CN116684421A - Service deployment method, management platform, equipment and medium of hybrid cloud platform - Google Patents

Abstract

The disclosure provides a service deployment method, a management platform, equipment and a medium of a hybrid cloud platform, and relates to the technical fields of cloud computing, hybrid cloud and the like. The specific implementation scheme comprises the following steps: acquiring a service deployment request of a user; the service deployment request carries functional information of a target model to be deployed and service deployment information of the target model; acquiring the residual quantity of each type of computing power resource on each node in each cluster created in advance for the user in a hybrid cloud platform; and deploying the services of the target model in the hybrid cloud platform based on the residual amounts of computing power resources of each type of each node in each cluster of the hybrid cloud platform and the service deployment information of the target model. According to the method and the device, resources in the hybrid cloud platform can be fully utilized, and the deployment efficiency of the services of the model in the hybrid cloud platform can be effectively improved.

Description

Service deployment method, management platform, equipment and media of hybrid cloud platform

technical field

The present disclosure relates to the field of computer technology, specifically to the technical fields of cloud computing and hybrid cloud, and in particular to a service deployment method, management platform, equipment and media of a hybrid cloud platform.

Background technique

The hybrid cloud platform integrates public cloud and private cloud, and is the main model and development direction of cloud computing in recent years.

In the prior art, the public cloud and the private cloud in the hybrid cloud platform are independent of each other. When deploying services in the public cloud and private cloud, you need to create clusters in the public cloud and private cloud respectively. Then deploy services in public cloud or private cloud respectively.

Contents of the invention

The present disclosure provides a hybrid cloud platform service deployment method, management platform, equipment and media.

According to an aspect of the present disclosure, a service deployment method of a hybrid cloud platform is provided, including:

Obtain the user's service deployment request; the service deployment request carries the function information of the target model to be deployed and the service deployment information of the target model;

Obtaining the remaining amount of each type of computing resources on each node in each cluster pre-created for the user in the hybrid cloud platform;

Deploy the service of the target model on the hybrid cloud platform based on the remaining amount of each type of computing resources of each node in each cluster of the hybrid cloud platform and the service deployment information of the target model.

According to another aspect of the present disclosure, a hybrid cloud management platform is provided, including:

A service deployment acquisition module, configured to acquire a user's service deployment request; the service deployment request carries the function information of the target model to be deployed and the service deployment information of the target model;

The remaining resource obtaining module is used to obtain the remaining amount of various types of computing power resources on each node in each cluster created in advance for the user in the hybrid cloud platform;

A service deployment module, configured to deploy the target in the hybrid cloud platform based on the remaining amount of each type of computing resources of each node in each cluster of the hybrid cloud platform and the service deployment information of the target model Model service.

According to still another aspect of the present disclosure, an electronic device is provided, including:

at least one processor; and

a memory communicatively coupled to the at least one processor; wherein,

The memory stores instructions executable by the at least one processor, the instructions are executed by the at least one processor, so that the at least one processor can perform the above aspects and any possible implementation way of way.

According to yet another aspect of the present disclosure, there is provided a non-transitory computer-readable storage medium storing computer instructions, the computer instructions are used to make the computer execute the method of the above aspect and any possible implementation manner .

According to yet another aspect of the present disclosure, a computer program product is provided, including a computer program, and when the computer program is executed by a processor, the above aspect and the method of any possible implementation manner are implemented.

According to the technology disclosed in the present disclosure, the deployment efficiency of model services in the hybrid cloud platform can be effectively improved.

It should be understood that what is described in this section is not intended to identify key or important features of the embodiments of the present disclosure, nor is it intended to limit the scope of the present disclosure. Other features of the present disclosure will be readily understood through the following description.

Description of drawings

The accompanying drawings are used to better understand the present solution, and do not constitute a limitation to the present disclosure. in:

FIG. 1 is a schematic diagram according to a first embodiment of the present disclosure;

FIG. 2 is a schematic diagram according to a second embodiment of the present disclosure;

Fig. 3 is a schematic diagram according to a third embodiment of the present disclosure;

FIG. 4 is a management diagram of a hybrid cloud management platform provided by the present disclosure;

FIG. 5 is a schematic diagram according to a fourth embodiment of the present disclosure;

FIG. 6 is a schematic diagram according to a fifth embodiment of the present disclosure;

FIG. 7 is a block diagram of an electronic device used to implement the method of an embodiment of the present disclosure.

Detailed ways

Exemplary embodiments of the present disclosure are described below in conjunction with the accompanying drawings, which include various details of the embodiments of the present disclosure to facilitate understanding, and they should be regarded as exemplary only. Accordingly, those of ordinary skill in the art will recognize that various changes and modifications of the embodiments described herein can be made without departing from the scope and spirit of the disclosure. Also, descriptions of well-known functions and constructions are omitted in the following description for clarity and conciseness.

Apparently, the described embodiments are some of the embodiments of the present disclosure, but not all of them. Based on the embodiments in the present disclosure, all other embodiments obtained by persons of ordinary skill in the art without creative efforts fall within the protection scope of the present disclosure.

It should be noted that the terminal devices involved in the embodiments of the present disclosure may include but not limited to mobile phones, personal digital assistants (Personal Digital Assistant, PDA), wireless handheld devices, tablet computers (TabletComputer) and other smart devices; display devices may include However, it is not limited to devices with display functions such as personal computers and televisions.

In addition, the term "and/or" in this article is only an association relationship describing associated objects, which means that there may be three relationships, for example, A and/or B may mean: A exists alone, A and B exist at the same time, There are three cases of B alone. In addition, the character "/" in this article generally indicates that the contextual objects are an "or" relationship.

In the existing hybrid cloud platform, each public cloud and private cloud needs to be managed separately, for example, based on different cluster creation methods, clusters are created in corresponding clouds respectively. After the cluster is created, related computing power components need to be independently installed in the nodes of the cluster to support the deployment of subsequent services. Then, services are deployed in public cloud or private cloud clusters, which cannot fully utilize the resources of the hybrid cloud platform, resulting in low service deployment efficiency of the hybrid cloud platform.

FIG. 1 is a schematic diagram according to a first embodiment of the present disclosure; as shown in FIG. 1 , this embodiment provides a service deployment method for a hybrid cloud platform, which may specifically include the following steps:

S101. Obtain a user's service deployment request; the service deployment request carries the function information of the target model to be deployed and the service deployment information of the target model;

In this embodiment, a hybrid cloud management platform may be provided for implementing management on the hybrid cloud platform, for example including service deployment on the hybrid cloud platform. Therefore, the service deployment method of the hybrid cloud platform in this embodiment can be applied to the hybrid cloud management platform. The hybrid cloud management platform can manage a cloud platform including at least one public cloud and at least one private cloud.

The hybrid cloud platform in this embodiment may include a public cloud or a private cloud. The hybrid cloud platform may be a hybrid cloud platform based on the container cluster management system kubernetes (K8s for short), so the hybrid cloud platform may also be called the K8s hybrid cloud platform.

The user in this embodiment is not an individual user, but can be a merchant account, or an account created based on a specified project, and needs services that can deploy various models required by the specified project on the hybrid cloud platform.

The target model in this embodiment may be a pre-trained neural network model capable of realizing certain functions.

S102. Obtain the remaining amount of various types of computing power resources on each node in each cluster created in advance for the user in the hybrid cloud platform;

In this embodiment, clusters are created in advance on each cloud platform in the hybrid cloud platform. And each node of each cluster is deployed with a collector component, which can collect the remaining amount of various types of computing power resources on the node. Of course, optionally, each collector component can also collect usage of various types of computing resources on each node.

S103. Deploy the service of the target model on the hybrid cloud platform based on the remaining amount of each type of computing resources of each node in each cluster of the hybrid cloud platform and the service deployment information of the target model.

