CN118034942B

CN118034942B - Cluster computing management method, device, equipment and storage medium

Info

Publication number: CN118034942B
Application number: CN202410440578.2A
Authority: CN
Inventors: 韩运恒; 陈长愿; 李利红
Original assignee: Shenzhen Jieyi Technology Co ltd
Current assignee: Shenzhen Jieyi Technology Co ltd
Priority date: 2024-04-12
Filing date: 2024-04-12
Publication date: 2024-06-25
Anticipated expiration: 2044-04-12
Also published as: CN118034942A

Abstract

The application relates to the technical field of power management, and discloses a cluster power management method, device, equipment and storage medium, which comprise the steps of acquiring input power management demand information, wherein the power management demand information comprises node power use information and node power distribution information; determining a power management mode according to the node power use information, wherein the power management mode comprises a single-dimensional power management mode and a multi-dimensional power management mode; in the single-dimensional power management mode, determining a first cluster power management result according to node power distribution information and preset single-dimensional power management basis information, and carrying out cluster power calling management according to the first cluster power management result and preset first management requirements; and in the multi-dimensional power management mode, determining a second cluster power management result according to the node power distribution information and the preset multi-dimensional power management basis information, and carrying out cluster power calling management according to the second cluster power management result. The application improves the efficiency of cluster calculation management.

Description

Cluster computing management method, device, equipment and storage medium

Technical Field

The present application relates to the field of power management technologies, and in particular, to a cluster power management method, device, equipment, and storage medium.

Background

With the high-speed development of computing power network technology, cluster computing power is widely applied to different computing power network scenes, and meanwhile, higher requirements are also provided for managing the cluster computing power.

The traditional cluster power management method directly and uniformly calls and manages the power resources of all power nodes, the management mode is based on the power resources which are used and not used, the cluster power management method has great defects, and when the power nodes are more, the power resources which are required to be called and managed are more and the management is complex because the power resources which are required to be called and managed are more.

Disclosure of Invention

The application mainly aims to provide a cluster power management method, cluster power management equipment and a storage medium, aiming at improving the efficiency of cluster power management.

In order to achieve the above object, the present application provides a cluster power management method, including:

Acquiring input power management demand information, wherein the power management demand information comprises node power use information and node power distribution information;

Determining a power management mode according to the node power use information, wherein the power management mode comprises a single-dimensional power management mode and a multi-dimensional power management mode;

if the power management mode is the single-dimensional power management mode, determining a first cluster power management result according to the node power distribution information and preset single-dimensional power management basis information, and carrying out cluster power calling management according to the first cluster power management result and preset first management requirements;

and if the power management mode is the multi-dimensional power management mode, determining a second cluster power management result according to the node power distribution information and preset multi-dimensional power management basis information, and carrying out cluster power calling management according to the second cluster power management result.

In addition, to achieve the above object, the present application also provides a cluster power management apparatus, including:

the information acquisition module is used for acquiring input power management demand information, wherein the power management demand information comprises node power use information and node power distribution information;

The mode judging module is used for determining a power management mode according to the node power use information, wherein the power management mode comprises a single-dimension power management mode and a multi-dimension power management mode;

The first management module is used for determining a first cluster power management result according to the node power distribution information and preset single-dimensional power management basis information if the power management mode is the single-dimensional power management mode, and carrying out cluster power calling management according to the first cluster power management result and preset first management requirements;

And the second management module is used for determining a second cluster power management result according to the node power distribution information and preset multi-dimensional power management basis information if the power management mode is the multi-dimensional power management mode, and carrying out cluster power calling management according to the second cluster power management result.

In addition, to achieve the above object, the present application also provides a cluster computing management device, including: the system comprises a memory, a processor and a computer program stored on the memory and capable of running on the processor, wherein the computer program is executed by the processor to realize the steps of the cluster computing management method.

In addition, to achieve the above object, the present application further provides a cluster power management storage medium, on which a cluster power management program is stored, which when executed by a processor, implements the steps of the cluster power management method as described above.

The method comprises the steps of obtaining input power management demand information, wherein the power management demand information comprises node power use information and node power distribution information; determining a power management mode according to the node power use information, wherein the power management mode comprises a single-dimensional power management mode and a multi-dimensional power management mode; if the power management mode is the single-dimensional power management mode, determining a first cluster power management result according to the node power distribution information and preset single-dimensional power management basis information, and carrying out cluster power calling management according to the first cluster power management result and preset first management requirements; if the management mode of the computing power is the multi-dimensional management mode of the computing power, a second cluster computing power management result is determined according to the node computing power distribution information and the preset multi-dimensional management basis information, cluster computing power calling management is conducted according to the second cluster computing power management result, the management mode of the computing power is determined according to computing power management requirement information, in different computing power management modes, the cluster computing power management result is determined based on the computing power management requirement information and different preset management basis information, cluster management is conducted based on the cluster computing power management result, cluster computing power calling management is conducted in a single-dimensional computing power management mode, and in addition, when computing power nodes are more, only a plurality of computing power resources to be called are needed to be managed according to management in the prior art, the phenomenon that the computing power resources to be called are more is managed is generated, in the cluster computing power management mode, in different management modes, cluster computing power management is conducted based on the computing power management requirement information and different preset computing power management basis, the computing power is not needed to be called according to the cluster management result, and the computing power is not needed to be called by the computing power management information, and the cluster management is conducted according to the cluster management result is not needed to the cluster management information.

Drawings

FIG. 1 is a schematic diagram of a cluster computing management device of a hardware runtime environment according to an embodiment of the present application;

FIG. 2 is a flow chart of a first embodiment of a cluster power management method according to the present application;

FIG. 3 is a schematic diagram of a cluster power management device module according to the present application;

FIG. 4 is a schematic diagram of a cluster power management system of the present application;

FIG. 5 is a flow chart of the cluster power management method of the present application.

The achievement of the objects, functional features and advantages of the present application will be further described with reference to the accompanying drawings, in conjunction with the embodiments.

Detailed Description

It should be understood that the specific embodiments described herein are for purposes of illustration only and are not intended to limit the scope of the application.

Referring to fig. 1, fig. 1 is a schematic structural diagram of a cluster computing management device of a hardware running environment according to an embodiment of the present application.

As shown in fig. 1, the cluster computing management device may include: processor 0003, such as central processing unit (CentralProcessingUnit, CPU), communication bus 0001, fetch interface 0002, processing interface 0004, memory 0005. Wherein a communication bus 0001 is used to enable connected communication between these components. The acquisition interface 0002 may comprise an information acquisition device, an acquisition unit such as a computer, and the optional acquisition interface 0002 may also comprise a standard wired interface, a wireless interface. Processing interface 0004 may optionally comprise a standard wired interface, a wireless interface. The memory 0005 may be a high-speed random access memory (RandomAccessMemory, RAM) or a stable nonvolatile memory (Non-VolatileMemory, NVM), such as a disk memory. The memory 0005 may alternatively be a storage device separate from the aforementioned processor 0003.

Those skilled in the art will appreciate that the structure shown in fig. 1 does not constitute a limitation of the cluster computing management device, and may include more or fewer components than shown, or may combine certain components, or may be a different arrangement of components.

