WO2021057825A1 - 一种云系统资源集合的推荐方法、装置及计算设备集群 - Google Patents

一种云系统资源集合的推荐方法、装置及计算设备集群 Download PDF

Info

Publication number
WO2021057825A1
WO2021057825A1 PCT/CN2020/117271 CN2020117271W WO2021057825A1 WO 2021057825 A1 WO2021057825 A1 WO 2021057825A1 CN 2020117271 W CN2020117271 W CN 2020117271W WO 2021057825 A1 WO2021057825 A1 WO 2021057825A1
Authority
WO
WIPO (PCT)
Prior art keywords
system resource
cloud system
resource set
functional component
component
Prior art date
Application number
PCT/CN2020/117271
Other languages
English (en)
French (fr)
Inventor
龚开
王新猴
王学友
Original Assignee
华为技术有限公司
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 华为技术有限公司 filed Critical 华为技术有限公司
Priority to EP20870185.4A priority Critical patent/EP4027240A4/en
Publication of WO2021057825A1 publication Critical patent/WO2021057825A1/zh
Priority to US17/702,224 priority patent/US20220214927A1/en

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F9/00—Arrangements for program control, e.g. control units
    • G06F9/06—Arrangements for program control, e.g. control units using stored programs, i.e. using an internal store of processing equipment to receive or retain programs
    • G06F9/46—Multiprogramming arrangements
    • G06F9/50—Allocation of resources, e.g. of the central processing unit [CPU]
    • G06F9/5061—Partitioning or combining of resources
    • G06F9/5077—Logical partitioning of resources; Management or configuration of virtualized resources
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F9/00—Arrangements for program control, e.g. control units
    • G06F9/06—Arrangements for program control, e.g. control units using stored programs, i.e. using an internal store of processing equipment to receive or retain programs
    • G06F9/46—Multiprogramming arrangements
    • G06F9/50—Allocation of resources, e.g. of the central processing unit [CPU]
    • G06F9/5005—Allocation of resources, e.g. of the central processing unit [CPU] to service a request
    • G06F9/5027—Allocation of resources, e.g. of the central processing unit [CPU] to service a request the resource being a machine, e.g. CPUs, Servers, Terminals
    • G06F9/505—Allocation of resources, e.g. of the central processing unit [CPU] to service a request the resource being a machine, e.g. CPUs, Servers, Terminals considering the load
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F11/00—Error detection; Error correction; Monitoring
    • G06F11/30—Monitoring
    • G06F11/3003—Monitoring arrangements specially adapted to the computing system or computing system component being monitored
    • G06F11/301—Monitoring arrangements specially adapted to the computing system or computing system component being monitored where the computing system is a virtual computing platform, e.g. logically partitioned systems
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F11/00—Error detection; Error correction; Monitoring
    • G06F11/30—Monitoring
    • G06F11/34—Recording or statistical evaluation of computer activity, e.g. of down time, of input/output operation ; Recording or statistical evaluation of user activity, e.g. usability assessment
    • G06F11/3442—Recording or statistical evaluation of computer activity, e.g. of down time, of input/output operation ; Recording or statistical evaluation of user activity, e.g. usability assessment for planning or managing the needed capacity
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F9/00—Arrangements for program control, e.g. control units
    • G06F9/06—Arrangements for program control, e.g. control units using stored programs, i.e. using an internal store of processing equipment to receive or retain programs
    • G06F9/46—Multiprogramming arrangements
    • G06F9/50—Allocation of resources, e.g. of the central processing unit [CPU]
    • G06F9/5005—Allocation of resources, e.g. of the central processing unit [CPU] to service a request
    • G06F9/5027—Allocation of resources, e.g. of the central processing unit [CPU] to service a request the resource being a machine, e.g. CPUs, Servers, Terminals
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F9/00—Arrangements for program control, e.g. control units
    • G06F9/06—Arrangements for program control, e.g. control units using stored programs, i.e. using an internal store of processing equipment to receive or retain programs
    • G06F9/46—Multiprogramming arrangements
    • G06F9/50—Allocation of resources, e.g. of the central processing unit [CPU]
    • G06F9/5061—Partitioning or combining of resources

Definitions

  • This application relates to the field of computer technology, and in particular to a method and device for recommending cloud system resource collections, and a computing device cluster.
  • Cloud-deployed cloud services usually require tenants to rent resources in the cloud in advance, and the collection of resources constituted by these resources is constructed as one Cloud system, in which cloud system can provide users with cloud services by running applications.
  • the cloud only provides a limited set of resources with multiple fixed configurations, and tenants can choose one of these resource sets according to their own needs.
  • these fixed-configuration resource collections usually cannot fully meet the needs of tenants.
  • tenant A needs to rent a resource set, the rent is 1000 yuan/month or less
  • the resource set needs to support 300 concurrent (processing 300 requests at the same time)
  • the central processing unit (CPU) utilization rate is greater than 40%
  • the cloud usually only provides a few fixed-configuration resource collections.
  • Each fixed resource configuration usually only supports one degree of concurrency, and several fixed resource configurations support a limited degree of concurrency. For example, several fixed resource configurations only support 100, 500, and 1000 concurrency.
  • Tenants can only choose a resource set that supports 500 concurrent resources provided by the cloud to ensure that they can meet their own needs. Although the cloud system built by this resource set can better improve cloud services, because the resource set is configured with more resources, the cost is relatively high. High, there is a waste of resources, and it cannot fit the user's own needs well.
  • This application provides a method and device for recommending cloud system resource collections, and a computing device cluster for recommending a more suitable cloud system resource collection for users.
  • this application provides a method for recommending cloud system resource collections.
  • the method can be executed by a recommending device.
  • the method includes: the user can send the cloud system resource collection to the recommending device by triggering methods such as clicking on the display screen or typing in parameters.
  • a recommendation request is used to request a collection of cloud system resources that can support the target cloud service and meet the performance requirements required by the user, where the target cloud service can be provided by at least one functional component, and each functional component implements different functions .