In this embodiment, the service of the target model can be deployed on the hybrid cloud platform by referring to the remaining amount of each type of computing resources of each node in each cluster of the hybrid cloud platform and the service deployment information of the target model. The deployment method of this embodiment, for the hybrid cloud management platform, does not need to distinguish between different clouds in the hybrid cloud platform and clusters in different clouds, but regards the hybrid cloud platform as a whole to deploy services of the target model. That is to say, in this embodiment, deploying the service of the target model on the hybrid cloud platform is not limited to only deploying the service of the target model on one cloud platform in the hybrid cloud platform; Deploy the service of the target model in more than two cloud platforms, so that the resources in the hybrid cloud platform can be fully utilized, and the deployment efficiency of the service of the target model can be improved.

The service deployment method of the hybrid cloud platform in this embodiment can regard the hybrid cloud platform as a whole based on the remaining amount of each type of computing resources of each node in each cluster of the hybrid cloud platform and the service deployment information of the target model, Deploying the service of the target model in the hybrid cloud platform is not limited to deploying the service of the target model on only one cloud platform in the hybrid cloud platform, which can make full use of the resources in the hybrid cloud platform and effectively improve the model in the hybrid cloud platform. service deployment efficiency.

Fig. 2 is a schematic diagram according to the second embodiment of the present disclosure; as shown in Fig. 2, this embodiment further introduces the technical solution of the present disclosure in more detail on the basis of the technical solution of the embodiment shown in Fig. 1 above, specifically, Including the following steps:

S201. Receive a service deployment request from an external user;

S202. Store the user's service deployment request into a queue;

The service deployment request in this embodiment carries the function information of the target model to be deployed and the service deployment information of the target model.

For example, in this embodiment, the service deployment information of the target model may include the type of the first computing power resource on which the service deployment of the target model depends, the amount of the first computing power resource relied on, and the total number of deployed target models, The total number of deployments can also be considered as the total number of copies of the service deploying the target model, or the total number of instances of the service deploying the target model. In this embodiment, the total number of deployments can be specified by the user according to requirements.

In the scenario of this embodiment, the types of computing power resources may include a central processing unit (Central Processing Unit; CPU), memory, graphics card, etc.; the graphics card may be divided into different types of graphics cards, and different types of graphics cards have different video memories. Further, the types of computing power resources may also include tensor processing units (Tensor Processing Unit; TPU), neural network processors (Neural Processing Unit; NPU); graphics processing units (Graphics Processing Unit; GPU); intelligent processors (Intelligent Processing Unit: IPU) and so on XPU.

In this embodiment, the type of the first computing resource that the service deployment of the target model depends on may include one type, two types, or multiple types. For example, some target models only depend on CPU and memory, while some target models also depend on a specified type of graphics card; some target models need to depend on TPU, NPU, or GPU.

In this embodiment, the service of the deployment target model depends on the amount of the first computing power resource may include: the number of cores of the CPU on which the service of the deployment target model depends, the size of the dependent memory (the unit can be megabytes), and the specified model of the dependence At least one of the number of graphics card blocks, the number of dependent TPU blocks, the number of dependent NPU blocks, the number of dependent GPU blocks, and the number of dependent IPU blocks. The hybrid cloud management platform in this embodiment takes a certain amount of time to deploy services, so each service deployment request received by the hybrid cloud management platform is not processed in real time when it is received.

In order to avoid the loss or omission of service deployment requests, in this embodiment, a queue can be set up, and after the hybrid cloud management platform receives the user's service deployment requests sent from the outside, it first stores the user's service deployment requests in the queue. Then when processing, read a user's service deployment request from the queue for service deployment. Specifically, when storing, it is stored at the end of the queue. When reading, read from the head of the queue.

In this embodiment, the external user's service deployment request may be input by the user through the application programming interface (Application Programming Interface; API) of the hybrid cloud management platform, or may also be input by the user through the interface of the hybrid cloud management platform.

S203. Obtain the user's service deployment request from the queue;

Specifically, the user's service deployment request is obtained from the head of the queue.

It should be noted that, in this embodiment, before step S201, it may also include: accessing the target model, that is, storing all information such as the parameters of the target model in the current hybrid cloud management platform to realize the access of the target model, This enables the hybrid cloud management platform to know which model service is to be deployed when receiving a service deployment request about the target model.

Further, after step S203, the method may further include: performing format conversion on the pre-accessed object model to obtain an object model in a standard format.

Since different target models may adopt different formats when the hybrid cloud management platform accesses the target model, the hybrid cloud management platform cannot recognize target models in each format. Therefore, when the hybrid cloud management platform receives the service deployment request, it can first convert the target model defined in the service deployment request into the standard format of the hybrid cloud management platform, so as to facilitate the identification of the target model.

S204. Obtain from the storage unit the remaining amount of each type of computing resources reported by each node in each cluster;

For example, in this embodiment, the remaining amounts of various types of computing power resources reported by each node may include the remaining number of cores of the CPU, the remaining size of the memory (the unit may be megabytes), and the remaining number of blocks of the graphics card. One, two or more graphics cards of specified models can be deployed on a node.

By adopting this method, the remaining amount of each type of computing resources of each node can be obtained accurately and efficiently.

Optionally, before this step S204, it may also include: receiving the remaining amount of each type of computing power resource reported by each node in each cluster, and storing it in the storage unit, which can effectively ensure that each node in the storage unit The accuracy of the remaining amount of various types of computing power resources.

Specifically, each cluster corresponding to the user is created on a cloud platform. For example, for a hybrid cloud platform including one private cloud platform and at least two public cloud platforms, corresponding clusters may be pre-created on the private cloud platform and each public cloud platform respectively. The manner of creating the cluster may be in any manner, which is not limited here. In addition, collector components can be deployed on each node in each cluster, which can collect the remaining amount of various types of computing power resources on each node. For example, for a multi-core CPU machine, the collector component can collect how many cores are not occupied by the CPU and how much memory is left. For a machine with a specified type of graphics card, the collection component can collect the number of blocks of the remaining graphics card, etc.

Optionally, in this embodiment, the remaining amount of each type of computing resources reported by each node in the cluster may also be obtained from each cluster in real time. Compared with the previous method, it takes longer, but the obtained data is more accurate. Of course, the reporting period of the remaining amount of each type of computing power resource reported by each node can also be set shorter, which can effectively improve the accuracy of the remaining amount of each type of computing power resource of each node stored in the storage unit.

S205. The remaining resource amount of each type of computing power resource of each node in each cluster based on the hybrid cloud platform, the type of the first computing power resource that the target model depends on, the amount that depends on the first computing power resource, and the service of deploying the target model total quantity, detect whether the deployment of the service of the total quantity target model can be implemented in the hybrid cloud platform; if yes, perform step S206; if not, perform step S209;

S206. Determine the target service deployment strategy of the target model; execute step S207;

The target service deployment strategy includes an identifier of at least one first cluster and a first quantity deployed in each first cluster; the sum of the first quantities deployed in each first cluster is equal to the total quantity. At this point, the target service deployment policy is only a pre-allocated solution for model deployment, which is used to guide the next steps. The first cluster in this embodiment is any cluster in the hybrid cloud platform.

In this embodiment, it can be specifically determined according to the amount of remaining resource space of each type of computing power resource of each node in each cluster of the hybrid cloud platform and the amount of resources required to deploy a target model service. The number of services of the target model that can be deployed in the cluster.

Based on the above, it can be known that when the amount of the first computing power resources in the hybrid cloud platform is sufficient, the target service deployment strategy of the target model can be directly determined, and the target service deployment strategy can realize the direct deployment of the overall cloud platform in the hybrid cloud platform. The number of target models to serve. During specific deployment, the total number of target model services may be deployed in one first cluster, or may be deployed in two or more first clusters. If there are more than two first clusters deployed, the sum of the first numbers deployed by each first cluster must be equal to the total number of services of the deployment target model.

S207. Generate a target image file corresponding to the target service deployment strategy; execute step S208;

Specifically, the target image file corresponding to the target service deployment strategy can be generated based on the target model in the standard format and the type of the dependent first computing resource, so that the generated target image file can be used on the first computing resource in the hybrid cloud platform Running, the service that implements the deployment target model is in the container of the corresponding cluster. By adopting this method, the accuracy of the generated image file can be effectively ensured.

Further optionally, after step S207, the method may further include: storing the generated target image file in the image file warehouse. That is, an image file warehouse is configured on the hybrid cloud management platform to store all generated image files, so that the generated image files can be stored in a timely and effective manner, so that they can be directly obtained when needed later.