As shown in fig. 1, an operating system, an acquisition interface module, a processing interface module, and a cluster power management program may be included in the memory 0005 as one type of storage medium.

In the cluster power management device shown in fig. 1, the communication bus 0001 is mainly used to implement connection communication between components; the acquisition interface 0002 is mainly used for connecting a background server and carrying out data communication with the background server; the processing interface 0004 is mainly used for connecting a deployment end (user end) and carrying out data communication with the deployment end; the processor 0003 and the memory 0005 in the cluster power management device of the present application may be provided in the cluster power management device, where the cluster power management device invokes a cluster power management program stored in the memory 0005 through the processor 0003, and executes the cluster power management method provided by the embodiment of the present application.

For clarity and brevity of description of the following embodiments, a brief description of implementation of a cluster power management method is first given:

The conventional cluster power management method is based on power resources on used and unused nodes, wherein the management mode at least comprises new demand scheduling, existing demand updating scheduling, existing demand management and the like, when the management is directly based on the used and unused power resources, the management is carried out by determining related information of the power resources of each node in turn when the management is carried out each time, and when the nodes are too many, the whole management process is quite complex, so that the problem of low efficiency of the cluster power management is caused. Therefore, based on the problems, the cluster power management method provided by the application is provided, and the problem of low efficiency of cluster power management is further solved.

The application relates to a cluster power management method, which comprises the steps of obtaining input power management demand information, wherein the power management demand information comprises node power use information and node power distribution information; determining a power management mode according to the node power use information, wherein the power management mode comprises a single-dimensional power management mode and a multi-dimensional power management mode; if the power management mode is the single-dimensional power management mode, determining a first cluster power management result according to the node power distribution information and preset single-dimensional power management basis information, and carrying out cluster power calling management according to the first cluster power management result and preset first management requirements; if the management mode of the computing power is the multi-dimensional management mode of the computing power, a second cluster computing power management result is determined according to the node computing power distribution information and the preset multi-dimensional management basis information, cluster computing power calling management is conducted according to the second cluster computing power management result, the management mode of the computing power is determined according to computing power management requirement information, in different computing power management modes, the cluster computing power management result is determined based on the computing power management requirement information and different preset management basis information, cluster management is conducted based on the cluster computing power management result, cluster computing power calling management is conducted in a single-dimensional computing power management mode, and in addition, when computing power nodes are more, only a plurality of computing power resources to be called are needed to be managed according to management in the prior art, the phenomenon that the computing power resources to be called are more is managed is generated, in the cluster computing power management mode, in different management modes, cluster computing power management is conducted based on the computing power management requirement information and different preset computing power management basis, the computing power is not needed to be called according to the cluster management result, and the computing power is not needed to be called by the computing power management information, and the cluster management is conducted according to the cluster management result is not needed to the cluster management information.

Based on the hardware structure, the embodiment of the cluster computing management method is provided.

An embodiment of the present application provides a cluster computing management method, and referring to fig. 2, fig. 2 is a schematic flow chart of a first embodiment of the cluster computing management method of the present application.

In this embodiment, the cluster computing management method includes:

Step S10, acquiring input power management demand information, wherein the power management demand information comprises node power use information and node power distribution information;

Step S20, determining a power management mode according to the node power use information, wherein the power management mode comprises a single-dimension power management mode and a multi-dimension power management mode;

In this embodiment, when it is determined that an computing power node needs to perform management, input computing power management requirement information is acquired, where the computing power management requirement information includes node computing power usage information and node computing power distribution information, the computing power management requirement information refers to information that the computing power node needs to perform management, and includes information such as computing power usage conditions of the computing power node, the computing power node refers to a node storing computing power resources in computing power, the node computing power usage information refers to information related to computing power usage in the node, such as first computing power usage time information and first computing power remaining time information, that is, first computing power usage time information formed by using time of each computing power requirement in the node, the first computing power remaining resources refer to remaining information of computing power resources of computing power nodes in different time periods, the node computing power distribution information refers to distribution of computing power resources in the node, characteristics of computing power resources, and the like, and the computing power demand A occupies a power size of A1, the computing power occupied by the node computing power demand is a 10-11 am, the computing power demand is a full-day, and the computing power demand is a full-day 1. Further, at this time, the management mode of the computing force may be determined based on the computing force usage information of the node, where the management mode of the computing force includes a single-dimensional computing force management mode and a multi-dimensional computing force management mode, where the single-dimensional computing force management mode refers to at least one computing force demand in the node requiring computing force management, and the computing force demand does not call all computing force resources of the computing force node or occupy all usage time, so that the dimension is managed in a single dimension from the remaining computing force resources and the remaining resources in a special time, and the multi-dimensional computing force management mode refers to at least one computing force demand in the node requiring computing force management, and when there is one computing force demand, the computing force demand does not call all computing force resources of the computing force node or occupy all usage time, and when there is a plurality of computing force demands, the dimension requires multi-dimensional management from the remaining force resources and the remaining resources in a special time. And then the power management mode can be determined based on the power management demand information so as to perform scheduling management of different modes based on the power management mode, and further the cluster power management functionality can be greatly improved.

Step S30, if the power management mode is the single-dimensional power management mode, determining a first cluster power management result according to the node power distribution information and preset single-dimensional power management basis information, and performing cluster power calling management according to the first cluster power management result and preset first management requirements;

step S40, if the power management mode is the multidimensional power management mode, determining a second cluster power management result according to the node power distribution information and the preset multidimensional power management basis information, and performing cluster power calling management according to the second cluster power management result.

In this embodiment, when the mode is a one-dimensional power management mode, a first cluster power management result is determined based on the node power distribution information and the preset one-dimensional power management basis information, and then cluster power calling management is performed based on the determined first cluster power management result and the preset first management requirement. That is, based on the basis of the information and the node computing power distribution information, how to perform cluster computing power management is determined, and management is performed based on the management result. The single-dimensional power management basis information refers to basis information for performing single-dimensional management set by a user, such as usage time information, power characteristics and power remaining resources of each node (including nodes needing to be managed in the implementation), the usage time information refers to usage time period information of the power resources in each node, the power characteristics refer to characteristics of the power resources of each node, such as high speed, high bandwidth and the like, the power remaining conditions of the power resources of each node of the power remaining resources, the first cluster power management result refers to management results determined based on the power distribution information of the node and preset single-dimensional power management basis information, and when a certain node is used for calling or a new power demand characteristic is S, the power resources of a certain node can be directly called, the first management demand refers to the power demand needing to management, and the first management demand refers to the power demand change of the power resources can be generated according to user input or time demand change, such as new power demand input, and the power demand change of the existing power demand. And then, cluster computing power calling management can be carried out based on different first management requirements and the obtained first cluster computing power management result, so that the efficiency of cluster computing power management is ensured, and meanwhile, the functionality of cluster computing power management is greatly improved. And when the multi-dimensional power management mode is adopted, determining a second cluster power management result based on the node power distribution information and the preset multi-dimensional power management basis information, and then carrying out cluster power calling management based on the determined second cluster power management result. That is, based on the basis of the information and the node computing power distribution information, how to perform cluster computing power management is determined, and management is performed based on the management result. The multi-dimensional power management basis information refers to multi-dimensional management basis information set by a user, such as usage time information, power calculation characteristics and power calculation residual resources of each node (including nodes needing to be managed in the implementation), the usage time information refers to usage time period information of the power calculation resources in each node, the power calculation characteristics refer to characteristics of the power calculation resources of each node, such as high speed, high bandwidth and the like, the residual condition of the power calculation resources of each node of the power calculation residual resources (the residual condition can be formed by approximately the same information as the single-dimensional power management basis information, but the difference exists in the basis of the processing flow), and the second cluster power management result refers to a management result determined based on the power calculation distribution information of the node and the preset multi-dimensional power management basis information. And then, cluster computing power calling management can be carried out based on different second cluster computing power management results, so that the efficiency of cluster computing power management is ensured, and meanwhile, the functionality of cluster computing power management is greatly improved.