  • the recommendation device receives the cloud system resource collection recommendation request triggered by the user, it can select the target cloud system resource collection from the existing cloud system resource collections that support the target cloud service according to the cloud system resource collection recommendation request.
  • the cloud system resource set includes the at least one functional component, and the recommending device may also feed back the target cloud system resource set to the user.
  • the user does not need to select the required cloud system resource set from a limited set of resources with fixed configurations, only the target cloud service that the cloud system resource set needs to support and the performance requirements that need to be met Inform the recommendation device, which effectively simplifies the user's operation.
  • the recommendation device selects from the existing cloud system resource collections that can support the target cloud service.
  • the selection range is no longer limited to a few cloud system resource collections, and the selection range is larger, which is conducive to choosing a more user-friendly cloud system resource collection. A collection of required cloud system resources.
  • the performance requirements may be expressed as the overall performance requirements of the target cloud system resource set and the performance requirements of each functional component in the target cloud system resource set.
  • the performance requirements include the concurrency that the target cloud system resource set needs to support.
  • it may also include one or more of the following: Central processing that the target cloud system resource set needs to achieve The average utilization rate of the server CPU, the average memory utilization rate that the target cloud system resource set needs to achieve, the average read and write rate of the disk that the target cloud system resource set needs to reach, and the CPU that any functional component in the target cloud system resource set needs to reach The average utilization rate of the target cloud system resource set, the average memory utilization rate of any functional component in the target cloud system resource set, and the average read and write rate of the disk that any functional component in the target cloud system resource set needs to achieve.
  • the performance requirements can include more types of information, which is convenient for the recommending device to select a target cloud system resource set that can truly meet the needs of the user.
  • the cloud system resource set recommendation request may also include the rent range that the target cloud system resource set needs to meet, and the recommending device may choose to satisfy the rent from an existing cloud system resource set that supports the target cloud service.
  • the rent is within the rent range of the target cloud system resource collection.
  • the cloud system resource set recommendation request carries a rent range, so that the selected target cloud system resource set can meet the user's demand for rent.
  • the recommending device when the recommending device selects a cloud system resource set that meets performance requirements based on an existing cloud system resource set that supports the target cloud service, it can use the existing cloud that supports the target cloud service.
  • the system load distribution model corresponding to the system resource set and the component load model corresponding to the at least one functional component are selected from the existing cloud system resource set supporting the target cloud service according to the data determined by the system load distribution model and the component load model A collection of target cloud system resources that meet performance requirements.
  • the system load distribution model corresponding to each cloud system resource collection is used to determine the load status of each functional component in the cloud system collection according to the load status of the cloud system resource collection, and the component load model corresponding to each functional component is used according to the function The load situation of the component determines the performance data of the functional component.
  • the recommendation device can quickly determine the target cloud system resource set with the aid of the system load distribution model and the component load model, which can effectively accelerate the selection efficiency of the target cloud system resource set.
  • the recommending device may generate the system load distribution model and the component load model in advance before determining the target cloud system resource set by using the system load distribution model and the component load model.
  • the recommending device may first collect the structure data and operating data of an existing cloud system resource set that supports the target cloud service, wherein the structure data of any cloud system resource set that supports the target cloud service indicates the cloud system resource set The composition of the functional components and the resource configuration of each functional component; the operation data of any cloud system resource collection that supports the target cloud service is used to indicate the performance of the cloud system resource collection during the operation of the cloud system resource collection and the cloud system resource collection Then, based on the structure data and operating data, the system load distribution model corresponding to the existing cloud system resource set supporting the target cloud service and the component load model corresponding to the functional component are generated.
  • the above description only takes the recommendation device to generate the system load distribution model and the component load model as an example, it is also possible to make other devices generate the system load distribution model and the component load model, and configure the system load distribution model and
  • the generation process of the system load distribution model and the component load model is based on the pre-collected structure data and operating data of the cloud system resource collection that supports the target cloud service, which can ensure the generation of effective system load distribution.
  • the data generated by the model and the component load model, the system load distribution model and the component load model are more accurate.
  • the recommending device when the recommending device generates the system load distribution model corresponding to the existing cloud system resource set supporting the target cloud service and the component load model corresponding to the at least one functional component according to the structure data and the operating data,
  • the structural data and operating data can be integrated into a training set of multiple component load models and a training set of multiple system load distribution models; based on the training set of the component load model, a component load model corresponding to the at least one functional component is generated; based on the system
  • the training set of the load distribution model generates the system load distribution model corresponding to the existing cloud system resource set that supports the target cloud service.
  • the recommendation device can sort structure data and operating data to generate training sets of multiple component load models and multiple system load distribution models, which is convenient for generating more effective system load distribution models and component load models.
  • the embodiments of the present application also provide a recommending device, and the beneficial effects can be referred to the description of the first aspect and will not be repeated here.
  • the device has the function of realizing the behavior in the method example of the first aspect described above.
  • the function can be realized by hardware, or by hardware executing corresponding software.
  • the hardware or software includes one or more modules corresponding to the above-mentioned functions.
  • the structure of the device includes a receiving module, a resource selection module, a sending module, a component load model, a system load distribution model, and a model training module. These modules or models can perform the above-mentioned method examples in the first aspect.
  • an embodiment of the present application also provides a computing device.
  • the computing device includes a processor and a memory, and may also include a communication interface.
  • it also includes a display screen.
  • the processor executes the program instructions in the memory to execute the foregoing first aspect or the method provided in any possible implementation manner of the first aspect.
  • the memory is coupled with the processor, and it stores program instructions and data necessary in the process of determining the traffic flow.
  • the communication interface is used to communicate with other devices, such as receiving a cloud system resource collection recommendation request.
  • the display screen is used to display information to the user under the trigger of the processor, such as displaying a target cloud system resource collection.
  • the present application provides a computing device cluster, which includes at least one computing device.
  • Each computing device includes a memory and a processor.
  • the processor of at least one computing device is configured to access the code in the memory to execute the first aspect or the method provided in any one of the possible implementation manners of the first aspect.
  • the present application provides a non-transitory readable storage medium.
  • the non-transitory readable storage medium executes the foregoing first aspect or any of the first aspects.
  • the storage medium stores the program.