S208. Deploy the corresponding first number of target image files to each first cluster, so as to realize the service of deploying a total number of target models in the hybrid cloud platform, and perform step S224;

For example, in this embodiment, by calling the API of each first cluster, the first cluster can be sent to the first cluster to carry the identification service deployment command corresponding to the first quantity and the target image file, so that the first cluster can pull it from the hybrid cloud management platform. Corresponding target image files, and deploying the first number of copies, so as to implement the service of deploying a total number of target models in the hybrid cloud platform.

By adopting the above method, when the first computing resources in the hybrid cloud platform are sufficient, the services of the total number of target models can be accurately and efficiently deployed on the hybrid cloud platform.

S209. Determine the first sub-service deployment strategy that can deploy the target model in the hybrid cloud platform; execute step S210;

The first sub-service deployment strategy in this embodiment includes an identifier of at least one second cluster and a second quantity deployed in each second cluster; the sum of the second quantities deployed in each second cluster is less than the total quantity. The second cluster is also any cluster in the hybrid cloud platform.

The first sub-service deployment strategy is to make full use of the first computing power resources that the target model depends on in the hybrid cloud platform, and deploy as many services as possible of the target model. That is, the first sub-service deployment strategy only completes part of the deployment tasks.

It should be noted that, in practical applications, if the first computing resource that the target model depends on in the hybrid cloud platform is very scarce, it may not be possible to complete the service deployment of a target model. At this time, the first sub-service deployment policy may also be empty. That is, the first sub-service deployment strategy does not include the identifier of the second cluster, and the second quantity deployed in each second cluster is empty. For the convenience of subsequent statistics, it is also possible to configure the first sub-service deployment strategy to include each cluster in the hybrid cloud platform, but the second number deployed in the cluster is 0.

S210. Based on the total quantity and the second quantity corresponding to each second cluster, determine the remaining quantity to be deployed; perform step S211;

Specifically, the sum of the second quantities corresponding to the second clusters is firstly taken as the deployed quantity. Then subtract the deployed quantity from the total quantity to get the remaining quantity deployed.

In this embodiment, after step S210, in order to implement the deployment of the remaining number of target models, the following deployment steps of model conversion can be performed: the remaining amount of each type of computing resources of each node in each cluster based on the hybrid cloud platform, As well as the function information of each model in the deployed multiple models and the service deployment information of each model, deploy the remaining number of services of the deployed model that have the same function as the target model in the hybrid cloud platform, so as to achieve the goal of replacing the remaining number Model service.

The service deployment information of each model may include the type and amount of computing power resources that the service deployment of the corresponding model depends on.

The deployment steps of this model conversion can be implemented when the computing power resources that the target model depends on on the hybrid cloud platform are insufficient, and the computing power resources that the deployed model on the hybrid cloud platform has the same function as the target model depend on and the computing power resources that the target model depends on In the case of different and sufficient computing power resources, the service of deploying the remaining number of target models is converted into the service of deploying the deployed model with the same function, which can make full use of all computing power resources on the hybrid cloud platform for model deployment , effectively improving the deployment efficiency of the model.

For example, the deploying step of model conversion may include all or part of the following steps S211-S222 during specific implementation. That is to say, in the process of model conversion, the model conversion and the service deployment of the converted corresponding model can be performed only once, or two or more model conversions and the service deployment of the converted corresponding model can be performed. For details, refer to the following implementation Example records.

S211. Based on the function information of the target model and the function information of each of the deployed models, obtain a first substitute model having the same function as the target model; perform step S212;

S212. The remaining amount of each type of computing power resources of each node in each cluster based on the hybrid cloud platform, the type of the second computing power resource that the service deployment of the first alternative model depends on, and the amount of the second target computing power resource that depends on and the remaining number of deployments, detecting whether the remaining number of services of the first alternative model can be deployed in the hybrid cloud platform; if yes, execute step S213; if not, execute step S216;

The type of the second computing power resource in this embodiment is different from the type of the first computing power resource.

In the detection process of this step, the hybrid cloud platform is regarded as a whole, and the remaining quantity of each type of computing resources of each node in each cluster in the hybrid cloud platform is fully considered, so that the hybrid cloud platform can make full use of Computing resources for model deployment.

S213. Determine the second sub-service deployment strategy for deploying the first alternative model in the hybrid cloud platform; execute step S214;

The second sub-service deployment strategy in this embodiment includes an identifier of at least one third cluster and a third quantity deployed in each third cluster; the sum of the third quantities deployed in each third cluster is equal to the remaining quantity. The third cluster is also any cluster in the hybrid cloud platform.

Based on the above, it can be known that at this time, the deployment services of the total number of target models can be split using the first sub-service deployment strategy to deploy part of the number of target model services, using the second sub-service deployment strategy to deploy the remaining number The first alternative model of service. At this point, the deployment service pre-allocation scheme for the total number of target models is completed.

S214. Respectively acquire the first image file corresponding to the first sub-service deployment strategy and the second image file corresponding to the second sub-service deployment strategy; execute step S215;

Among them, since the target model is a newly accessed model, not a deployed model, there will be no corresponding image file in the image file warehouse. Therefore, based on the target model in the standard format and the type of the first computing power resource that depends on, A first image file corresponding to the first sub-service deployment strategy is generated. The generated first image file can be run on the first computing power resource, and the service that implements the deployment target model is in the container of the corresponding cluster.

The first alternative model deployed by the second sub-service deployment strategy is a model that has already been deployed in the hybrid cloud platform. At this time, it can be detected whether there is a second image file corresponding to the second sub-service deployment strategy in the image file warehouse; If not, at this time, the second image file corresponding to the second sub-service deployment strategy may be further generated based on the first alternative model in the standard format and the type of the second computing power resource to be relied upon. The generated second image file can be run on the second computing resource, so that the service of the first alternative model can be deployed in the container of the corresponding cluster.

By adopting the above method, the second image file corresponding to the deployment strategy of the second sub-service can be obtained efficiently and accurately.

S215. Deploy the first image files corresponding to the second number to each second cluster, and deploy the second image files corresponding to the third number to each third cluster, so as to implement the total number of target models deployed in the hybrid cloud platform service; execute step S224;

Because the second numbers corresponding to different second clusters may be different, the third numbers corresponding to different third cluster deployments may also be different. In order to ensure the accuracy of deployment, it is necessary to ensure that the number of first image files deployed in each second cluster is the second number corresponding to the current second cluster, and the number of second image files deployed in each third cluster is the current The third cluster corresponds to the third quantity.

That is to say, in this embodiment, in order to realize the deployment of the service of the target model, the deployment of the first alternative model that has the same function as the target model can be deployed when the first computing resource that the target model depends on is insufficient. In the case of sufficient resources, the services of the deployment target model are converted to the services of the first alternative model, so as to ensure that the deployed services have the same functions and the same service quantity and service quality.

With this deployment method, the computing resources in the hybrid cloud platform can be fully utilized for service deployment, ensuring the quality of service deployment and improving service deployment efficiency.

S216. Determine the third sub-service deployment strategy for deploying the first alternative model in the hybrid cloud platform; execute step S217;

At this time, the third sub-service deployment strategy includes at least one fourth cluster identifier and the fourth quantity deployed in each fourth cluster; the sum of the fourth quantities deployed in each fourth cluster is less than the remaining quantity. Similarly, the fourth cluster is also any cluster in the hybrid cloud platform.

That is to say, resources of the hybrid cloud platform can be fully utilized at this time, and a part of the services of the first alternative model can be deployed to replace a part of the services of the target model.

Similarly, referring to the description in step 209 above, the third sub-service deployment strategy can also be empty. That is, the third sub-service deployment strategy does not include the identifier of the fourth cluster, and the fourth quantity deployed in each fourth cluster is empty. Similarly, for the convenience of subsequent statistics, it is also possible to configure the third sub-service deployment strategy to include each cluster in the hybrid cloud platform, but the fourth number deployed in the cluster is 0.

S217. Based on the fourth quantity deployed in each fourth cluster, update the remaining quantity deployed; perform step S218;

Specifically, the sum of the fourth quantities deployed by each fourth cluster can be taken as the quantity deployed by the third sub-service deployment strategy; quantity.