When cluster management is performed based on the first cluster power management result or the second cluster power management result, the management node of the first cluster power management result is determined, and then call management is timely performed based on the management node, so that the time for searching the optimal node for new power demand is shortened, and the replacement node is quickly found for call when the power resource of the existing power demand changes, and the use efficiency of cluster power can be greatly improved. And determining a second management node of the second cluster power management result, and further using power resources on the second management node, wherein the second management node is a node capable of calling for management, and further the accuracy of cluster power operation can be ensured when the power resource requirement is increased by the node to be managed.

Further, a flow chart of a cluster management method is provided for the present embodiment, referring to fig. 5, in this embodiment, after acquiring the related information of the management (that is, the input management requirement information is received), it is determined whether to perform the management in a single dimension based on the related information, that is, it is not determined that the management is performed directly after receiving the management request, but it is determined that the management is performed in a direction required to be performed in the request itself, for example, the management in a single dimension (only one management requirement exists in the management requirement information of a node, and the management requirement occupies almost all resources or almost all time of the node), the management in a multiple dimension (there are a plurality of management requirements in the management requirement information of a node, and there is a little free management resource in the node), and so on, and then it is determined that the management control is performed in different management modes. When the related information is determined to be single-dimensional management, the mode is determined to be single-dimensional management, and management is further carried out based on the related information and a single-dimensional management basis, namely, a management mode on a single dimension is carried out on a calculation management request at the moment, wherein the single-dimensional management basis is a management basis which is designed by a user in a single-dimensional management mode; otherwise, when the related information is determined not to be single-dimensional management, the mode is determined to be multi-dimensional management, and management is further performed based on the related information and multi-dimensional management basis, namely, a management mode in multi-dimensions is performed on the calculation management request at the moment, wherein the multi-dimensional management basis is a management basis for management design of a user in a plurality of dimension management modes. And the cluster power management result is determined based on power management demand information and preset management basis information according to different power management modes, and is managed not simply according to the used and unused power resources, but according to the related information of power management (during power resource use), so that the cluster power use efficiency is improved.

It should be noted that, the implementation is a cluster computing power management method executed after determining that the computing power resources are used by the node scheduling, the determining how to perform node scheduling based on the computing power demand is a cluster computing power scheduling method, the optimizing based on the change information of the real-time computing power resources is a cluster computing power optimizing method after scheduling, and the cluster computing power inertia method of the embodiment needs to be executed after optimizing or scheduling, so that the efficiency of subsequent management is improved, the computing power resources efficiently meet the computing power demands, the scheduling resources do not meet the demands due to the increase of the demand resources or the change of the using time, and the accuracy of cluster computing power control can be ensured.

In the embodiment, input power management requirement information is acquired, wherein the power management requirement information comprises node power use information and node power distribution information; determining a power management mode according to the node power use information, wherein the power management mode comprises a single-dimensional power management mode and a multi-dimensional power management mode; if the power management mode is the single-dimensional power management mode, determining a first cluster power management result according to the node power distribution information and preset single-dimensional power management basis information, and carrying out cluster power calling management according to the first cluster power management result and preset first management requirements; if the management mode of the computing power is the multi-dimensional management mode of the computing power, a second cluster computing power management result is determined according to the node computing power distribution information and the preset multi-dimensional management basis information, cluster computing power calling management is conducted according to the second cluster computing power management result, the management mode of the computing power is determined according to computing power management requirement information, in different computing power management modes, the cluster computing power management result is determined based on the computing power management requirement information and different preset management basis information, cluster management is conducted based on the cluster computing power management result, cluster computing power calling management is conducted in a single-dimensional computing power management mode, and in addition, when computing power nodes are more, only a plurality of computing power resources to be called are needed to be managed according to management in the prior art, the phenomenon that the computing power resources to be called are more is managed is generated, in the cluster computing power management mode, in different management modes, cluster computing power management is conducted based on the computing power management requirement information and different preset computing power management basis, the computing power is not needed to be called according to the cluster management result, and the computing power is not needed to be called by the computing power management information, and the cluster management is conducted according to the cluster management result is not needed to the cluster management information.

Further, based on the first embodiment of the cluster computing management method of the present application, a second embodiment of the cluster computing management method of the present application is provided, the node computing force usage information includes first computing force usage time information and first computing force remaining resources of a node to be managed, and the step of determining a computing force management mode according to the node computing force usage information includes:

Step S21, if the first computing power use time information meets the preset use duration and the first computing power residual resources meet the preset computing power residual resources, determining that the computing power management mode is a multi-dimensional computing power management mode;

step S22, if the first computing power usage time information does not meet the preset usage time length or the first computing power remaining resources do not meet the preset computing power remaining resources, determining that the computing power management mode is a single-dimensional computing power management mode.

In this embodiment, when determining the power management mode, the node power usage information may be directly determined, so as to determine the power management mode. The node computing power use information comprises first computing power use time information of the node to be managed and first computing power residual resources, wherein the first computing power use time information refers to the use time of computing power demands on the node to be managed, the use time is actually time information because a plurality of computing power demands can exist, the first computing power residual resources refer to the residual computing power resources of different time points on the node to be managed, and the first computing power residual resources are the residual computing power resources of different time points because the node to be managed possibly exists time-sharing use computing power resources. And when the first computing power use time information meets the preset use time length and the first computing power residual resources meet the preset computing power residual resources, determining that the computing power management mode is a multi-dimensional computing power management mode, otherwise, determining that the computing power management mode is a single-dimensional computing power management mode when the first computing power use time information does not meet the preset use time length or the first computing power residual resources do not meet the preset computing power residual resources. The preset use duration refers to a defined use duration, such as 2 hours, or the total use duration of the same computing resource. That is, whether the use time of each calculation force demand in the first calculation force use time information meets the preset use time length is detected, the preset use time length is not met, or the total use time length of each calculation force demand does not exceed the preset use time length, the calculation force residual resources are defined calculation force resource residual thresholds, and the preset calculation force residual resources are larger than the calculation force resource residual thresholds. And when the preset computing power residual resources are determined to be met and the preset using time length is determined to be met, the node is known to be actually managed in two dimensions, and then the time division and the residual resources are respectively managed, otherwise, when the preset computing power residual resources are determined to be met or the preset using time length is determined to be met, the node is known to be actually managed in one dimension, and then the time division or the residual resources are respectively managed. And then the management mode of the computing power is determined, and a mode management basis is provided for management of the computing power resources of the subsequent clusters.

