then S510. The embodiments of the present application do not specifically limit this.

[0120] S530. The computing device controls the work tasks of the computing power cluster according to the first load state information and the second load state information.

[0121] Specifically, it can include the following steps:

[0122] ① If the first computing power cluster is in an overloaded state, the computing device can control the work tasks of the computing power cluster as follows:

[0123] The computing device controls the offline work tasks of the first computing power cluster to stop execution or postpone execution through the first cluster management system, and compares the first electricity price and the second electricity price. If the first electricity price is greater than the second electricity price, the computing device controls the connection of user requests in the online work tasks of the first computing power cluster to be switched to the second computing power cluster through the first cluster management system, and the second computing power cluster responds to the user requests.

[0124] Alternatively, if the second utilization rate is less than the highest threshold of the preset load range, the computing device controls, through the first cluster management system, the switching of the user request connections in the online working tasks of the first computing power cluster to the second computing power cluster, and the second computing power cluster responds to the user requests.

[0125] Finally, after a period of time, when the first load status information obtained in real time by the computing device indicates that the first computing power cluster is not in an overloaded state, the computing device controls, through the first cluster management system, the first computing power cluster to take over the user requests at the location of the first computing power cluster. At the same time, the computing device controls, through the second cluster management system, the second computing power cluster to stop receiving and responding to the user requests at the location of the first computing power cluster.

[0126] ② If the first computing power cluster is in a low-load state, the computing device can control the working tasks of the computing power cluster as follows: The computing device controls, through the first cluster management system, the offline working tasks of the first computing power cluster to start execution or resume execution.

[0127] The embodiments of the present application can perform intelligent control on the working tasks of the computing power cluster. Specifically, when the first computing power cluster is in an overloaded state, by comparing the first and second electricity prices or considering the processor utilization rate of the second computing power cluster, the computing device can switch the online working tasks to the second computing power cluster to reduce the load pressure. At the same time, when the load of the second computing power cluster is lower than the set threshold, a response switch will also be made to ensure the reasonable allocation of resources. When the state of the first computing power cluster changes and it is no longer overloaded, the computing device can actively take over its tasks and at the same time stop receiving and responding to the tasks of the standby cluster, thereby improving the overall processing efficiency of the intelligent computing center for working tasks and ensuring the efficient operation of the computing power cluster. Further, the immediacy of the switch is also considered, improving the user experience.

[0128] In addition, the embodiments of the present application further provide a task processing device, which is applied to a computing device. The computing device is respectively communicatively connected to the main computing power cluster, the standby computing power cluster, the first power grid and the second power grid, wherein the main computing power cluster and the standby computing power cluster are located in different regions, the main computing power cluster is powered by the first power grid, and the standby computing power cluster is powered by the second power grid.

[0129] Appendix Figure 6 FIG. is a schematic structural diagram of a task processing device provided by an embodiment of the present application. The device 600 includes:

[0130] An obtaining unit 601, configured to obtain target power load information. The target power load includes target power status information and / or target load status information, the target power status information is used to describe the power status of the first power grid, and the target load status information includes the load status of the main computing power cluster.

[0131] A processing unit 602, configured to process the work tasks of the computing power cluster according to the target power consumption status information and / or the target load status information. The work tasks of the computing power cluster include the work tasks of the primary computing power cluster and / or the work tasks of the standby computing power cluster.

[0132] Optionally, the processing unit 602 is specifically configured to: when the target power consumption status information indicates that the first power grid is in the first power consumption state, that is, the first power grid is in the peak power consumption period, reduce the work tasks being executed by the primary computing power cluster; when the target power consumption status information indicates that the first power grid is in the second power consumption state, that is, the first power grid is in the valley power consumption period, increase the work tasks being executed by the primary computing power cluster.

[0133] Alternatively, when the target load status information indicates that the primary computing power cluster is in the first load state, that is, the primary computing power cluster is in the overloaded state, reduce the work tasks being executed by the primary computing power cluster; when the target load status information indicates that the primary computing power cluster is in the second load state, that is, the primary computing power cluster is in the low load state, increase the work tasks being executed by the primary computing power cluster.