S218. Based on the function information of the target model and the function information of each of the deployed models, obtain a second substitute model having the same function as the target model; perform step S219;

S219. The remaining amount of each type of computing power resources of each node in each cluster based on the hybrid cloud platform, the type of the third computing power resource that the service deployment of the second alternative model depends on, and the amount of the third target computing power resource that depends on , and the updated remaining quantity, detecting whether the hybrid cloud platform can implement the deployment of the service of the updated remaining quantity of the second alternative model; if yes, perform step S220; if not, perform step S223;

In this embodiment, the type of the third computing resource is different from the type of the first computing resource.

S220. Determine the fourth sub-service deployment strategy for deploying the second alternative model in the hybrid cloud platform; execute step S221;

The fourth sub-service deployment strategy in this embodiment includes an identifier of at least one fifth cluster and a fifth quantity deployed in each fifth cluster; the sum of the fifth quantities corresponding to each fifth cluster is equal to the updated remaining quantity. Similarly, the fifth cluster may also be any cluster in the hybrid cloud platform.

Based on the above, it can be known that at this time, the deployment services of the total number of target models can be split into services that use the first sub-service deployment strategy to deploy part of the number of target models, and use the third sub-service deployment strategy to deploy the part The number of services of the first alternative model, and the remaining number of services of the second alternative model are deployed using the fourth sub-service deployment strategy.

Each sub-service deployment strategy in this embodiment can be regarded as a pre-allocation scheme for service deployment, which is used to guide subsequent service deployment.

S221. Obtain the first image file corresponding to the first sub-service deployment strategy, the third image file corresponding to the third sub-service deployment strategy, and the fourth image file corresponding to the fourth sub-service deployment strategy; perform step S222;

For the specific implementation of this step, reference may be made to the above step S214, which will not be repeated here.

S222. Deploy second image files corresponding to the second number to each second cluster, deploy third image files corresponding to the fourth number to each fourth cluster, and deploy fourth image files corresponding to the fifth number to each fifth cluster , to realize the service of deploying a total number of target models in the hybrid cloud platform; perform step S224;

That is to say, in this embodiment, in order to realize the deployment of the service of the target model, the first alternative model with the same function as the target model may also In the case of partial computing resources, the service of deploying a part of the target model is converted into the service of deploying a part of the first alternative model; further, the service of deploying a part of the target model can also be converted into a deployment of A partial number of services for a second alternative model that has the same functionality as the target model. This deployment method can make full use of various types of computing resources in the hybrid cloud platform for service deployment, ensure service deployment quality, and improve service deployment efficiency.

S223. If it is determined that there is no other alternative model having the same function as the target model in the deployed multiple models, it is determined that the total number of services of the target model cannot be deployed in the hybrid cloud platform; perform step S225;

It should be noted that, if it is determined that there are other alternative models having the same function as the target model among the deployed models, the deployment can be continued in the manner of the above steps S219-S222. However, if the final pre-allocation plan for deploying the total number of target model services has not been determined, and there are no other alternative models with the same functions among the deployed multiple models, at this time, it is determined that the total number of targets cannot be deployed in the hybrid cloud platform Model service.

S224, returning to the user that the service deployment of the target model is completed, and ends.

S225. Return to the user the service that the target model cannot be deployed, and end.

The service deployment method of the hybrid cloud platform in this embodiment can regard the hybrid cloud platform as a whole based on the remaining resources of each type of computing resources of each node in each cluster of the hybrid cloud platform and the service deployment information of the target model , deploying the service of the target model in the hybrid cloud platform, without limiting the deployment of the service of the target model on some single cloud platforms in the hybrid cloud platform, can make full use of the resources in the hybrid cloud platform and effectively improve the efficiency of the hybrid cloud platform. Deployment efficiency of services in the target model;

Furthermore, in this embodiment, when deploying the service of the target model on the hybrid cloud platform, if the computing power resources on which the target model depends are insufficient, only a part of the target model can be deployed by using the remaining computing power resources. For the deployment of the service of the remaining number of target models, an alternative model with the same function as the target model can be obtained through model conversion, and when there are dependent computing resources of the alternative model in the hybrid cloud platform, the remaining number of target models Deployment of the service, converted to a service that deploys the remaining number of alternative models. Moreover, in the scenario of this embodiment, the deployment service of one target model can be converted into deploying one, two or The service of more than two alternative models can make full use of various types of computing resources of the hybrid cloud platform and effectively improve the efficiency of service deployment.

Fig. 3 is a schematic diagram according to the second embodiment of the present disclosure; as shown in Fig. 3, this embodiment further introduces the technical solution of the present disclosure in more detail on the basis of the technical solution of the above-mentioned embodiment shown in Fig. 1 or Fig. 2 , specifically may include the following steps:

S301. Receive a cluster creation request from a user; the cluster creation request carries a cluster creation platform identifier and information of various types of computing resources to be configured;

The technical solution of this embodiment is mainly used to provide a hybrid cloud-based management platform to realize cluster creation. The creation request of the cluster may be input by the user through the API of the hybrid cloud management platform, or through the interface of the hybrid cloud management platform.

Specifically, the user in this embodiment is the same as the user in the embodiment shown in FIG. 1 above, and may be a merchant account or an account created based on a specified item.

The cluster creation request can be generated and initiated based on the business plan or project plan. For example, project planning requires the creation of clusters with certain computing power resources in a specified public cloud or private cloud platform.

S302. Based on the identification of the cluster creation platform and the information of various types of computing power resources to be configured, obtain multiple nodes in the corresponding cloud platform to build a user cluster; each node can provide at least one type of computing power resources;

In this embodiment, a node may correspond to a device, and the device can increase at least one type of computing resources. For example, some nodes in the cluster have multi-core CPUs and memory of a certain size; while some nodes are equipped with graphics cards of specified models.

For example, a cluster creation request for a certain project can directly request to create a cluster in the private cloud platform that meets the following conditions: 5 devices with 8-core CPUs, 3 devices with specified graphics cards, and the total memory can reach the first preset threshold. Based on the above conditions, the nodes corresponding to each condition can be screened sequentially in the corresponding private cloud platform to obtain nodes that meet the conditions and build a cluster. For example, first screen 5 devices with 8-core CPUs, and then check whether the 5 devices have the specified graphics card. If so, check whether the total memory of the 5 devices reaches the first budget threshold. If so, determine the 5 devices. Devices with 8-core CPUs can form a cluster. Otherwise, if it is detected that these 5 devices do not have the graphics card of the specified model, another 3 devices with the graphics card of the specified model need to be selected. Then further check whether the total memory of 5 devices with 8-core CPUs and 3 devices with the specified graphics card reaches the first preset threshold, and if so, determine whether the total memory of 5 devices with 8-core CPUs and 3 devices with the Devices with specified graphics cards can form a cluster. If not, you need to select a device with a certain amount of memory to form a cluster with 5 devices with 8-core CPUs and 3 devices with the specified graphics card.

The above example is only an example, and is not intended to limit the cluster creation request of this application. In an actual application scenario, on the premise of not violating the implementation idea of this application, a corresponding cluster can be created on a corresponding platform based on any form of cluster creation request.

The above cluster creation method in this embodiment can be based on the hybrid cloud management platform to implement cluster creation in each cloud platform in the hybrid cloud platform, which is very simple and convenient to implement. Moreover, compared with the prior art of manually creating clusters on each cloud platform, the cluster creation efficiency can be effectively improved.

S303. Obtain the collector component from the component warehouse;

In this embodiment, the hybrid cloud management platform is further configured with a component warehouse, and the component warehouse is configured with multiple components, for example, may include a collector component, and may also include a graphics card service component and the like.

S304. Deploy a collector component in each node of the cluster, so as to collect remaining resource amounts of various types of computing power resources of each node;

Optionally, by adopting the above steps S301 and S302, the cluster can be created on any cloud platform in the hybrid cloud platform.