Further, based on the first embodiment and the second embodiment of the cluster power management method of the present application, a third embodiment of the cluster power management method of the present application is provided, the node power distribution information includes a first power feature of the node to be managed, the power usage time information and a power demand, the single-dimensional power management basis information includes second usage time information, a second power feature and a second power remaining resource of each node, and the step of determining a first cluster power management result according to the node power distribution information and a preset single-dimensional power management basis information includes:

Step S301, if the first computing force usage time information does not meet the preset usage time length, determining a first target feature of the computing force demand in the first computing force features;

step S302, determining a first target node matched with the first target feature in the second computing power feature, and determining a first target residual resource of the first target node in the second computing power residual resources;

Step S303, determining a second target node of which the first target residual resource is matched with the demand resource of the computing power demand, and taking the first target node and the second target node as a first cluster computing power management result;

In the present embodiment, management is performed in two cases in the single-dimensional management mode. When the first calculation force using time information does not meet the preset using time (at this time, the remaining calculation force resources of the node do not call the calculation force resources due to the calculation force demand, but the remaining calculation force resources exist), a first target feature of the calculation force demand in the first calculation force features, that is, the calculation force features of the calculation force demand to be managed are determined as the first target feature, and the first calculation force features refer to the calculation force features corresponding to the calculation force demands in the calculation force nodes (calculation force nodes to be managed). In this case, a first target node that matches the first target feature may be determined from the second computing features of the other computing nodes, that is, a node that matches the computing feature may be determined from the other computing nodes as the first target feature, where the computing feature is the same as or close to the first target feature. And then, a first target node meeting the calculation force characteristics required by the calculation force node can be determined, and then, a first target residual resource of the first target node in second calculation force residual resources is determined, namely, whether the residual resources of the first target node can realize the calling of the calculation force requirements in the calculation force node or not is determined, wherein the second calculation force residual resources refer to the residual calculation force resources of each calculation force node in each time period, and at the moment, the calling of the whole time is ensured because the situation is that the first calculation force use time information does not meet the preset use duration. And then, a second target node, namely a node, which is larger than the demand resource of the calculation force demand, in the first target residual resource is used as a second target node, and the first target node and the second target node are used as first cluster calculation force management results at the moment, so that cluster calculation force management is conveniently carried out based on the first cluster calculation force management results when the first calculation force use time information does not meet the preset use time.

Step S304, if the first computing power residual resources do not meet the preset multi-computing power residual use resources, determining a second target feature of the computing power demand in the first computing power features;

Step S305, determining, at the second computing power feature, a third target node that matches the second target feature, and determining a first target usage time of the third target node in the second usage time information;

step S306, determining a fourth target node of which the first target use time is not coincident with the calculated force use time information, and taking the third target node and the fourth target node as a first cluster calculated force management result;

In this embodiment, when the first remaining computing power resources do not meet the preset remaining usage resources of multiple computing powers (at this time, the remaining computing power resources of the node call the computing power resources due to the computing power demand, but there are computing power resources that are not called and remain in full time), a second target feature of the computing power demand in the first computing power feature, that is, a computing power feature of the computing power demand that needs to be managed is determined as a second target feature, where the second computing power feature refers to a computing power feature corresponding to each computing power demand in the computing power node (computing power node to be managed). In this case, a third target node that matches the second target feature may be determined from the second computing features of the other computing nodes, that is, a node that matches the computing feature may be determined from the other computing nodes as the second target feature, where the computing feature is the same as or close to the first target feature. And then a third target node meeting the calculation force characteristics required by the calculation force node can be determined, and then a first target use time of the third target node in the second use time information is determined, namely whether the calling of the calculation force requirements in the calculation force node can be realized on the residual resource time of the third target node is determined, wherein the first target use time refers to the residual calculation force resources of each calculation force node in each time period, and the calling of the time division is ensured at the moment because the situation is that the residual first calculation force resources do not meet the preset multi-calculation force residual use resources. And then, a fourth target node, in which the first target usage time and the calculation force usage time information are not coincident, namely, a node of a node resource, in which the first target usage time and the calculation force usage time information are not coincident, is used as a fourth target node, and at this time, the third target node and the fourth target node are used as a first cluster calculation force management result, so that cluster calculation force management is conveniently performed based on the first cluster calculation force management result when the residual first calculation force resource does not meet the preset multi-calculation force residual usage resource.

Further, the first management requirement includes a temporary call of the target calculation requirement and a requirement call of the new calculation requirement, and the step of performing cluster calculation power call management according to the first cluster calculation power management result and a preset first management requirement includes:

Step S311, if the preset first management requirement is the temporary call, determining a third target feature of the target calculation force requirement of the temporary call in the first calculation force features;

Step S312, determining a first target calling node corresponding to the target power demand, determining a temporary calling node in the first target calling node, wherein target features in the first target calling node are matched with the third target features, and performing cluster power calling management on the temporary calling node, wherein the first target calling node comprises the fourth target node and the second target node;

In this embodiment, after determining the first cluster computing management result, cluster computing power call management is performed based on the preset first management requirement and the first cluster computing management result. Since the first management requirement includes temporary call of the target calculation force requirement and requirement call of the new calculation force requirement, and further when the preset first management requirement is temporary call, determining a third target feature of the temporary call target calculation force requirement in the first calculation force feature, wherein the temporary call refers to a situation that the existing calculation force requirement increases or decreases the required calculation force resource, only the situation of increase is discussed here, the decrease is not needed to be processed, and the third target feature refers to a calculation force feature of the temporary call target calculation force requirement in the first calculation force feature. And determining a third target feature, and further determining a first target call node corresponding to the target calculation power demand, wherein the first target call node comprises the fourth target node and the second target node, namely two target nodes for node scheduling and time sharing scheduling, determining a node matched with the target feature in the first target call node for the third target feature as a temporary call node, and finally performing cluster calculation power call management on the temporary call node, thereby realizing cluster management. The first target calling node corresponding to the target power demand refers to the standby node (i.e., the fourth target node for standby at the time-sharing point and the second target node for standby at the node-sharing point) of each power demand in the power-computing node determined in the previous embodiment, so that when the power-computing resource of the power-computing demand increases, the standby node can be directly called, and further, the management efficiency can be greatly improved. The target feature in the first target call node means that the computing power resource of the computing power demand is increased, and the computing power feature corresponding to the computing power demand can further realize rapid time-sharing or node-division scheduling management, so that the overall cluster computing power management scheduling efficiency is greatly improved.

Step S313, if the preset first management requirement is the requirement call, determining a fourth target feature of the new requirement feature in the first calculation power feature;

step S314, determining a second target call node corresponding to the new calculation force demand, and determining a first demand call node with target features matched with the fourth target features in the second target call node, wherein the first demand call node comprises the third target node and the first target node;

Step S315, determining a required resource size of the new calculation force requirement, determining a second requirement calling node satisfying the required resource size in the first requirement calling node, and performing cluster calculation force calling management on the second requirement calling node.