  • the storage medium includes, but is not limited to, volatile memory, such as random access memory, non-volatile memory, such as flash memory, hard disk drive (HDD), and solid state drive (SSD).
  • the present application provides a computer program product.
  • the computer program product includes computer instructions. When executed by a computing device, the computing device executes the foregoing first aspect or any possible implementation of the first aspect Provided method.
  • the computer program product may be a software installation package. In the case where the method provided in the foregoing first aspect or any possible implementation of the first aspect needs to be used, the computer program product may be downloaded and executed on a computing device. Program product.
  • the present application also provides a computer chip, which is connected to a memory, and the chip is used to read and execute a software program stored in the memory, and execute the aforementioned first aspect or any possible implementation of the first aspect. method.
  • Figure 1 is a schematic structural diagram of a recommendation device provided by this application.
  • 2A is a schematic structural diagram of a computing device cluster provided by this application.
  • Figure 2B is a schematic diagram of a system architecture provided by this application.
  • FIG. 3 is a schematic diagram of the training method of the component load model 140 and the system load distribution model 150 provided by this application;
  • FIG. 4 is a schematic diagram of the system recommendation method provided by this application.
  • FIG. 5 is a schematic diagram of a selection interface displayed to the user provided by this application.
  • FIG. 6 is a schematic diagram of another selection interface displayed to the user provided by this application.
  • FIG. 7 is a schematic diagram of another selection interface displayed to the user provided by this application.
  • Figure 8 is a schematic diagram of an interface provided by this application for displaying a target cloud system resource collection to a user.
  • a cloud system resource collection can be used to support a cloud service (a cloud service can also be called a cloud business), and the cloud services that a cloud system resource collection can support depends on each cloud system resource collection.
  • the applications run by each functional component. From the perspective of cloud services, a cloud service can be deployed on different cloud system resource collections, and the performance of each cloud system resource collection affects the quality of the cloud services provided.
  • the performance of any cloud system resource set is related to the types of functional components included in the cloud system resource set and the resource configuration of each functional component.
  • the cloud system resource collection may include one or more functional components, each functional component is used to implement a function, and multiple functional components cooperate to support a cloud service.
  • a cloud system resource collection that supports call services can include three functional components, which are divided into functional components for receiving voice, functional components for processing voice (such as denoising), and output Functional components of speech.
  • a cloud system resource collection that provides call services it can include two functional components, which are divided into functional components for receiving voice and functional components for outputting voice. Both of these two cloud system resource sets can be used to provide call services, but the functional components included are different, and the quality of the cloud services provided is also different.
  • the received voice can be processed to make the final output voice clearer.
  • This cloud system resource collection can Users in noisy environments provide better call services; the latter cloud system resource collection that provides call services does not include functional components for processing voice, omitting voice processing operations, which can speed up the efficiency of voice transmission, although The clarity of the output voice cannot be better guaranteed, but the timeliness of voice transmission can be better guaranteed.
  • This kind of cloud system resource collection is suitable for scenarios that have higher requirements for voice transmission speed.
  • the two cloud system resource sets are also different. Belong to different cloud system resource collections.
  • the resource configuration of any functional component includes any one or more of the following: the number of virtual machines (VM) included in the functional component, the number of containers included in the functional component, and the physical components included in the functional component
  • a virtual machine or a physical machine in the functional component is referred to as an instance of the functional component.
  • an embodiment of the present application provides a method for recommending a cloud system resource set. Before describing the method for recommending the system, the system architecture to which the embodiment of the present application is applicable will be described first.
  • the embodiment of the present application provides a method for recommending cloud system resources, and the method is executed by a recommending device.
  • FIG. 1 a schematic diagram of the structure of a recommendation device provided by an embodiment of this application.
  • the recommendation device 100 includes a receiving module 110, a resource selection module 120, and a sending module 130.
  • the recommendation device 100 may include a component load model 140 and System load distribution model 150.
  • This embodiment of the application does not limit the number of component load models 140 and system load distribution models 150.
  • the number of system load distribution models 150 is related to the number of existing cloud system resource sets.
  • the component load model 140 The number of is related to the number of types of functional components that may be included in the existing cloud system resource collection.
  • the receiving module 110 is configured to receive a cloud system resource collection recommendation request triggered by a user, and the cloud system resource collection recommendation request is used to request a cloud system resource collection that can support a certain type of cloud service.
  • the system load distribution model 150 a cloud system resource collection corresponds to a system load distribution model, and a cloud system resource collection corresponds to a system load distribution model for determining each of the cloud resource collections according to the load situation of the input cloud system resource collection The load status of the functional components.
  • the component load model 140 in the embodiment of the present application, a functional component corresponds to a component load model, and a component load model corresponding to a functional component is used to determine the corresponding performance data of the functional component according to the input load condition of the functional component.
  • the resource selection module 120 is configured to select a cloud system resource collection from an existing cloud system resource collection that supports this type of cloud service. Specifically, the resource selection module may use the component load model 140 and the system load distribution model 150 to select a cloud system resource set from an existing cloud system resource set that supports this type of cloud service.
  • the cloud system resource collection 170 includes a functional component 171, a functional component 172, and a functional component 173.
  • the functional component 171 includes a PM and a VM.
  • the PM in the functional component 171 includes two CPUs, memory and disks, and the functional component 171
  • the memory and disk capacities of the PM in the are 128 gigabytes (Gigabyte, GB) and 500GB, respectively.
  • the functional component 172 includes a PM, and the PM in the functional component 172 includes a CPU, memory, and disk.
  • the memory and disk capacities of the VM in the functional component 172 are 256 GB and 1 terabyte (Terabyte, T), respectively.
  • the functional component 173 includes a VM and three containers.
  • the VM in the functional component 173 includes a VCPU, memory, and disk.
  • the memory and disk capacities of the VM in the functional component 173 are 256GB and 1T, respectively.
  • the cloud system resource collection 180 includes a functional component 181, a functional component 182, and a functional component 183.
  • the functional component 181 includes a PM.
  • the PM in the functional component 181 includes two CPUs, memory and disk.