Furthermore, in order to more effectively implement the deployment of services in the cluster, the above steps S303 and S304 can be used to deploy collector components in each node in the cluster, so as to realize the remaining Real-time collection of resources and reporting. Optionally, the collector component in each node can also collect usage of various types of resources on each node, report it to the hybrid cloud management platform, and store it in the storage unit.

For example, in specific implementation, after the cluster is created in step S302. The hybrid cloud management platform can call the API of the cluster and initiate a deployment command carrying the collector component identifier. The cluster can pull the collector component from the hybrid cloud management platform and install it on each node to realize the The collector component is deployed on the node.

In this embodiment, by deploying the collector component on each node in the created cluster, the remaining amount of each type of computing resources of each node can be obtained in a timely and accurate manner, which facilitates the deployment of subsequent services.

S305. Obtain the graphics card service component from the component warehouse;

S306. Deploy the graphics card service component in the nodes of the cluster that provide the graphics card resource;

Steps S305 and S306 are a supplementary solution. When there are graphics card resources in the cluster, it is necessary to deploy the graphics card service component on the device with the graphics card resource at the same time to empower the device so that the device can improve the graphics card service.

The graphics card service component can be a device plugin. When used, the graphics card service component can report the graphics card usage of the current node to the kubelet component in the cluster, so that the cluster can refer to the graphics card usage of the node when deploying services. , for more accurate service deployment.

S307. Receive a cluster modification request from a user;

The cluster modification request in this embodiment may carry the identifier of the cluster to be modified; the cluster modification request is used to instruct to delete the cluster to be modified, or to add nodes in the cluster, or to delete nodes in the cluster.

S308. Based on the user's cluster modification request, update the corresponding cluster.

For example, if the cluster modification request indicates to delete the cluster, the corresponding cluster can be deleted.

If the cluster modification request indicates to add a node with a specified type of resource to the cluster, then according to the cluster modification request, obtain a machine with the specified type of resource from the cloud platform where the cluster ID is located, as long as the cluster is sufficient. In addition, it should be noted that, for a newly added node, it is necessary to refer to the above step S303 and step S304 to deploy the collector component in the newly added node. If the newly added node has graphics card resources, step S305 and step S306 need to be adopted to deploy the graphics card service component in the node.

If the cluster modification request indicates to delete a node of a specified type of resource for the cluster, the corresponding node in the cluster can be deleted according to the cluster modification request.

The above technical solution of this embodiment can create clusters in each cloud platform based on the same hybrid cloud management platform, without configuring a corresponding cluster creation platform for each cloud platform separately, and the management is very convenient.

Moreover, in this embodiment, collector components can also be deployed in each node of the cluster to empower the cluster and realize the ability to monitor the remaining resources of each node in the cluster, providing effective support for subsequent service deployment in the cluster.

Furthermore, in this embodiment, the graphics card service component can also be deployed on the nodes supporting the graphics card in the cluster, so as to endow the cluster with the ability to implement graphics cards, and provide effective support for the subsequent deployment of services that require graphics cards.

Furthermore, in this embodiment, the modification of the cluster can also be implemented based on the hybrid cloud management platform, which is very convenient for the management and maintenance of the cluster.

For example, FIG. 4 is a management diagram of a hybrid cloud management platform provided by the present disclosure. As shown in FIG. 4 , taking a hybrid cloud platform managed by a hybrid cloud management platform including a public cloud platform and a private cloud platform as an example, more public cloud platforms may also be included in practical applications. As shown in FIG. 4 , in this embodiment, the hybrid cloud management platform includes a scheduler, a computing unit, a storage unit, a model conversion unit, an adapter, a component warehouse, and an image file warehouse as an example. In practical applications, it can also be divided according to other functional units, which is not limited here.

Referring to the description of the embodiment shown in FIG. 3 above, the hybrid cloud management platform can receive an external cluster creation request through an API, and then create a corresponding cluster on the public cloud platform or private cloud platform based on the cluster creation request. In practical applications, traditional methods or other methods may also be used to create clusters in each cloud platform of the hybrid cloud platform, which is not limited here. Further, after the cluster is created, the collector component can be obtained from the component warehouse and deployed on each node of the cluster by referring to the description of the above embodiment shown in FIG. 3 . For a node including a graphics card component, it is also necessary to obtain the graphics card service component from the component warehouse and deploy it on the corresponding node.

The hybrid cloud management platform can also receive external service deployment requests through the API and store them in a queue (Queue). Then the scheduler reads a service deployment request from the queue, as in step S203 of the above embodiment. Further, the scheduler schedules the computing unit to execute steps S204-S207 of the above embodiment, and store the generated image file in the image file repository. And the scheduler schedules the adapter to implement the service of the deployment target model by executing step S208.

And if the calculation unit determines that the service deployment of the total number of target models cannot be implemented in each cluster of the hybrid cloud platform during the calculation process, the scheduler schedules the calculation unit to perform steps S209 and S210; and schedules the model conversion unit to execute Step S211 converts the model, and further executes steps S212-S213 by the calculation unit to obtain the image files corresponding to the deployment strategies of each sub-service, and store them in the image file warehouse. And the scheduler schedules the adapter to implement the service of the deployment target model by executing step S215. By analogy, the above-mentioned steps S216-S222 can also be executed based on the above-mentioned units until the service deployment of the target model is finally returned to the user, or the service that cannot deploy the target model is returned to the user.

FIG. 5 is a schematic diagram according to a fourth embodiment of the present disclosure; as shown in FIG. 5, the hybrid cloud management platform 500 of this embodiment includes:

The service deployment acquiring module 501 is configured to acquire a user's service deployment request; the service deployment request carries the function information of the target model to be deployed and the service deployment information of the target model;

The remaining resource acquisition module 502 is used for the remaining resource acquisition module, and is used for acquiring the remaining amount of various types of computing power resources on each node in each cluster pre-created for the user in the hybrid cloud platform;

The service deployment module 503 is configured to deploy said The serving of the target model.

The hybrid cloud management platform 500 of this embodiment uses the above-mentioned modules to realize the implementation principle and technical effect of service deployment based on the hybrid cloud platform, which is the same as the implementation of the above-mentioned related method embodiments. For details, please refer to the records of the above-mentioned related method embodiments. I won't repeat them here.

Fig. 6 is a schematic diagram according to the fifth embodiment of the present disclosure; as shown in Fig. 6, the hybrid cloud management platform 600 of this embodiment includes modules with the same name and the same function as shown in Fig. 5 above: service deployment acquisition module 601, remaining resource acquisition Module 602 and service deployment module 603.

In this embodiment, the service deployment obtaining module 601 is configured to obtain the user's service deployment request from the queue.

As shown in FIG. 6, the hybrid cloud management platform 600 of this embodiment also includes:

A receiving module 604, configured to receive an external service deployment request from the user;

The storage module 605 is configured to store the user's service deployment request into the queue.

Further optionally, in an embodiment of the present disclosure, the remaining resource acquisition module 502 is configured to:

The remaining amount of each type of computing resources reported by each node in each cluster is obtained from the storage unit.

Further optionally, in one embodiment of the present disclosure, the receiving module 604 is also configured to receive the remaining amount of each type of computing resources reported by each node in each cluster;

The storage module 605 is configured to store, in the storage unit, the remaining resource amounts of each type of computing resources reported by each node in each cluster.

Further optionally, in an embodiment of the present disclosure, the service deployment module 603 includes:

The calculation unit 6031 is configured to be based on the remaining amount of each type of computing power resource of each node in each cluster of the hybrid cloud platform, the type of the first computing power resource on which the target model depends, and the number of dependent first computing power resources. amount and the total number of services deploying the target model, detecting whether the deployment of the total number of services of the target model can be implemented in the hybrid cloud platform;

The determining unit 6032 is configured to, if possible, determine a target service deployment strategy of the target model; the target service deployment strategy includes an identifier of at least one first cluster and a first number of deployments in each first cluster; each of the The sum of the first number of deployments of the first cluster is equal to the total number of first clusters on which the service of the target model is deployed and the first number of deployments in the first cluster;

A generating unit 6033, configured to generate a target image file corresponding to the target service deployment strategy;

The deploying unit 6034 is configured to deploy a corresponding first quantity of target image files to each of the first clusters, so as to implement the service of deploying the total quantity of the target models in the hybrid cloud platform.