In this embodiment, when the preset first management requirement is a requirement call, a fourth target feature of the target calculation force requirement of the requirement call in the first calculation force features is determined, where the requirement call refers to a situation that the new calculation force requirement needs calculation force resources, and the fourth target feature refers to a calculation force feature of the new calculation force requirement of the requirement call in the first calculation force features. And determining a fourth target feature, further determining a second target call node corresponding to the target calculation power demand, wherein the second target call node comprises a third target node and the first target node, namely two target nodes for node scheduling and time sharing scheduling, determining a node matched with the target feature in the second target call node for the fourth target feature as a first demand call node, and finally performing cluster calculation power call management based on the first demand call node, thereby realizing cluster management. And then the node of the new calculation power demand dispatching is rapidly determined at the moment, and then the second demand calling node meeting the demand resource size is determined in the first demand calling node by determining the demand resource size of the new calculation power demand, so that cluster calculation power calling management can be carried out on the second demand calling node, and rapid cluster dispatching is realized. The second target calling node corresponding to the new calculation force demand refers to a standby node of the new calculation force demand (namely, a third target node for standby at a time-sharing point and a first target node for standby at a partition node), so that when the new calculation force demand is input, the standby node can be directly called, and the management efficiency can be greatly improved. The target features in the second target calling node are the computing power features corresponding to the computing power resources of the new computing power requirements, so that the rapid time-sharing or node-division scheduling management can be realized, and the overall cluster computing power management scheduling efficiency is greatly improved.

If it is determined that there is no new result (i.e., a plurality of nodes are screened based on conditions, but there is no result) after a certain step, the technical solution of this embodiment may not be continuously executed, or the execution is stopped after waiting for a delay time, or when the priority is higher, the invocation of the computing resource at the special position is directly stopped, so as to directly satisfy the computing requirement of this time. The above is only one case where the management scheduling result cannot be specified, and the steps of the method may be executed, and are not limited thereto.

Referring to fig. 4, an exemplary illustration of a scenario of the cluster computing management system of the present application is shown in fig. 4, where the cluster computing management controller is connected to a plurality of management nodes, and when the management node 1 sends out information to be managed, the information of each management node is managed in the cluster computing management controller, and at this time, the information of the plurality of management nodes may be determined based on the information to be managed, and then the management/scheduling process 1-n of each management node may be sent to each management node, so as to complete the cluster computing management, and may be managed in a computing resource use process, so as to improve the computing efficiency of the cluster computing use. The cluster power management method can acquire respective related power resource information from a node server corresponding to each management node, wherein the power resource information comprises the used power resource, the power resource characteristics, the service time of the power resource (namely, power management demand information) and the like of the own node, and meanwhile, the related power resource information (comprising second service time information, second power characteristics and second power residual resources of each node) of each node is sent to a cluster power management controller for management processing. When the respective nodes have the requirement to be managed, the relevant computing power resource information and the requirement information of the nodes are included in the requirement information, such as the use time, the real-time requirement computing power resource, the total residual computing power resource and the like of the computing power resource of the node to be managed, so that the cluster computing power management method of the embodiment is executed in the cluster computing power management controller based on the relevant computing power resource information of the respective nodes and the requirement information of the node to be managed, the cluster computing power management result is obtained, the cluster computing power management result is sent to the server of each node, the server controls the respective nodes to complete the operations of calling, time multiplexing and the like of the computing power resource in the node, and meanwhile, when the new computing power requirement arrives, the computing power resource can be rapidly distributed based on the cluster computing power management result. Different information can be selectively acquired in different modes, instead of directly acquiring all relevant computing power resource information of each node, so that management efficiency can be greatly improved, and meanwhile, cluster computing power management accuracy is improved.

Further, based on the first embodiment, the second embodiment and the third embodiment of the present application, a fourth embodiment of the present application is a cluster power management method, where node power distribution information includes the power demand and total power resources of the nodes to be managed, and the step of determining a second cluster power management result according to the node power distribution information and preset multi-dimensional power management basis information includes:

Step S401, determining a first calculation force demand meeting a preset first condition in the calculation force demands, wherein the preset first condition comprises the calculation force demand that the first calculation force use time information of the calculation force demand meets a preset use time length;

Step S402, determining a second computing force demand meeting a preset second condition in the computing force demands, wherein the preset first condition comprises that the first computing force residual resource of the computing force demand meets the computing force demand of a preset computing force residual resource;

Step S403 of determining an computational power resource sum value between a second demand resource of the second computational power demand and a first demand resource of the first computational power demand, and determining a remaining computational power resource between the total computational power resource and the computational power resource sum value;

In this embodiment, in the multidimensional management mode, management is performed in two cases, and there is a difference from the single-dimensional management mode, and management based on the occupation ratio of the multidimensional is required. By determining a first calculation force demand satisfying a preset first condition and a second calculation force demand satisfying a preset second condition among the calculation force demands, respectively, wherein the preset first condition includes the calculation force demand of which the first calculation force use time information satisfies a preset use time length, that is, the calculation force demand which can be managed in a time-sharing manner as described in the previous embodiment, the preset first condition includes the calculation force demand of which the first calculation force remaining resources satisfy the preset calculation force remaining resources, that is, the calculation force demand which can be managed in a node-by-node manner as described in the previous embodiment, that is, the resource quantity of the calculation force resources which can be managed in a time-sharing manner and in a node-by-node manner in the multidimensional management mode is determined. Further, the computing power resource and the value between the second demand resource of the second computing power demand and the first demand resource of the first computing power demand are determined, that is, the sum value of the computing power resource is used as the computing power resource and the value by the whole computing power node, instead of using the remaining computing power resource, because the remaining computing power resource is different at different times in the embodiment, the determination of the computing power resource and the value is required to be more accurate. And then determining the total computing power resource and the remaining computing power resource between the computing power resource and the value, and further determining the remaining computing power resource under the condition of the most use, wherein the total computing power resource refers to the total computing power resource of the computing power node. At this time, because the first calculation force requirement meeting the preset first condition and the second calculation force requirement meeting the preset second condition are determined, the situation that time sharing calling cannot be performed is considered, for example, time sharing calling of the calculation force requirement S1 and the calculation force requirement S2 is considered, and at this time, the determined second calculation force requirement is the sum of the resources of the calculation force requirement S1 and the calculation force requirement S2, so that management errors caused by incapability of time sharing of the calculation force resources can be avoided, and further, the accuracy of cluster calculation force management can be greatly improved.

Step S404, if the remaining power resources are greater than a preset remaining threshold, respectively taking the first demand resources and the second demand resources as the power demand, and executing the step of determining a first cluster power management result according to the node power distribution information and the preset single-dimensional power management basis information;

step S405, if the remaining power resources are smaller than or equal to a preset remaining threshold, determining a size relationship between the second demand resources and the first demand resources, and determining a second cluster power management result according to the size relationship, the node power distribution information and preset multi-dimensional power management basis information.