  • the capacities are 256GB and 500GB respectively.
  • the functional component 182 includes one VM.
  • the VM in the functional component 182 includes two VCPUs, memory and disk.
  • the memory and disk capacities of the VM in the functional component 182 are 256GB and 1T, respectively.
  • the functional component 183 includes two VMs.
  • Each VM in the functional component 183 includes a VCPU, memory, and disk.
  • the memory and disk capacities of each VM in the functional component 183 are 128 GB and 500 GB, respectively.
  • the sending module 130 is configured to send the target cloud system resource set selected by the resource selection module 120 to the user.
  • the component load model 140 and the system load distribution model 150 in the recommender device 100 may be trained by devices other than the recommender device 100 and configured in the recommender device.
  • the recommendation device 100 may include a model training module 160, which is used to train the component load model and the system load distribution model, and to use the trained component load model and the system load distribution model. It is arranged in the recommendation device 100.
  • a computer cluster provided by the embodiments of the present application includes at least one computing device 10, and each computing device 10 passes through The communication network establishes a communication path.
  • Each computing device 10 includes a bus 101, a processor 102, a communication interface 103, and a memory 104.
  • the computing device 10 may also include a display screen 105.
  • the processor 102, the memory 104, and the communication interface 103 communicate through a bus 101.
  • the processor 102 may be composed of one or more general-purpose processors, such as a central processing unit (CPU), or a combination of a CPU and a hardware chip.
  • the aforementioned hardware chip may be an application-specific integrated circuit (ASIC), a programmable logic device (PLD), or a combination thereof.
  • ASIC application-specific integrated circuit
  • PLD programmable logic device
  • the above-mentioned PLD may be a complex programmable logic device (CPLD), a field-programmable gate array (FPGA), a general array logic (generic array logic, GAL), or any combination thereof.
  • the memory 104 may include a volatile memory (volatile memory), such as a random access memory (random access memory, RAM).
  • volatile memory such as a random access memory (random access memory, RAM).
  • the memory 104 may also include non-volatile memory (NVM), such as read-only memory (ROM), flash memory, hard disk drive (HDD), or solid-state drive (solid-disk drive, HDD). state drive, SSD).
  • NVM non-volatile memory
  • ROM read-only memory
  • HDD hard disk drive
  • solid-state drive solid-disk drive
  • state drive SSD
  • the memory 104 may also include a combination of the above-mentioned types.
  • the memory 104 stores executable codes.
  • the processor 102 can read the executable codes in the memory 104 to realize functions. It can also communicate with other computing devices through the communication interface 103.
  • the processor 102 can also trigger the display 105 to report to the user. Display information, such as the information included in the interface as shown in FIGS. 5 to 8 in the embodiment of the present application.
  • the processor 102 may implement one or more modules of the recommendation apparatus 100 (such as the receiving module 110, the resource selection module 120, the sending module 130, the component load model 140, the system load distribution model 150, and the model training module 160).
  • the memory 104 stores the receiving module 110, the resource selection module 120, the sending module 130, the component load model 140, and the system load distribution model 150 in the recommendation device 100 in the memory 104. .
  • the processors 102 in the multiple computing devices 10 may coordinate work to execute the cloud system resource collection recommendation method provided in the embodiment of the present application.
  • a system architecture provided by this embodiment of the application includes a client 200 and a cloud device 300 on which a recommendation device is deployed.
  • the client 200 and the cloud device 300 are connected via a network, and the cloud device 300 is located in a cloud environment and can be a server or a virtual machine deployed in a cloud data center.
  • the recommended device is deployed on a cloud device 300 as an example.
  • the recommended device It can be deployed on multiple cloud devices 300 in a distributed manner.
  • the client 200 includes a bus 201, a processor 202, a communication interface 203, and a memory 204.
  • the client 200 may also include a display screen 205.
  • the processor 202, the memory 204, and the communication interface 203 communicate through a bus 201.
  • the memory 204 stores executable code, and the processor 202 can read the executable code in the memory 204 to implement functions.
  • the processor 202 may also trigger the display screen 205 to display information to the user, such as the information included in the interface shown in FIGS. 5 to 8 in the embodiment of the present application.
  • the processor 202 may also communicate with the cloud device through the communication interface 203.
  • the processor 202 may prompt the user to input information through the display screen, and the information may be information carried in the cloud system resource collection recommendation request, such as the type of cloud service that the cloud system resource collection needs to support and the performance requirements of the cloud system resource collection;
  • the cloud system resource collection recommendation request carrying information such as the type of cloud service that the cloud system resource collection needs to support and the performance requirements of the cloud system resource collection is sent to the cloud device 300 through the communication interface 203.
  • the cloud device 300 includes a bus 301, a processor 302, a communication interface 303, and a memory 304.
  • the processor 302, the memory 304, and the communication interface 303 communicate through a bus 301.
  • Executable code is stored in the memory 304, and the processor 302 can read the executable code in the memory 304 to implement functions, and can also communicate with the client 200 through the communication interface 303.
  • the processor 302 of multiple cloud devices 300 can cooperate to realize the function of the recommending apparatus 100.
  • the memory 304 stores a receiving module 110, a resource selection module 120, and a sending module.
  • the processor 302 After the processor 302 receives the cloud system resource collection recommendation request from the client 200 through the communication interface 303, it can call the module stored in the memory 304 to implement the cloud system resource collection recommendation method provided by the embodiment of the present application, and select the cloud system resource collection for the user The processor 302 may send the selected cloud system resource set to the client 200 through the communication interface 303.
  • the processor 202 in the client 200 may trigger the display screen 205 to display the selected cloud system resource set to the user.
  • Each functional component corresponds to a component load model 140.
  • the component load model 140 corresponding to a functional component is based on the load situation of the functional component and the functional component in the collection of various cloud system resources. In this way, the component load model 140 corresponding to the functional component can determine the performance data of the functional component according to the input load condition of the functional component.
  • the load condition of the functional component can be characterized by the concurrency of the functional component (that is, the number of internal requests received by the functional component at a certain moment), or it can be characterized by other parameters used to reflect the load condition of the functional component.