[0134] Optionally, the target power consumption status information includes the first power consumption price and the first power consumption load of the first power grid; when the first power consumption price is greater than the first preset price, and / or, the first power consumption load is greater than the highest threshold of the first load interval, it indicates that the first power grid is in the first power consumption state; when the first power consumption price is less than or equal to the first preset price, and / or, the first power consumption load is less than or equal to the lowest threshold of the first load interval, it indicates that the first power grid is in the second power consumption state.

[0135] Optionally, if the target load status information includes the first utilization rate, the first utilization rate is the processor utilization rate in the primary computing power cluster; when the first utilization rate is greater than the highest threshold of the preset load interval, it indicates that the primary computing power cluster is in the first load state; if the first utilization rate is less than the lowest threshold of the preset load interval, it indicates that the primary computing power cluster is in the second load state.

[0136] Optionally, the processing unit 602 is specifically configured to: stop or postpone the execution of the offline work tasks on the primary computing power cluster.

[0137] Optionally, the target power consumption status information includes the first power consumption price and the first power consumption load of the first power grid, and the target load status information includes the first utilization rate. The first utilization rate is the processor utilization rate in the main computing power cluster. The processing unit 602 is further configured to: if one of the first preset condition, the second preset condition, and the third preset condition is satisfied, control the online working tasks of the main computing power cluster to be switched to the backup computing power cluster for execution; wherein, the first preset condition is that the second power consumption price of the second power grid is less than the first power consumption price; the second preset condition is that the second power consumption price is less than or equal to the second preset price, or the second power consumption load of the second power grid is less than the highest threshold of the second load interval; the third preset condition is that the second utilization rate is less than the highest threshold of the preset load interval; the second utilization rate is the processor utilization rate in the backup computing power cluster.

[0138] Further, if the fourth preset condition is satisfied, or the fifth preset condition is satisfied, the processing unit 602 is further configured to: control the online working tasks of the backup computing power cluster to be switched to the main computing power cluster for execution; wherein, the fourth preset condition is that the first power consumption price is less than or equal to the first preset price, or the first power consumption load is less than or equal to the highest threshold of the first load interval; the fifth preset condition is that the first utilization rate is less than the highest threshold of the preset load interval.

[0139] The processing unit 602 is further configured to: when the target power consumption status information indicates that the first power grid is in the first power consumption state and the target load status information indicates that the main computing power cluster is in the first load state, or when the target power consumption status information indicates that the first power grid is in the second power consumption state, or the first power grid is in the first power consumption state and the target load status information indicates that the main computing power cluster is in the second load state, increase the working tasks being executed by the main computing power cluster; wherein, the power consumption of the first power consumption state is higher than that of the second power consumption state, and the load of the first load state is higher than that of the second load state.

[0140] The processing unit 602 is further configured to: when the target power consumption status information indicates that the first power grid is in the first power consumption state and the target load status information indicates that the main computing power cluster is in the first load state, the computing device stops or postpones the execution of the offline work tasks on the main computing power cluster when the target power consumption status information indicates that the first power grid is in the first power consumption state and the target load status information indicates that the main computing power cluster is in the first load state; when the electricity price of the first power grid is greater than that of the second power grid, control the online work tasks of the main computing power cluster to be switched to the standby computing power cluster for execution; when the target power consumption status information indicates that the first power grid is in the first power consumption state and the target load status information indicates that the main computing power cluster is in the third load state, control the online work tasks of the standby computing power cluster to be switched to the main computing power cluster for execution; when the target power consumption status information indicates that the first power grid is in the second power consumption state and the target load status information indicates that the main computing power cluster is in the second load state, start or re-execute the offline work tasks of the main computing power cluster; wherein, the load of the first load state is higher than that of the third load state, and the load of the third load state is higher than that of the second load state.