In this embodiment, after the remaining computing power resources are determined, the case-division processing is performed based on the remaining computing power resources. When the remaining computing power resource is larger than a preset remaining threshold, the first demand resource and the second demand resource are respectively used as computing power demands, and the step of determining a first cluster computing power management result according to the node computing power distribution information and the preset single-dimensional computing power management basis information is executed. That is, the first demand resource capable of being scheduled in a time-sharing manner and the second demand resource capable of being scheduled in a node-division manner are respectively executed in a single-dimensional power management mode, that is, the single-dimensional power management mode can execute only one or two-dimensional management simultaneously. And further, the management of the multidimensional management mode can be realized. That is, the above case is that it is determined that the satisfaction of the remaining threshold value greater than the preset remaining threshold value (the threshold value set by the user, if at least the remaining power resource of A2 is to exist, the A2 is set to the preset remaining threshold value) is satisfied in the multidimensional management mode, and then the first demand resource and the second demand resource are determined to perform the one-dimensional power management mode, respectively. When the remaining computing power resource is smaller than or equal to a preset remaining threshold value, the size relation between the second demand resource and the first demand resource is determined, and then a second cluster computing power management result is determined based on the size relation, the node computing power distribution information and the preset multi-dimensional computing power management basis information, namely management scheduling in a multi-dimensional management mode is further realized in the second case based on the size relation, the node computing power distribution information and the preset multi-dimensional computing power management basis information, so that the functionality of cluster computing power management can be improved, meanwhile, targeted management is carried out aiming at different situations, and the specificity of cluster computing power management on management in different situations can be improved.

Further, the size relationship includes a first relationship that the second demand resource is greater than the first demand resource and a second relationship that the second demand resource is less than or equal to the first demand resource, the node calculation power distribution information further includes the first calculation power feature and the calculation power usage time information, the multidimensional calculation power management basis information includes the second usage time information, the second calculation power feature and the second calculation power remaining resource, and the step of determining a second cluster calculation power management result according to the size relationship, the node calculation power distribution information and the preset multidimensional calculation power management basis information includes:

Step S4051, if the size relationship is the first relationship, determining a third target feature of the first demand resource in the first calculation power feature;

step S4052, determining a fifth target node matched with the third target feature in the second computing power feature, and determining a second target usage time of the fifth target node in the second usage time information;

step S4053, determining a sixth target node, where the second target usage time does not coincide with the power usage time information of the first demand resource, and using the sixth target node as a second cluster power management result;

In this embodiment, when the multidimensional management mode is further required, the multidimensional management is further performed based on the first relationship and the second relationship by dividing the size relationship into the first relationship in which the second demand resource is greater than the first demand resource and the second relationship in which the second demand resource is less than or equal to the first demand resource. When the magnitude relation is the first relation, a third target feature of the first demand resource in the first calculation force features is determined, and then a fifth target node matched with the third target feature is determined in the second calculation force features, namely when the magnitude relation is the first relation, the second demand resource is used as the only calculation force demand of the calculation force resource capable of calling the calculation force node, the calculation force demand of the first demand resource is then called out of the calculation force node, normal management of the second demand resource is further ensured (the phenomenon that the calculation force node collapses due to the increase of the calculation force resource is avoided), wherein the third target feature refers to the calculation force feature corresponding to each calculation force demand in the first calculation force feature in the first demand resource, and the first calculation force feature refers to the calculation force feature of all calculation force demands in the calculation force node. And further determining a fifth target node matched with the third target feature in the second computing force features of the computing force features of all the nodes, wherein the fifth target node is a node matched with the third target feature (the computing force features are the same or similar) in the computing force features of all the nodes. And determining a second target use time in second use time information of the fifth target node in which the use time of each calculation force demand in all the nodes is recorded, wherein the second target use time refers to the use time of each calculation force demand in the fifth target node. And finally, determining a sixth target node with non-coincident second target use time and first demand resource calculation force use time information, namely determining a node which can call calculation force demand corresponding to the first demand resource as the sixth target node, wherein the non-coincident node can be used for time-sharing scheduling, or directly determining a node with calculation force resource larger than the first demand resource as the sixth target node, and further, taking the sixth target node as a second cluster calculation force management result to perform subsequent cluster calculation force management, thereby ensuring the accuracy of cluster calculation force management, and simultaneously improving the efficiency of cluster calculation force management.

Step S4054, if the size relationship is the second relationship, determining a fourth target feature of the second demand resource in the first calculation power feature;

Step S4055, determining, at the second computing power feature, a seventh target node that matches the fourth target feature, and determining a second target remaining resource of the seventh target node among the second computing power remaining resources;

Step S4056, determining an eighth target node where the second target remaining resource matches the second demand resource, and taking the eighth target node as a second cluster power management result.

In this embodiment, when the second relationship is the second relationship, the first demand resource is used as the only power demand of the power resource capable of calling the power node, so that the power demand of the second demand resource is called out of the power node, and normal management of the first demand resource is further ensured (the phenomenon that the power node collapses due to the increase of the power resource is avoided), wherein the fourth target feature is the power feature corresponding to each power demand in the second demand resource in the first power feature, and the first power feature is the power feature of all power demands in the power node. And then determining a seventh target node matched with the fourth target feature in the second computing force features of the computing force features of all the nodes, wherein the seventh target node is the node matched with the fourth target feature (the computing force feature is the same or similar) in the computing force features of all the nodes. And determining a second target residual resource in the second residual resources of the residual power resources of the power demands in all the nodes, wherein the second target residual resource refers to the residual power resource in the seventh target node. Finally, an eighth target node, in which the second target remaining resource matches (is greater than) the second demand resource, may be determined, that is, a node, in which the computing power demand call corresponding to the second demand resource may be determined as the eighth target node. And the eighth target node can be further used as a second cluster computing management result to perform subsequent cluster computing management, so that the accuracy of cluster computing management can be ensured, and meanwhile, the efficiency of cluster computing management is improved. The step of performing cluster computing power calling management according to the second cluster computing power management result comprises at least one of the following steps:

Step S411, performing cluster computing power call management on the second required resource based on the eighth target node;

Step S412, performing cluster computing power call management on the first demand resource based on the sixth target node.

In this embodiment, after the eighth target node and the sixth target node are determined, the second demand resource is scheduled to the eighth target node or the first demand resource is scheduled to the sixth target node, so that the efficiency of subsequent cluster computing power management can be improved, the phenomenon of computing power collapse can be avoided, and the efficiency of cluster computing power management can be improved.

Further, based on the first, second, third and fourth embodiments of the present application, a fifth embodiment of the present application is provided, and before the step of obtaining the input power management requirement information, at least one of the following is included:

step S11, real-time computing power resource information of the nodes is sequentially obtained based on a preset timing threshold to serve as computing power management requirement information;

Step S12, acquiring real-time computing power resource information of a node, and taking the real-time computing power resource information as computing power management demand information when the real-time computing power resource information meets a preset change condition, wherein the change condition comprises that computing power resources of the node are scheduled or residual computing power resources are changed.

In this embodiment, when acquiring the power management requirement information, two acquisition modes may be adopted. The method comprises the steps of sequentially acquiring real-time computing power resource information of nodes based on a preset timing threshold as computing power management requirement information, wherein the timing threshold is a time value defined by a user, namely, each node is managed based on a fixed time length, and the real-time computing power resource information is real-time computing power related information of the node and at least comprises node computing power use information and node computing power distribution information. And the other is to acquire the real-time computing power resource information of the node, and then take the real-time computing power resource information as computing power management demand information when the real-time computing power resource information meets a preset change condition, wherein the change condition comprises that computing power resources of the node are scheduled or residual computing power resources are changed, that is, each node is monitored, and then the computing power resource information of the monitored node is taken as computing power management demand information when the computing power resources of the node are changed, such as scheduled or residual computing power resources are changed. The method comprises the steps of detecting in real time, detecting the change of the computing power resources of the nodes or acquiring the computing power resources based on preset time, acquiring computing power management requirement information of the nodes to be managed, judging the mode of the computing power management requirement information, and continuously executing the step of acquiring the input computing power management requirement information when the mode is determined to be neither a single-dimensional computing power management mode nor a multi-dimensional computing power management mode. And further, the demand information of the power management is acquired pertinently, and a management basis can be provided for the power management of the cluster.