  • the performance data of the functional component includes any one or more of the following: the average utilization of the CPU (including VCPU) of the functional component, the average utilization of the memory of the functional component, the average read/write rate of the disk in the functional component (input/output operations per second, IOPS).
  • IOPS input/output operations per second
  • the CPU (including VCPU) average utilization of the functional component is equal to the average of the CPU utilization of all instances (an instance is a virtual machine or a physical machine) in the functional component, and the CPU utilization of an instance reflects a period of time The proportion of the CPU running time to the total time.
  • the average memory utilization rate of a functional component is the ratio of the occupied memory of the functional component to the total memory of the functional component.
  • the average IOPS of the disk in the functional component is equal to the average of the IOPS of the disks of all instances in the functional component, and the IOPS of the disk of an instance is the number of reads and writes per second of the disk.
  • the performance data of the above three functional components are only examples, and the embodiments of the present application do not limit the specific data types of the performance data of the functional components. All data that can reflect the performance of the functional components is applicable to the embodiments of the present application.
  • the other type is the system load distribution model 150.
  • a cloud system resource collection corresponds to a system load distribution model 150
  • a cloud system resource collection corresponds to a system load distribution model 150. It is based on the existing load situation of the cloud system resource collection. Is generated with the load status of each functional component in the cloud system resource set. In this way, the system load distribution model 150 corresponding to the cloud system resource set can determine the load of each functional component in the cloud system resource set according to the load situation of the cloud system resource set. Load situation.
  • the load situation of the cloud system resource collection can be the concurrency of the cloud system resource collection (that is, the number of external requests received by the cloud system at a certain moment), and other parameters that reflect the load situation of the cloud system resource collection can also be used Characterization.
  • the concurrency of the cloud system resource set is used to characterize the load situation of the cloud system resource set.
  • the requests involved in the concurrency in the cloud system resource collection are external requests, and the so-called external requests refer to requests initiated by devices or users outside the cloud system resource collection. For example, it is a request triggered when a user needs a cloud service.
  • the requests involved in the concurrency of functional components are internal requests initiated by the functional components.
  • the so-called internal requests refer to requests initiated by the functional components within the cloud system resource collection.
  • a cloud service provided by a cloud system resource collection for users is a web service.
  • the cloud system resource collection can be divided into three functional components, which are divided into front-end components, back-end components, and database components.
  • the conversion operations of the back-end webpage when the user clicks on the pictures, links, and icons in the webpage are executed by the back-end component; and the database component saves the relevant signals of the webpage, such as text, pictures, and the webpage.
  • the page information that can be redirected to the page.
  • the front-end component is triggered to initiate an application programming interface (API) request to the back-end component.
  • the API request is triggered by the user and is an external request; the back-end After the component receives the API request of the front-end component, it initiates a data query request to the database component.
  • the back-end component returns the query result fed back by the database component to the front-end component.
  • the data query request is an API request triggered by the functional component in the cloud system resource collection. It is an internal request.
  • the recommendation device is configured with the above-mentioned two types of different models.
  • the resource selection module 120 can be based on these two types of different models, according to the user's requirements for the cloud system resource collection (such as the cloud service type supported by the cloud system resource collection and the user's The performance requirements of the cloud system resource set), and the cloud system resource set that meets the demand is selected.
  • the embodiments of this application do not limit the type of the model.
  • the two different types of models can be neural network models or models established based on statistical machine learning algorithms, such as polynomial regression algorithms, ridge regression algorithms, and elastic net regression algorithms. The model established by the algorithm. The specific details of the resource selection module 120 selecting a cloud system resource set that meets the demand will be described below.
  • the device for training these two types of neural network models is not limited, and the recommendation device may include the model training module 160 to train the two types of neural network models. It may also be that other devices train the two types of neural network models, and after the training is completed, the two types of neural network models that have been trained are configured in the recommendation device. In this embodiment of the application, only the model training module 160 is used. Take the training of these two types of neural network models as an example.
  • FIG. 3 it is a schematic diagram of the training method of the component load model 140 and the system load distribution model 150.
  • the method includes:
  • Step 301 The model training module 160 collects data from the existing cloud system resource collection.
  • the collected data can be divided into two types, one is the structure data of the cloud system resource collection, and the other is the operation data of the cloud system resource collection .
  • the structural data of the cloud system resource collection is used to indicate the cloud services supported by the cloud system resource collection, the composition of the functional components in the cloud system resource collection and the resource configuration of each functional component, and the structural data of the cloud system resource collection Including any one or more of the following: cloud services supported by each cloud system resource collection, functional components included in each cloud system resource collection, the number of virtual machines in each functional component in each cloud system resource collection, The number of physical machines, the number of CPUs (including VCPUs) of each virtual machine/physical machine, the memory capacity of each virtual machine/physical machine, and the disk capacity of each virtual machine/physical machine.
  • the operating data of the cloud system resource collection is the data collected during the operation of the cloud system resource collection.
  • the operating data can characterize the overall performance of the cloud system resource collection, as well as the specific functions within the cloud system resource collection.
  • the performance of the components and the operating data of the cloud system resource collection include any one or more of the following: API call records of the cloud system resource collection, the CPU utilization, memory utilization, and memory utilization of each instance in each functional component in the cloud system resource collection And disk IOPS.
  • the embodiment of the application does not limit the data types included in the performance data of the cloud system resource collection.
  • the data types included in the performance data of the cloud system resource collection are related to the component load model 140 that needs to be trained.
  • the performance of the cloud system resource collection collected by the model training module 160 when other performance data can be determined in addition to the average utilization of the CPU (including VCPU), the average utilization of the memory of the functional component, and the average read and write rate of the disk in the functional component
  • the data also needs to include the same type of performance data, or data that can calculate the other performance data.
  • the embodiment of the present application only takes the performance data of the cloud system resource collection including the CPU utilization rate, memory utilization rate, and disk IOPS of each instance in each functional component as an example for description.
  • the API call record records the execution information of each API request processed by the cloud system resource collection (including any external request executed by the cloud system resource collection and any internal request triggered within the cloud system resource collection), such as the API request Identity (ID), trigger time of API request, end time of API request, return result of API request, etc.