[0141] Further, the computing device can control the online work tasks with a target delay greater than or equal to the preset delay in the online work tasks of the main computing power cluster to be switched to the standby computing power cluster for execution; the target delay is the maximum delay that the user can wait for the response of the online work task, and the preset delay is the switching delay for the online work task to be switched from the main computing power cluster to the standby computing power cluster.

[0142] Further, an embodiment of the present application also provides a server.

[0143] As Figure 7 shown, an embodiment of the present application also provides a server 700. The application scenario of the server 700 is not specifically limited. For example, the server 700 is introduced by taking a server as an example. Specifically, the type of the server is not limited either. For example, the server can be a rack server or an edge server. The server can be located in a data center or in other areas, which is not specifically limited in the embodiment of the present application.

[0144] The server 700 includes a processor 701 and a memory 703. The memory 703 is electrically connected to the processor 701 respectively; the memory 703 is used to store program instructions for the task processing method involved in the above embodiments; the processor 701 is used to call the corresponding part of the above program instructions so that the server can execute the task processing method involved in the above embodiments.

[0145] In the above embodiments, it can be implemented in whole or in part by software, hardware, firmware, or any combination thereof. When implemented using a software program, it can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer instructions are loaded and executed on a computer, the processes or functions in the embodiments of the present application are generated in whole or in part.

[0146] The embodiments of the present application also provide a computer program product containing instructions. The computer program product can be software or a program product containing instructions that can run on a computing device or be stored in any available medium. When the computer program product runs on a computing device, it causes the computing device to execute the above task processing method.

[0147] The embodiments of the present application also provide a computer-readable storage medium. The computer-readable storage medium can be any available medium that a computing device can store or a data storage device such as a data center containing one or more available media. The available medium can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium (e.g., solid-state drive), etc. The computer-readable storage medium includes instructions that direct the computing device to execute the above task processing method.

[0148] The descriptions of the processes or structures corresponding to the above respective drawings each have their own focuses. For parts not detailed in a certain process or structure, reference can be made to the relevant descriptions of other processes or structures.

[0149] As described above, the above are only specific embodiments of the present application, but the protection scope of the present application is not limited thereto. Any changes or substitutions within the technical scope disclosed in the present application should be covered by the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

Claims

1. A task processing method, characterized in that, Applied to a computing device, the computing device is used to communicate with a primary computing power cluster, a standby computing power cluster, a first power grid, and a second power grid. The primary computing power cluster is powered by the first power grid, and the standby computing power cluster is powered by the second power grid. The method includes: Obtain target power consumption load information, where the target power consumption load information includes target power consumption status information and / or target load status information; the target load status information is used to describe the load status of the primary computing power cluster; the target power consumption status information is used to describe the power consumption status of the first power grid; Process the work tasks of the computing power cluster according to the target power consumption status information and / or the target load status information; the work tasks of the computing power cluster include the work tasks of the primary computing power cluster and / or the work tasks of the standby computing power cluster.

2. The method according to claim 1, wherein The processing of the work tasks of the computing power cluster according to the target power consumption status information and / or the target load status information includes: When the target power consumption status information indicates that the first power grid is in a first power consumption state, reduce the work tasks being executed by the primary computing power cluster; when the target power consumption status information indicates that the first power grid is in a second power consumption state, increase the work tasks being executed by the primary computing power cluster; Or, When the target load status information indicates that the primary computing power cluster is in a first load state, reduce the work tasks being executed by the primary computing power cluster; when the target load status information indicates that the primary computing power cluster is in a second load state, increase the work tasks being executed by the primary computing power cluster; Wherein, the power consumption of the first power consumption state is higher than that of the second power consumption state, and the load of the first load state is higher than that of the second load state.

3. The method according to claim 2, wherein The target power consumption status information includes the first power consumption price and the first power consumption load of the first power grid; When the first power consumption price is greater than the first preset power price, and / or, the first power consumption load is greater than the highest threshold of the first load interval, it is indicated that the first power grid is in the first power consumption state; When the first power consumption price is less than or equal to the first preset power price, and / or, the first power consumption load is less than or equal to the lowest threshold of the first load interval, it is indicated that the first power grid is in the second power consumption state.