The present application also provides a schematic block diagram of a cluster computing management device, referring to fig. 3, where the cluster computing management device includes:

The information acquisition module A10 is used for acquiring input power management demand information, wherein the power management demand information comprises node power use information and node power distribution information;

the mode judging module A20 is used for determining a power management mode according to the node power use information, wherein the power management mode comprises a single-dimension power management mode and a multi-dimension power management mode;

The first management module a30 is configured to determine a first cluster power management result according to the node power distribution information and preset single-dimensional power management basis information if the power management mode is the single-dimensional power management mode, and perform cluster power calling management according to the first cluster power management result and a preset first management requirement;

And the second management module a40 is configured to determine a second cluster power management result according to the node power distribution information and preset multi-dimensional power management basis information if the power management mode is the multi-dimensional power management mode, and perform cluster power calling management according to the second cluster power management result.

Optionally, the mode determining module a20 is further configured to:

If the first computing power using time information meets the preset using time length and the first computing power residual resources meet the preset computing power residual resources, determining that the computing power management mode is a multi-dimensional computing power management mode;

And if the first computing power using time information does not meet the preset using time length or the first computing power residual resources do not meet the preset computing power residual resources, determining that the computing power management mode is a single-dimension computing power management mode.

Optionally, the first management module a30 is further configured to:

If the first calculation force using time information does not meet the preset using time length, determining a first target feature of the calculation force demand in the first calculation force features;

Determining a first target node matched with the first target feature in the second computing power feature, and determining a first target residual resource of the first target node in the second computing power residual resources;

Determining a second target node of which the first target residual resources are matched with the demand resources of the computing power demand, and taking the first target node and the second target node as a first cluster computing power management result;

if the first computing power residual resources do not meet the preset multi-computing power residual use resources, determining a second target feature of the computing power demand in the first computing power features;

Determining a third target node matched with the second target feature in the second computing force feature, and determining a first target use time of the third target node in the second use time information;

And determining a fourth target node of which the first target use time is not coincident with the calculated force use time information, and taking the third target node and the fourth target node as a first cluster calculated force management result.

Optionally, the first management module a30 is further configured to:

if the preset first management requirement is the temporary call, determining a third target characteristic of the temporary call target calculation force requirement in the first calculation force characteristic;

determining a first target calling node corresponding to the target computing power demand, determining a temporary calling node in which target features in the first target calling node are matched with the third target features, and performing cluster computing power calling management on the temporary calling node, wherein the first target calling node comprises the fourth target node and the second target node;

If the preset first management requirement is the requirement call, determining a fourth target feature of the requirement feature of the new calculation power requirement in the first calculation power feature;

Determining a second target call node corresponding to the new calculation force demand, and determining a first demand call node in which target features in the second target call node are matched with the fourth target features, wherein the first demand call node comprises the third target node and the first target node;

and determining the size of the required resources of the new calculation force requirement, determining a second requirement calling node meeting the size of the required resources in the first requirement calling node, and carrying out cluster calculation force calling management on the second requirement calling node.

Optionally, the second management module a40 is further configured to:

determining a first calculation force demand meeting a preset first condition in the calculation force demands, wherein the preset first condition comprises the calculation force demand that the first calculation force use time information of the calculation force demand meets a preset use time length;

Determining a second one of the computational power demands that satisfies a preset second condition, wherein the preset first condition includes a computational power demand that the first computational power remaining resources of the computational power demand satisfy a preset computational power remaining resources;

Determining an computational power resource sum value between a second demand resource of the second computational power demand and a first demand resource of the first computational power demand, and determining a remaining computational power resource between the total computational power resource and the computational power resource sum value;

If the residual power resource is larger than a preset residual threshold value, respectively taking the first demand resource and the second demand resource as the power demand, and executing the step of determining a first cluster power management result according to the node power distribution information and preset single-dimensional power management basis information;

and if the residual computational power resource is smaller than or equal to a preset residual threshold value, determining the size relation between the second demand resource and the first demand resource, and determining a second cluster computational power management result according to the size relation, the node computational power distribution information and preset multi-dimensional computational power management basis information.

Optionally, the second management module a40 is further configured to:

If the size relation is the first relation, determining a third target feature of the first demand resource in the first calculation force feature;

determining a fifth target node matched with the third target feature in the second computing force feature, and determining a second target use time of the fifth target node in the second use time information;

Determining a sixth target node, the second target use time of which is not coincident with the computing power use time information of the first required resource, and taking the sixth target node as a second cluster computing power management result;

If the size relation is the second relation, determining a fourth target feature of the second required resource in the first calculation force feature;

Determining a seventh target node matched with the fourth target feature in the second computing power feature, and determining a second target residual resource of the seventh target node in second computing power residual resources;

Determining an eighth target node of which the second target residual resources are matched with the second required resources, and taking the eighth target node as a second cluster calculation management result;

the step of performing cluster computing power call management according to the second cluster computing power management result comprises at least one of the following steps:

performing cluster computing power calling management on the second demand resource based on the eighth target node;

and carrying out cluster computing power calling management on the first demand resource based on the sixth target node.

Optionally, the information acquisition module a10 is further configured to:

sequentially acquiring real-time computing power resource information of the nodes based on a preset timing threshold value to serve as computing power management demand information;

acquiring real-time computing power resource information of a node, and taking the real-time computing power resource information as computing power management demand information when the real-time computing power resource information meets a preset change condition, wherein the change condition comprises that computing power resources of the node are scheduled or residual computing power resources are changed.

The application also provides cluster power management equipment.

The device of the application comprises: the system comprises a memory, a processor and a cluster computing management program stored on the memory and capable of running on the processor, wherein the cluster computing management program realizes the steps of the cluster computing management method when being executed by the processor.

The application also provides a storage medium.

The storage medium of the present application has stored thereon a cluster power management program which, when executed by a processor, implements the steps of the cluster power management method as described above.

The method implemented when the cluster computing management program running on the processor is executed may refer to various embodiments of the cluster computing management method of the present application, which are not described herein again.

It should be noted that, in this document, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or system that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or system. Without further limitation, an element defined by the phrase "comprising one … …" does not exclude the presence of other like elements in a process, method, article, or system that comprises the element.

The foregoing embodiment numbers of the present application are merely for the purpose of description, and do not represent the advantages or disadvantages of the embodiments.

From the above description of embodiments, it will be clear to a person skilled in the art that the above embodiment method may be implemented by means of software plus a necessary general hardware platform, but may of course also be implemented by means of hardware, but in many cases the former is a preferred embodiment. Based on this understanding, the technical solution of the present application may be embodied essentially or in a part contributing to the prior art in the form of a software product stored in a storage medium (e.g. ROM/RAM, magnetic disk, optical disk) as described above, comprising instructions for causing a terminal device (which may be a mobile phone, a computer, a server, or a network device, etc.) to perform the method according to the embodiments of the present application.