  • ID the API request Identity
  • trigger time of API request the time of API request
  • end time of API request the time of API request
  • return result of API request etc.
  • Step 302 The model training module 160 organizes the collected data, and generates a training set of the component load model 140 and a training set of the system load distribution model 150 respectively.
  • each component load model 140 corresponds to a functional component
  • each system load distribution model 150 corresponds to a cloud system resource set; that is, when there are multiple functional components, the model training module 160 needs to generate each functional component The training set of the corresponding component load model 140; when there are multiple cloud system resource sets, the model training module 160 needs to generate a training set of the component load model 140 corresponding to each cloud system resource set.
  • the training set of component load model 140 corresponding to functional component A includes the concurrency of functional component A and functional component A.
  • the number of medium instances, the number of CPUs (including VCPUs) of a single instance in functional component A, the memory capacity of a single instance in functional component A, the disk capacity of a single instance in functional component A; the average of CPUs (including VCPU) in functional component A Utilization, the average utilization of the memory in functional component A, the average read and write rate of the disk in functional component A.
  • the number of CPUs, memory capacity, and disk capacity of each instance of functional component A are the same as an example. If there are multiple different instances in functional component A, the number of CPUs in each instance, If the memory capacity or the disk capacity is different, the training set of the component load model 140 needs to include the number of CPUs, memory capacity, or disk capacity of each instance.
  • the model training module 160 needs to further organize the collected data. For example, the model training module 160 obtains the number of internal requests received by the functional components in the cloud system resource collection at time 1 from the API call record Determine the concurrency of functional component A at time 1; the model training module 160 analyzes the structural data of the cloud system resource collection to determine the number of instances in functional component A, the number of CPUs (including VCPUs) of a single instance in functional component A, and the number of functional components in functional component A The memory capacity of a single instance and the disk capacity of a single instance in functional component A; the model training module 160 determines the CPU utilization, memory utilization, and disk IOPS of each instance in each functional component in the cloud system resource set at time 1, through The method of solving the average value determines the average utilization rate of CPU (including VCPU) in functional component A at time 1, the average utilization rate of memory in functional component A, and the average read and write rate of disk in functional component A.
  • the training set of the component load model 140 can be divided into multiple sets of training data according to time. It is necessary to ensure that when the component load model 140 is subsequently trained, the input data and the preset output data correspond to the same time.
  • the model training module 160 can number the supported cloud services, functional components, and instances of the functional components to form an organization tree, and integrate the collected data into the organization tree. If the corresponding data is queried, the training set of the component load model 140 can be generated quickly.
  • the training set of the system load distribution model 150 corresponding to cloud system resource set 1 includes cloud system resources.
  • the concurrency of set 1 and the concurrency of cloud system resource set 1 are the concurrency of each functional component included in cloud system resource set 1.
  • the model training module 160 needs to further organize the collected data. For example, the model training module 160 obtains the internal information received by each functional component in the cloud system resource set at time 1 from the API call record. The number of requests determines the concurrency of each functional component A at time 1, and the number of external requests received by the cloud system resource collection at time 1 determines the concurrency of the cloud system resource collection at time 1.
  • the training set of the system load distribution model 150 can be divided into multiple sets of training data according to time. It is necessary to ensure that when the system load distribution model 150 is subsequently trained, the input data and the preset output data correspond to the same time.
  • step 303 is the training of the component load model 140 Process
  • step 304 is the training process of the system load distribution model 150.
  • Step 303 The model training module 160 trains the component load model 140 according to the training set of the component load model 140.
  • the model training module 160 sets the training set of the component load model 140 corresponding to functional component A to the concurrency of functional component A, the number of instances in functional component A, and the CPU of each instance.
  • the number of (including VCPU), the memory capacity of each instance, and the disk capacity of each instance are input values, the average utilization of CPU (including VCPU) in functional component A, the average utilization of memory in functional component A, functional components
  • the average read and write rate of the disk in A is the output value, and the component load model 140 corresponding to functional component A is trained in a supervised learning manner until the loss function converges.
  • Step 304 The model training module 160 trains the system load distribution model 150 according to the training set of the system load distribution model 150.
  • the model training module 160 takes the concurrency of cloud system resource set 1 in the training set of the system load distribution model 150 corresponding to cloud system resource set 1 as the input value, and Under the input concurrency of cloud system resource set 1, the concurrency of each functional component in cloud system resource set 1 is used as the output value, and the system load distribution model 150 corresponding to cloud system resource set 1 is trained by supervised learning until the loss The function converges.
  • FIG. 4 it is a schematic diagram of a system recommendation method provided by an embodiment of this application, and the method includes:
  • Step 401 The receiving module 110 receives a cloud system resource collection recommendation request triggered by a user.
  • the cloud system resource collection recommendation request includes the user's demand for the cloud system resource collection (referred to as user demand for short), and the demand includes the cloud system resource collection.
  • the cloud service that needs to be supported can also include any one or more of the following information: the degree of concurrency that the cloud system resource collection needs to support, the average CPU utilization that the cloud system resource collection needs to achieve, and the cloud system resource collection needs to achieve.
  • FIG. 5 it is a selection interface displayed on the display screen of the client 200 or the display screen of the computing device 10.
  • the user is prompted to select the cloud service required by the cloud system resource collection. The user can select by clicking the corresponding option.
  • Fig. 6 is another selection interface displayed to the user by the display screen in the client 200 or the display screen in the computing device 10, the user can load the cloud system collection (for example, the cloud system resource collection needs to support Concurrency) and performance (the average utilization of the CPU required by the cloud system resource collection, the average memory utilization required by the cloud system resource collection, and the average disk read and write rate required by the cloud system resource collection) are set.
  • the options in Figure 6 also include defaults. When the user has no requirements for the load and performance of the cloud system collection, or cannot determine the load and performance that the cloud system collection needs to achieve, the default can be selected.
  • This "default” option corresponds to the concurrency that the cloud system resource collection needs to support, the average CPU utilization that the cloud system resource collection needs to achieve, the average memory utilization that the cloud system resource collection needs to reach, or the disk that the cloud system resource collection needs to reach
  • the average read and write rate is a preset value.