4. The method according to claim 2, wherein If the target load status information includes a first utilization rate, the first utilization rate is the processor utilization rate in the primary computing power cluster; When the first utilization rate is greater than the highest threshold of the preset load interval, it is indicated that the primary computing power cluster is in the first load state; If the first utilization rate is less than the lowest threshold of the preset load interval, it is indicated that the primary computing power cluster is in the second load state.

5. The method according to claim 2, wherein The reducing of the work tasks being executed by the primary computing power cluster includes: stopping or postponing the execution of the offline work tasks on the primary computing power cluster.

6. The method according to any one of claims 2 to 5, characterized in that, The target power consumption status information includes the first power consumption price and the first power consumption load of the first power grid. The target load status information includes the first utilization rate, which is the processor utilization rate in the main computing power cluster. Reducing the work tasks being executed by the main computing power cluster includes: If one of the first preset condition, the second preset condition, and the third preset condition is satisfied, controlling the online work tasks of the main computing power cluster to be switched to the standby computing power cluster for execution; Among them, the first preset condition is that the second power consumption price of the second power grid is less than the first power consumption price; the second preset condition is that the second power consumption price is less than or equal to the second preset price, or the second power consumption load of the second power grid is less than the highest threshold of the second load interval; the third preset condition is that the second utilization rate is less than the highest threshold of the preset load interval; the second utilization rate is the processor utilization rate in the standby computing power cluster.

7. The method according to claim 6, characterized in that, The method further includes: If the fourth preset condition is satisfied, or the fifth preset condition is satisfied, controlling the online work tasks of the standby computing power cluster to be switched to the main computing power cluster for execution; The fourth preset condition is that the first power consumption price is less than or equal to the first preset price, or the first power consumption load is less than or equal to the highest threshold of the first load interval; the fifth preset condition is that the first utilization rate is less than the highest threshold of the preset load interval.

8. The method according to claim 1, wherein Controlling the work tasks of the computing power cluster according to the target power consumption status information and / or the target load status information includes: When the target power consumption status information indicates that the first power grid is in the first power consumption state and the target load status information indicates that the main computing power cluster is in the first load state, reducing the work tasks being executed by the main computing power cluster; When the target power consumption status information indicates that the first power grid is in the second power consumption state, or the first power grid is in the first power consumption state and the target load status information indicates that the main computing power cluster is in the second load state, increasing the work tasks being executed by the main computing power cluster; Among them, the power consumption of the first power consumption state is higher than that of the second power consumption state, and the load of the first load state is higher than that of the second load state.

9. The method according to claim 8, characterized in that, When the target power consumption status information indicates that the first power grid is in the first power consumption state and the target load status information indicates that the main computing power cluster is in the first load state, reducing the work tasks being executed by the main computing power cluster includes: When the target power consumption status information indicates that the first power grid is in the first power consumption state and the target load status information indicates that the main computing power cluster is in the first load state, stopping or postponing the execution of the offline work tasks on the main computing power cluster; when the power price of the first power grid is greater than that of the second power grid, controlling the online work tasks of the main computing power cluster to be switched to the standby computing power cluster for execution; The method further includes: when the target power consumption status information indicates that the first power grid is in the first power consumption state and the target load status information indicates that the main computing power cluster is in the third load state, controlling the online work tasks of the standby computing power cluster to be switched to the main computing power cluster for execution; The increasing of the work tasks being executed by the main computing power cluster includes: starting to execute or re-executing the offline work tasks of the main computing power cluster; Wherein, the load of the first load state is higher than the load of the third load state, and the load of the third load state is higher than the load of the second load state.

10. A computing device, characterized in that, Comprising a processor and a memory, the processor and the memory are coupled; The memory is used for storing programs; The processor is used for executing the programs stored in the memory. When the programs stored in the memory are executed, the processor is used for executing the method according to any one of claims 1-9.

Images