Further optionally, in an embodiment of the present disclosure, it also includes:

A format conversion module 606, configured to perform format conversion on the pre-accessed target model to obtain a target model in a standard format;

generating unit 6033 for:

Generate the target image file corresponding to the target service deployment strategy based on the target model in the standard format and the type of the dependent first computing resource;

And store the generated target image file in the image file repository.

Further optionally, in one embodiment of the present disclosure, the deployment unit 6034 is configured to:

By invoking the application program interface of each of the first clusters, a service deployment command carrying the corresponding first quantity and the identification of the target image file is sent to the first cluster, so that the corresponding first cluster can receive from the hybrid cloud The management platform pulls the corresponding target image file, and deploys the first number of copies, so as to implement the service of deploying the total number of the target models in the hybrid cloud platform.

Further optionally, as shown in FIG. 6 , in an embodiment of the present disclosure, the service deployment module 603 further includes a conversion deployment unit 6035 .

The determination unit 6032 is also used for:

If it is detected and determined that the deployment of the total number of services of the target model cannot be implemented in the hybrid cloud platform, determine a first sub-service deployment strategy in the hybrid cloud platform that can deploy the target model; the second A sub-service deployment policy includes an identifier of at least one second cluster and a second quantity deployed in each second cluster; the sum of the second quantities deployed in each second cluster is less than the total quantity;

Based on the total quantity and the second quantity corresponding to each of the second clusters, determine a remaining quantity for deployment.

Conversion Deployment Unit 6035 for:

Based on the remaining amount of each type of computing resources of each node in each cluster of the hybrid cloud platform, and the function information of each model in the deployed multiple models and the service deployment information of each model, the hybrid cloud platform Deploy the services of the remaining number of deployed models that have the same functionality as the target model in .

Further optionally, in an embodiment of the present disclosure, the conversion and deployment unit 6035 is configured to:

Based on the remaining amount of each type of computing power resource of each node in each cluster of the hybrid cloud platform, the type of the second computing power resource that the service deployment of the first alternative model depends on, and the second target computing power resource that depends on amount, and the remaining number of deployments, detecting whether the remaining number of services of the first alternative model can be deployed in the hybrid cloud platform; the type of the second computing power resource is the same as the first computing power resource different types of human resources;

If yes, determine the second sub-service deployment strategy for deploying the first alternative model in the hybrid cloud platform; the second sub-service deployment strategy includes at least one third cluster identifier and each third cluster deployed A third quantity; the sum of the third quantities deployed by each of the third clusters is equal to the remaining quantity;

Respectively obtain a first image file corresponding to the first sub-service deployment strategy and a second image file corresponding to the second sub-service deployment strategy;

Deploying the first image files corresponding to the second number to each of the second clusters, and deploying the second image files corresponding to the third number to each of the third clusters, so as to implement the deployment of all the images on the hybrid cloud platform. The total number of services for the target model.

Further optionally, in an embodiment of the present disclosure, the conversion and deployment unit 6035 is configured to:

Detecting and determining that the second image file corresponding to the second sub-service deployment strategy exists in the image file repository.

Further optionally, in an embodiment of the present disclosure, the conversion and deployment unit 6035 is configured to:

If it is detected and determined that the second image file corresponding to the second sub-service deployment strategy does not exist in the image file warehouse, the second sub-service is generated based on the first substitution model in the standard format and the type of the second computing resource that depends on it. The second image file corresponding to the service deployment policy.

Further optionally, in one embodiment of the present disclosure, the conversion and deployment unit 6035 is also used to:

If it is detected and determined that the deployment of the remaining number of services of the first alternative model cannot be implemented in the hybrid cloud platform, determine a third sub-service deployment strategy for deploying the first alternative model in the hybrid cloud platform ; The third sub-service deployment strategy includes an identifier of at least one fourth cluster and a fourth quantity deployed in each fourth cluster; the sum of the fourth quantities deployed in each fourth cluster is less than the remaining quantity;

updating said remaining number of deployments based on a fourth number of deployments of each of said fourth clusters;

Acquiring a second substitute model having the same function as the target model based on the function information of the target model and the function information of each of the deployed models;

Based on the remaining amount of each type of computing power resource of each node in each cluster of the hybrid cloud platform, the type of the third computing power resource that the service deployment of the second alternative model depends on, and the third target computing power resource that depends on and the updated remaining quantity, detecting whether the hybrid cloud platform can implement the deployment of the updated remaining quantity of services of the second alternative model; the type of the third computing resource, the The type of the second computing power resource is different from the type of the first computing power resource;

If yes, determine the fourth sub-service deployment strategy for deploying the second alternative model in the hybrid cloud platform; the fourth sub-service deployment strategy includes at least one fifth cluster identifier and each fifth cluster deployment strategy The fifth quantity; the sum of the fifth quantities corresponding to each of the fifth clusters is equal to the updated remaining quantity;

Respectively obtain the first image file corresponding to the first sub-service deployment strategy, the third image file corresponding to the third sub-service deployment strategy, and the fourth image file corresponding to the fourth sub-service deployment strategy;

Deploy the second image files corresponding to the second quantity to each of the second clusters, deploy the third image files corresponding to the fourth quantity to each of the fourth clusters, and deploy the fifth image files corresponding to the fifth to each of the fifth clusters. A quantity of fourth image files, so as to implement the deployment of the total quantity of services of the target model in the hybrid cloud platform.

Further optionally, as shown in FIG. 6, in an embodiment of the present disclosure, in an embodiment of the present disclosure, the hybrid cloud management platform 600 further includes:

The determination module 607 is configured to detect and determine that the hybrid cloud platform cannot implement the deployment of the updated remaining number of services of the second alternative model, and it is determined that there is no such thing among the deployed multiple models When using other alternative models of the target model that have the same function, it is determined that the total number of services of the target model cannot be deployed on the hybrid cloud platform.

Further optionally, as shown in FIG. 6, in an embodiment of the present disclosure, the hybrid cloud management platform 600 further includes:

The cluster request receiving module 608 is configured to receive the user's cluster creation request; the cluster creation request carries the cluster creation platform identifier and information of various types of computing resources to be configured;

The cluster creation module 609 is configured to acquire multiple nodes in the corresponding cloud platform to build the user's cluster based on the creation platform identifier of the cluster and the information of various types of computing resources to be configured; each of the nodes can Provide at least one type of computing resources.

Further optionally, as shown in FIG. 6, in an embodiment of the present disclosure, the hybrid cloud management platform 600 further includes:

A component acquisition module 610, configured to acquire the collector component from the component warehouse;

The component deployment module 611 is configured to deploy collector components in each node of the cluster, so as to collect remaining resource amounts of various types of computing power resources of each node.

Further optionally, in an embodiment of the present disclosure, the component acquisition module 610 is also configured to acquire the graphics card service component from the component warehouse;

The component deployment module 611 is further configured to deploy the graphics card service component in the nodes of the cluster that provide graphics card resources.

Further optionally, as shown in FIG. 6, in an embodiment of the present disclosure, the hybrid cloud management platform 600 further includes:

The cluster request receiving module 608 is also configured to receive the user's cluster modification request;

The cluster update module 612 is configured to update the corresponding cluster based on the user's cluster modification request.

The hybrid cloud management platform 600 of this embodiment uses the above-mentioned modules to realize the implementation principle and technical effect of service deployment based on the hybrid cloud platform, which is the same as the implementation of the above-mentioned related method embodiments. For details, please refer to the records of the above-mentioned related method embodiments. I won't repeat them here.

In the technical solution of the present disclosure, the acquisition, storage and application of the user's personal information involved are in compliance with relevant laws and regulations, and do not violate public order and good customs.

According to the embodiments of the present disclosure, the present disclosure also provides an electronic device, a readable storage medium, and a computer program product.

FIG. 7 shows a schematic block diagram of an example electronic device 700 that may be used to implement embodiments of the present disclosure. Electronic device is intended to represent various forms of digital computers, such as laptops, desktops, workstations, personal digital assistants, servers, blade servers, mainframes, and other suitable computers. Electronic devices may also represent various forms of mobile devices, such as personal digital processing, cellular telephones, smart phones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions, are by way of example only, and are not intended to limit implementations of the disclosure described and/or claimed herein.