  • the embodiment of this application does not limit the way to determine the preset value. It can be an empirical value, or it can be a set of data corresponding to all existing cloud system resource sets supporting the cloud service average value.
  • Users can also only check the concurrency required by the cloud system resource collection, the average CPU utilization required by the cloud system resource collection, the average memory utilization required by the cloud system resource collection, and the disk that the cloud system resource collection needs to reach.
  • One or more parameters in the average read/write rate can be set. For the parameters that are not set by the user, it can be defaulted to have no specific requirements, or the parameters that have not been set by the user can be set to a preset value.
  • the preset value can be an empirical value or It can be a value matched according to the parameters set by the user.
  • users can not only support the concurrency required by the cloud system resource collection, the average CPU utilization required by the cloud system resource collection, the average memory utilization required by the cloud system resource collection, and the cloud system
  • One or more parameters in the average read and write rate of the disk that the resource collection needs to reach can be set, and it can also be specific to the inside of the cloud system resource collection, and the CPU required to be reached by one or more functional components in the cloud system resource collection
  • Average utilization, one or more parameters in the average utilization of the memory required by one or more functional components and the average read/write rate of the disk required by one or more functional components are set.
  • the premise of the setting is that the user can first determine the types of functional components included in the cloud system resource set.
  • the user can configure cloud system resources in the interface displayed to the user by the display screen in the client 200 or the display screen in the computing device 10
  • the collection recommendation request requested cloud system resource collection supporting web services includes front-end components, back-end components, and database components.
  • the user configures the average utilization rate of the CPU to be achieved by one or more functional components in the cloud system resource set on the display screen in the client 200 or the display screen in the computing device 10 displayed to the user, one or more One or more parameters of the average memory utilization rate required by each functional component and the average read/write rate of the disk that one or more functional components must achieve.
  • the user can set the display screen in the client 200 or the computer in the computing device 10
  • the average utilization rate of the CPU, the average utilization rate of the memory and the average read/write rate of the disk to be achieved by configuring the front-end components on the interface displayed by the display screen to the user.
  • FIG. 7 it is another selection interface displayed to the user by the display screen in the client 200 or the display screen in the computing device 10, and the user can select the cloud system resource collection payment mode (monthly settlement, daily settlement, and annual settlement) And the specific cost.
  • the cloud system resource collection payment mode monthly settlement, daily settlement, and annual settlement
  • the user can trigger the client 200 to send a cloud system resource collection recommendation request to the recommendation device by clicking the corresponding option.
  • Step 402 After the receiving module 110 receives the system resource set recommendation request, the resource selection module 120 may select a target cloud system resource set that meets the user's needs based on the existing cloud system resource sets that support the cloud service.
  • the resource selection module 120 may use the pre-trained system load distribution model 150 corresponding to the cloud system resource set supporting the cloud service and the component load model 140 corresponding to each functional component to select a cloud system resource set that meets the needs of the user.
  • the average CPU utilization rate illustrates the method of selecting cloud system resource sets that meet user needs.
  • Step 1 The resource selection module 120 may first determine the existing cloud system resource set corresponding to the system load distribution model 150 used to support the cloud service based on the degree of concurrency that the cloud system resource set needs to support. The concurrency of each functional component in the cloud system resource collection of the cloud service.
  • the resource selection module 120 may query the system load distribution corresponding to the existing cloud system resource set used to support the cloud service according to the cloud service required by the cloud system resource set carried in the cloud system resource set recommendation request. Model 150.
  • the resource selection module 120 can input the concurrency that the cloud system resource set needs to support into the system corresponding to the cloud system resource set In the load model, the concurrency of each functional component in the cloud system resource collection can be determined.
  • cloud system resource collection 1 Including functional component A, functional component B, and functional component C
  • cloud system resource collection 2 includes functional component A and functional component B
  • cloud system resource collection 3 includes functional component A, functional component B, functional component C, and functional component D.
  • the resource selection module 120 inputs the concurrency (300) that the cloud system resource collection needs to support into the system load distribution model f1 (system concurrency) to obtain the concurrency f1 (300)-A of the functional component A and the concurrency f1 of the functional component B (300)-B and functional component C concurrency f1(300)-C.
  • the resource selection module 120 inputs the concurrency (300) that the cloud system resource collection needs to support into the system load distribution model f2 (300) to obtain the concurrency of functional component A f2 (300)-A and the concurrency of functional component B f2 (300) )-B.
  • the resource selection module 120 inputs the concurrency (300) that the cloud system resource collection needs to support into the system load distribution model f3 (system concurrency) to obtain the function component A concurrency f3 (300)-A and the function component B concurrency f3 (300)-B, functional component C concurrency f3(300)-C, and functional component D concurrency f4(300)-D.
  • system concurrency system concurrency
  • Step 2 Based on the component load model 140, the resource selection module 120 determines the existing cloud system resources used to support the cloud service according to the concurrency of each functional component in the existing cloud system resource set used to support the cloud service The average CPU utilization of each functional component in the set.
  • the resource selection module 120 may input the concurrency of each functional component in the cloud system resource set calculated in the first step into each corresponding function.
  • the average utilization rate of the CPU of each functional component is determined.
  • the resource selection module 120 can input the functional component A concurrency degree f1(300)-A determined in the first step into the component load model Fa corresponding to the functional component A, and set the functional component B concurrency f1( 300)-B is input to the component load model Fb corresponding to functional component B, and the concurrency f1(300)-C of function component C is input to the component load model Fc corresponding to functional component C.
  • the output value of the component load model Fa corresponding to the functional component A the average CPU utilization of the functional component A can be determined.
  • the average CPU utilization of the functional component B can be determined.
  • the functional component C The output value of the corresponding component load model Fc can determine the average CPU utilization of the functional component C.
  • the resource selection module 120 can input the concurrency f2(300)-A of the functional component A determined in the first step into the component load model Fa corresponding to the functional component A, and set the concurrency f2( 300)-B is input to the component load model Fb corresponding to the functional component B.