The foregoing description is only of the preferred embodiments of the present application, and is not intended to limit the scope of the application, but rather is intended to cover any equivalents of the structures or equivalent processes disclosed herein or in the alternative, which may be employed directly or indirectly in other related arts.

Claims

1. A cluster power management method, characterized in that the cluster power management method comprises the steps of:

Determining a power management mode according to the node power usage information, wherein the power management mode comprises a single-dimensional power management mode and a multi-dimensional power management mode, the node power usage information comprises first power usage time information and first power remaining resources of a node to be managed, and the step of determining the power management mode according to the node power usage information comprises the following steps:

If the first computing power using time information meets the preset using time length and the first computing power residual resources meet the preset computing power residual resources, determining that the computing power management mode is a multi-dimensional computing power management mode; if the first computing power using time information does not meet the preset using time length or the first computing power residual resources do not meet the preset computing power residual resources, determining that the computing power management mode is a single-dimensional computing power management mode;

If the power management mode is the single-dimensional power management mode, determining a first cluster power management result according to the node power distribution information and preset single-dimensional power management basis information, wherein the node power distribution information comprises first power characteristics of the nodes to be managed, power use time information and power requirements, the single-dimensional power management basis information comprises second use time information, second power characteristics and second power residual resources of each node, and the step of determining the first cluster power management result according to the node power distribution information and the preset single-dimensional power management basis information comprises the following steps:

If the first calculation force using time information does not meet the preset using time length, determining a first target feature of the calculation force demand in the first calculation force features; determining a first target node matched with the first target feature in the second computing power feature, and determining a first target residual resource of the first target node in the second computing power residual resources; determining a second target node of which the first target residual resources are matched with the demand resources of the computing power demand, and taking the first target node and the second target node as a first cluster computing power management result; if the first computing power residual resources do not meet the preset multi-computing power residual use resources, determining a second target feature of the computing power demand in the first computing power features; determining a third target node matched with the second target feature in the second computing force feature, and determining a first target use time of the third target node in the second use time information; determining a fourth target node, of which the first target usage time is not coincident with the calculation force usage time information, using the third target node and the fourth target node as a first cluster calculation force management result, and performing cluster calculation force calling management according to the first cluster calculation force management result and a preset first management requirement, wherein the first management requirement comprises temporary calling of a target calculation force requirement and requirement calling of a new calculation force requirement, and performing cluster calculation force calling management according to the first cluster calculation force management result and the preset first management requirement, and the method comprises the following steps:

If the preset first management requirement is the temporary call, determining a third target characteristic of the temporary call target calculation force requirement in the first calculation force characteristic; determining a first target calling node corresponding to the target computing power demand, determining a temporary calling node in which target features in the first target calling node are matched with the third target features, and performing cluster computing power calling management on the temporary calling node, wherein the first target calling node comprises the fourth target node and the second target node; if the preset first management requirement is the requirement call, determining a fourth target feature of the requirement feature of the new calculation power requirement in the first calculation power feature; determining a second target call node corresponding to the new calculation force demand, and determining a first demand call node in which target features in the second target call node are matched with the fourth target features, wherein the first demand call node comprises the third target node and the first target node; determining the size of a required resource of the new calculation force requirement, determining a second requirement calling node meeting the size of the required resource in the first requirement calling node, and carrying out cluster calculation force calling management on the second requirement calling node;

If the power management mode is the multi-dimensional power management mode, determining a second cluster power management result according to the node power distribution information and preset multi-dimensional power management basis information, and performing cluster power calling management according to the second cluster power management result, wherein the node power distribution information comprises the power demand and total power resources of the nodes to be managed, and the step of determining the second cluster power management result according to the node power distribution information and the preset multi-dimensional power management basis information comprises the following steps:

Determining a first calculation force demand meeting a preset first condition in the calculation force demands, wherein the preset first condition comprises the calculation force demand that the first calculation force use time information of the calculation force demand meets a preset use time length; determining a second one of the computational power demands that satisfies a preset second condition, wherein the preset first condition includes a computational power demand that the first computational power remaining resources of the computational power demand satisfy a preset computational power remaining resources; determining an computational power resource sum value between a second demand resource of the second computational power demand and a first demand resource of the first computational power demand, and determining a remaining computational power resource between the total computational power resource and the computational power resource sum value; if the residual power resource is larger than a preset residual threshold value, respectively taking the first demand resource and the second demand resource as the power demand, and executing the step of determining a first cluster power management result according to the node power distribution information and preset single-dimensional power management basis information; and if the residual computational power resource is smaller than or equal to a preset residual threshold value, determining the size relation between the second demand resource and the first demand resource, and determining a second cluster computational power management result according to the size relation, the node computational power distribution information and preset multi-dimensional computational power management basis information.

2. The clustered power management method of claim 1 wherein the size relationship includes a first relationship in which the second demand resource is greater than the first demand resource and a second relationship in which the second demand resource is less than or equal to the first demand resource, the node power distribution information further includes the first power feature and the power usage time information, the multidimensional power management basis information includes the second usage time information, the second power feature and the second power remaining resource, and the step of determining a second clustered power management result from the size relationship, the node power distribution information and a preset multidimensional power management basis information includes:

3. The clustered power management method of any one of claims 1-2 wherein prior to the step of obtaining the input power management demand information, at least one of:

4. A clustered power management apparatus, the clustered power management apparatus comprising:

The mode judging module is configured to determine a power management mode according to the node power usage information, where the power management mode includes a single-dimensional power management mode and a multi-dimensional power management mode, the node power usage information includes first power usage time information and first power remaining resources of a node to be managed, and the step of determining the power management mode according to the node power usage information includes: if the first computing power using time information meets the preset using time length and the first computing power residual resources meet the preset computing power residual resources, determining that the computing power management mode is a multi-dimensional computing power management mode; if the first computing power using time information does not meet the preset using time length or the first computing power residual resources do not meet the preset computing power residual resources, determining that the computing power management mode is a single-dimensional computing power management mode;

The first management module is configured to determine a first cluster power management result according to the node power distribution information and preset single-dimensional power management basis information if the power management mode is the single-dimensional power management mode, where the node power distribution information includes a first power feature of a node to be managed, the power usage time information, and a power demand, the single-dimensional power management basis information includes second usage time information, second power feature, and second power remaining resources of each node, and the step of determining the first cluster power management result according to the node power distribution information and the preset single-dimensional power management basis information includes:

The second management module is configured to determine a second cluster power management result according to the node power distribution information and preset multi-dimensional power management basis information if the power management mode is the multi-dimensional power management mode, and perform cluster power calling management according to the second cluster power management result, where the node power distribution information includes the power demand and total power resources of the node to be managed, and the step of determining the second cluster power management result according to the node power distribution information and the preset multi-dimensional power management basis information includes:

5. A clustered power management apparatus, the clustered power management apparatus comprising: a memory, a processor and a cluster computing management program stored on the memory and executable on the processor, which when executed by the processor implements the steps of the cluster computing management method of any one of claims 1 to 3.

6. A storage medium having stored thereon a cluster power management program which when executed by a processor performs the steps of the cluster power management method of any of claims 1 to 3.