As shown in FIG. 7, the device 700 includes a computing unit 701 that can execute according to a computer program stored in a read-only memory (ROM) 702 or loaded from a storage unit 708 into a random-access memory (RAM) 703. Various appropriate actions and treatments. In the RAM 703, various programs and data necessary for the operation of the device 700 can also be stored. The computing unit 701 , ROM 702 , and RAM 703 are connected to each other through a bus 704 . An input/output (I/O) interface 705 is also connected to the bus 704 .

Multiple components in the device 700 are connected to the I/O interface 705, including: an input unit 706, such as a keyboard, a mouse, etc.; an output unit 707, such as various types of displays, speakers, etc.; a storage unit 708, such as a magnetic disk, an optical disk, etc. ; and a communication unit 709, such as a network card, a modem, a wireless communication transceiver, and the like. The communication unit 709 allows the device 700 to exchange information/data with other devices over a computer network such as the Internet and/or various telecommunication networks.

The computing unit 701 may be various general-purpose and/or special-purpose processing components having processing and computing capabilities. Some examples of computing units 701 include, but are not limited to, central processing units (CPUs), graphics processing units (GPUs), various dedicated artificial intelligence (AI) computing chips, various computing units that run machine learning model algorithms, digital signal processing processor (DSP), and any suitable processor, controller, microcontroller, etc. The computing unit 701 executes the various methods and processes described above, such as the above-mentioned methods of the present disclosure. For example, in some embodiments, the above-described methods of the present disclosure may be implemented as a computer software program tangibly embodied on a machine-readable medium, such as the storage unit 708 . In some embodiments, part or all of the computer program may be loaded and/or installed on the device 700 via the ROM 702 and/or the communication unit 709 . When the computer program is loaded into the RAM 703 and executed by the computing unit 701, one or more steps of the above-mentioned method of the present disclosure described above may be performed. Alternatively, in other embodiments, the computing unit 701 may be configured in any other appropriate way (for example, by means of firmware) to execute the above-mentioned method of the present disclosure.

Various implementations of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), systems on chips Implemented in a system of systems (SOC), complex programmable logic device (CPLD), computer hardware, firmware, software, and/or combinations thereof. These various embodiments may include being implemented in one or more computer programs executable and/or interpreted on a programmable system including at least one programmable processor, the programmable processor Can be special-purpose or general-purpose programmable processor, can receive data and instruction from storage system, at least one input device, and at least one output device, and transmit data and instruction to this storage system, this at least one input device, and this at least one output device an output device.

Program codes for implementing the methods of the present disclosure may be written in any combination of one or more programming languages. These program codes may be provided to a processor or controller of a general-purpose computer, a special purpose computer, or other programmable data processing devices, so that the program codes, when executed by the processor or controller, make the functions/functions specified in the flow diagrams and/or block diagrams Action is implemented. The program code may execute entirely on the machine, partly on the machine, as a stand-alone software package partly on the machine and partly on a remote machine or entirely on the remote machine or server.

In the context of the present disclosure, a machine-readable medium may be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, apparatus, or device. A machine-readable medium may be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium may include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination of the foregoing. More specific examples of machine-readable storage media would include one or more wire-based electrical connections, portable computer discs, hard drives, random access memory (RAM), read only memory (ROM), erasable programmable read only memory (EPROM or flash memory), optical fiber, compact disk read only memory (CD-ROM), optical storage, magnetic storage, or any suitable combination of the foregoing.

To provide for interaction with the user, the systems and techniques described herein can be implemented on a computer having a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user. ); and a keyboard and pointing device (eg, a mouse or a trackball) through which a user can provide input to the computer. Other kinds of devices can also be used to provide interaction with the user; for example, the feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and can be in any form (including Acoustic input, speech input or, tactile input) to receive input from the user.

The systems and techniques described herein can be implemented in a computing system that includes back-end components (e.g., as a data server), or a computing system that includes middleware components (e.g., an application server), or a computing system that includes front-end components (e.g., as a a user computer having a graphical user interface or web browser through which a user can interact with embodiments of the systems and techniques described herein), or including such backend components, middleware components, Or any combination of front-end components in a computing system. The components of the system can be interconnected by any form or medium of digital data communication, eg, a communication network. Examples of communication networks include: Local Area Network (LAN), Wide Area Network (WAN) and the Internet.

A computer system may include clients and servers. Clients and servers are generally remote from each other and typically interact through a communication network. The relationship of client and server arises by computer programs running on the respective computers and having a client-server relationship to each other. The server can be a cloud server, a server of a distributed system, or a server combined with a blockchain.

It should be understood that steps may be reordered, added or deleted using the various forms of flow shown above. For example, each step described in the present disclosure may be executed in parallel, sequentially, or in a different order, as long as the desired result of the technical solution disclosed in the present disclosure can be achieved, no limitation is imposed herein.

The specific implementation manners described above do not limit the protection scope of the present disclosure. It should be apparent to those skilled in the art that various modifications, combinations, sub-combinations and substitutions may be made depending on design requirements and other factors. Any modifications, equivalent replacements and improvements made within the spirit and principles of the present disclosure shall be included within the protection scope of the present disclosure.

Claims (22)

1. A service deployment method of a hybrid cloud platform comprises the following steps:

acquiring a service deployment request of a user; the service deployment request carries functional information of a target model to be deployed and service deployment information of the target model;

acquiring the residual quantity of each type of computing power resource on each node in each cluster created in advance for the user in a hybrid cloud platform;

and deploying the services of the target model in the hybrid cloud platform based on the residual amounts of computing power resources of each type of each node in each cluster of the hybrid cloud platform and the service deployment information of the target model.

2. The method of claim 1, wherein obtaining a service deployment request for a user comprises:

and acquiring the service deployment request of the user from a queue.

3. The method of claim 2, wherein prior to retrieving the user's service deployment request from a queue, the method further comprises:

receiving an external service deployment request of the user;

and storing the service deployment request of the user into the queue.

4. The method of claim 1, wherein obtaining the remaining amount of each type of computing power resource on each node in each cluster created in advance for the user in the hybrid cloud platform comprises:

and acquiring the residual quantity of each type of computing power resource reported by each node in each cluster from a storage unit.

5. The method of claim 4, wherein before obtaining, from a storage unit, a remaining amount of each type of computing power resource reported by each node in each cluster, the method further comprises:

and receiving the residual quantity of each type of computing power resource reported by each node in each cluster and storing the residual quantity in the storage unit.

6. The method of claim 1, wherein deploying the services of the target model in the hybrid cloud platform based on the remaining amount of each type of computing power resource for each node in each cluster of the hybrid cloud platform and the service deployment information of the target model comprises:

Detecting whether deployment of services of the target model of the total number can be implemented in the hybrid cloud platform based on a remaining amount of each type of computing power resource of each node in each cluster of the hybrid cloud platform, a type of first computing power resource on which the target model depends, an amount of first computing power resource on which the target model depends, and a total number of services of the target model are deployed;

if yes, determining a target service deployment strategy of the target model; the target service deployment strategy comprises at least one first cluster identifier and a first number deployed in each first cluster; the sum of the first number of deployments of each of the first clusters is equal to the total number;

generating a target image file corresponding to the target service deployment strategy;

and deploying a corresponding first number of target image files to each first cluster so as to realize the service of deploying the total number of target models in the hybrid cloud platform.

7. The method of claim 6, wherein after obtaining the service deployment request of the user, the method further comprises:

performing format conversion on the target model which is accessed in advance to obtain a target model with a standard format;

The generating the target image file corresponding to the target service deployment strategy comprises the following steps:

generating the target image file corresponding to the target service deployment strategy based on the target model in the standard format and the type of the first computing power resource depended on;

after generating the target image file corresponding to the target service deployment policy, the method further includes:

and storing the generated target image file in an image file warehouse.

8. The method of claim 6, wherein deploying a corresponding first number of target image files to each of the first clusters to enable deployment of the total number of services of the target model in the hybrid cloud platform comprises:

and sending a service deployment command carrying the corresponding first number and the identification of the target image files to the first clusters by calling the application program interfaces of the first clusters so as to enable the corresponding first clusters to pull the corresponding target image files from a hybrid cloud management platform and deploy the first number so as to realize the deployment of the total number of services of the target model in the hybrid cloud platform.

9. The method of claim 6, wherein the method further comprises:

if the deployment of the total number of services of the target model which cannot be implemented in the hybrid cloud platform is detected and determined, determining a first sub-service deployment strategy capable of deploying the target model in the hybrid cloud platform; the first sub-service deployment strategy comprises at least one identifier of a second cluster and a second number deployed in each second cluster; the sum of the second number of deployments of each of the second clusters is less than the total number;

determining the residual quantity of deployment based on the total quantity and the second quantity corresponding to each second cluster; and deploying the services of the residual quantity of deployed models with the same function as the target model in the hybrid cloud platform based on the residual quantity of computing power resources of each type of each node in each cluster of the hybrid cloud platform, the function information of each model in the deployed models and the service deployment information of each model.

10. The method of claim 9, wherein deploying the remaining number of services of the deployed models having the same function as the target model in the hybrid cloud platform based on the remaining amount of each type of computing power resource of each node in each cluster of the hybrid cloud platform, and the function information of each model of the deployed plurality of models, and the service deployment information of each model, comprises:

Acquiring a first alternative model with the same function as the target model based on the function information of the target model and the function information of each model in the deployed multiple models;

detecting whether deployment of services of the first surrogate model of the remaining number can be implemented in the hybrid cloud platform based on a remaining amount of each type of computing power resource of each node in each cluster of the hybrid cloud platform, a type of second computing power resource on which service deployment of the first surrogate model depends, an amount of second target computing power resource on which service deployment of the first surrogate model depends, and a remaining number of deployments; the type of the second computing power resource is different from the type of the first computing power resource;

if yes, determining a second sub-service deployment strategy for deploying the first alternative model in the hybrid cloud platform; the second sub-service deployment strategy comprises at least one identifier of a third cluster and a third number deployed in each third cluster; the sum of the third numbers of the third cluster deployments is equal to the remaining number;

respectively acquiring a first image file corresponding to a first sub-service deployment strategy and a second image file corresponding to a second sub-service deployment strategy;

and respectively deploying the first image files corresponding to the second quantity to each second cluster, and deploying the second image files corresponding to the third quantity to each third cluster so as to realize the service of deploying the total quantity of the target models in the hybrid cloud platform.

11. The method of claim 10, wherein obtaining a second image corresponding to a second sub-service deployment policy comprises:

and detecting and determining that a second image file corresponding to the second sub-service deployment strategy exists in the image file warehouse.

12. The method of claim 11, wherein obtaining a second image corresponding to a second sub-service deployment policy further comprises:

if the fact that the second image file corresponding to the second sub-service deployment strategy does not exist in the image file warehouse is detected and determined, and the second image file corresponding to the second sub-service deployment strategy is generated based on the first alternative model in the standard format and the type of the second computing power resource which is depended on.

13. The method of claim 10, wherein deploying the remaining number of services of the deployed models having the same function as the target model in the hybrid cloud platform based on the remaining amount of each type of computing power resource of each node in each cluster of the hybrid cloud platform, and the function information of each model of the deployed plurality of models, and the service deployment information of each model, further comprises:

if the deployment of the residual quantity of the services of the first alternative model in the hybrid cloud platform can not be implemented is detected and determined, determining a third sub-service deployment strategy for deploying the first alternative model in the hybrid cloud platform; the third sub-service deployment strategy comprises at least one identifier of a fourth cluster and a fourth number deployed in each fourth cluster; the sum of the fourth numbers of the fourth cluster deployments is smaller than the remaining number;

Updating the remaining number of deployments based on a fourth number of deployments of each of the fourth clusters;

acquiring a second alternative model with the same function as the target model based on the function information of the target model and the function information of each model in the deployed multiple models;

detecting whether deployment of services of the second surrogate model of the updated remaining number can be implemented in the hybrid cloud platform based on a remaining amount of each type of computing power resource of each node in each cluster of the hybrid cloud platform, a type of third computing power resource on which service deployment of the second surrogate model depends, an amount of third target computing power resource on which service deployment of the second surrogate model depends, and the updated remaining number; the type of the third computing power resource, the type of the second computing power resource and the type of the first computing power resource are different;

if yes, determining a fourth sub-service deployment strategy for deploying the second alternative model in the hybrid cloud platform; the fourth sub-service deployment strategy comprises at least one identifier of a fifth cluster and a fifth number deployed in each fifth cluster; the sum of the fifth numbers corresponding to the fifth clusters is equal to the updated residual number;

Respectively acquiring a first image file corresponding to a first sub-service deployment strategy, a third image file corresponding to a third sub-service deployment strategy and a fourth image file corresponding to a fourth sub-service deployment strategy;

and respectively deploying a second quantity of second image files corresponding to each second cluster, deploying a fourth quantity of third image files corresponding to each fourth cluster, and deploying a fifth quantity of fourth image files corresponding to each fifth cluster so as to realize the deployment of the total quantity of services of the target model in the hybrid cloud platform.

14. The method of claim 13, wherein the method further comprises:

and if the deployment of the updated residual number of services of the second alternative model cannot be implemented in the hybrid cloud platform is detected and determined, and if other alternative models with the same function as the target model do not exist in the deployed multiple models, determining that the total number of services of the target model cannot be deployed in the hybrid cloud platform.

15. The method of any of claims 1-14, wherein prior to obtaining the remaining amount of each type of computing power resource on each node in each cluster created in advance for the user in a hybrid cloud platform, the method further comprises:

Receiving a cluster creation request of the user; the cluster creation request carries information of a cluster creation platform identifier and various computing power resources to be configured;

based on the creation platform identification of the cluster and the information of various types of computing power resources to be configured, acquiring a plurality of nodes from the corresponding cloud platform to construct the cluster of the user; each of the nodes is capable of providing at least one type of computational power resource.

16. The method of claim 15, wherein the cluster of the user is constructed by acquiring a plurality of nodes in the corresponding cloud platform based on the creation platform identification of the cluster and the information of various types of computing power resources to be configured; after each node is capable of providing at least one type of computational power resource, the method further comprises:

acquiring a collector component from a component warehouse;

a collector component is deployed in each node of the cluster for collecting the remaining amount of each type of computational power resource for each node.

17. The method of claim 15, wherein the cluster of the user is constructed by acquiring a plurality of nodes in the corresponding cloud platform based on the creation platform identification of the cluster and the information of various types of computing power resources to be configured; after each node is capable of providing at least one type of computational power resource, the method further comprises:

Acquiring a display card service component from a component warehouse;

and deploying the display card service component in a node of the cluster providing display card resources.

18. The method of claim 15, wherein the cluster of the user is constructed by acquiring a plurality of nodes in the corresponding cloud platform based on the creation platform identification of the cluster and the information of various types of computing power resources to be configured; after each node is capable of providing at least one type of computational power resource, the method further comprises:

receiving a cluster modification request of the user;

and updating the corresponding clusters based on the cluster modification request of the user.

19. A hybrid cloud management platform, comprising:

the service deployment acquisition module is used for acquiring a service deployment request of a user; the service deployment request carries functional information of a target model to be deployed and service deployment information of the target model;

the residual resource acquisition module is used for acquiring the residual quantity of each type of computing power resource on each node in each cluster created in advance for the user in the hybrid cloud platform;

and the service deployment module is used for deploying the services of the target model in the hybrid cloud platform based on the residual quantity of the computing power resources of each type of each node in each cluster of the hybrid cloud platform and the service deployment information of the target model.

20. An electronic device, comprising:

at least one processor; and

a memory communicatively coupled to the at least one processor; wherein,,

the memory stores instructions executable by the at least one processor to enable the at least one processor to perform the method of any one of claims 1-18.

21. A non-transitory computer readable storage medium storing computer instructions for causing the computer to perform the method of any one of claims 1-18.

22. A computer program product comprising a computer program which, when executed by a processor, implements the method according to any of claims 1-18.