  • the average CPU utilization of the functional component A can be determined according to the output value of the component load model Fa corresponding to the functional component A
  • the average CPU utilization of the functional component B can be determined according to the output value of the component load model Fb corresponding to the functional component B.
  • the resource selection module 120 can input the concurrency f3(300)-A of the functional component A determined in the first step into the component load model Fa corresponding to the functional component A, and set the concurrency f3( 300)-B is input to the component load model Fb corresponding to function component B, and the concurrency degree of function component C f3(300)-C is input to the component load model Fc corresponding to function component C, and the concurrency degree of function component D is f4(300) -D is input to the component load model Fd corresponding to the functional component D.
  • the average CPU utilization of the functional component A can be determined.
  • the average CPU utilization of the functional component B can be determined.
  • the functional component C The output value of the corresponding component load model Fc can determine the average CPU utilization of the functional component C, and the average CPU utilization of the functional component D can be determined according to the output value of the component load model Fd corresponding to the functional component D.
  • Step 3 The resource selection module 120 selects a cloud system resource set whose average utilization rate of the CPU of each functional component in the cloud system resource set is greater than the average utilization rate of the CPU that the cloud system resource set needs to support, and the cloud system resource set is Target resource collection.
  • the resource selection module 120 may also select a cloud system resource set whose average value of the average CPU utilization of each functional component in the cloud system resource set is greater than the average CPU utilization that the cloud system resource set needs to support, and the cloud system resource set is the target Resource collection.
  • the resource selection module 120 may further filter the target cloud system resource set that meets the rent range from the cloud system resource set that meets the average utilization rate of the cloud system resource set CPU. Of course, the resource selection module 120 may also remove cloud system resource sets that do not meet the rent range before performing the first step, and execute the first step only for cloud system resource sets that meet the rent range.
  • Step 403 The sending module 130 feeds back the selected target cloud system resource set to the user.
  • the sending module 130 may feed back the selected target cloud system resource set to the client 200.
  • the client 200 receives the cloud system resource set fed back by the sending module 130, and may push the target cloud system resource set fed back by the sending module 130 to the user. For users to view.
  • the sending module 130 may push the selected target cloud system resource set to the user through the display screen in the computing device 10 for the user to view.
  • the client 200 may list the cloud system resource collection fed back by the sending module 130 and present it to the user.
  • the client 200 can arrange the target cloud system resource set fed back by the sending module 130 in an ascending or descending order of price.
  • the disclosed system, device, and method may be implemented in other ways.
  • the device embodiments described above are merely illustrative, for example, the division of the units is only a logical function division, and there may be other divisions in actual implementation, for example, multiple units or components may be combined or It can be integrated into another system, or some features can be ignored or not implemented.
  • the displayed or discussed mutual coupling or direct coupling or communication connection may be indirect coupling or communication connection through some interfaces, devices or units, and may be in electrical, mechanical or other forms.
  • the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the objectives of the solutions of the embodiments.

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Software Systems (AREA)
  • Physics & Mathematics (AREA)
  • General Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Quality & Reliability (AREA)
  • Computing Systems (AREA)
  • Computer Hardware Design (AREA)
  • Mathematical Physics (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)

Abstract

一种云系统资源集合的推荐方法、装置及计算设备集群。推荐装置接收到由用户触发的用于请求获取能够支持目标云服务、且满足性能需求的云系统资源集合的云系统资源集合推荐请求（401）后，根据云系统资源集合推荐请求（401），从已有的支持目标云服务的云系统资源集合中选择满足性能需求的目标云系统资源集合（402），并将目标云系统资源集合反馈给用户（403）。用户并不需要从有限的几种固定配置的资源集合中，选择所需的云系统资源集合，只需将该云系统资源集合所需支持的目标云服务和性能需求告知推荐装置，简化了用户操作。由推荐装置从已有的、能够支持目标云服务的云系统资源集合中进行选择，选择范围更大，有利于选择出更加贴合用户的需求的云系统资源集合。

Description

一种云系统资源集合的推荐方法、装置及计算设备集群 技术领域
本申请涉及计算机技术领域，尤其涉及一种云系统资源集合的推荐方法、装置及计算设备集群。
背景技术
随着云计算技术的高速发展，云端可以部署越来越多的云服务，以供给用户使用，而云端部署的云服务，通常需要租户预先在云端租用资源，这些资源构成的资源集合构建为一个云系统，在该云系统可以通过运行应用程序为用户提供云服务。
目前，云端只会提供有限的几种多种固定配置的资源集合，租户可以根据自身需求从这几种资源集合中选择其中一种。但是这些固定配置的资源集合通常不能完全贴合租户的自身需求。举例来说，租户A需要租用一个资源集合，租金为1000元/月以下、该资源集合需要可支持300并发(同时处理300个请求)、以及中央处理器(central processing unit，CPU)利用率大于40％，而云端提供的通常只提供几个固定配置的资源集合。每个固定的资源配置通常也只能支持一种并发度，几种固定的资源配置支持的并发度有限，例如几种固定的资源配置仅支持100并发、500并发、1000并发。
租户只能选择云端提供的支持500并发的资源集合才能保证达到自身需求，该资源集合构建的云系统虽然能够较好的提高云服务，但是由于该资源集合中配置了较多的资源，成本较高，存在资源浪费，不能很好的贴合用户的自身需求。
发明内容
本申请提供一种云系统资源集合的推荐方法、装置及计算设备集群，用以为用户推荐较为合适的云系统资源集合。
第一方面，本申请提供了一种云系统资源集合推荐方法，该方法可以由推荐装置执行，该方法包括：用户可以通过点击显示屏或键入参数等触发方式，向推荐装置发送云系统资源集合推荐请求，用以请求获取能够支持目标云服务的、且满足用户所要求的性能需求的云系统资源集合，其中，目标云服务能由至少一个功能组件配合提供，每个功能组件实现的功能不同。当推荐装置接收到接收由用户触发的云系统资源集合推荐请求后，可以根据云系统资源集合推荐请求,从已有的支持目标云服务的云系统资源集合中选择目标云系统资源集合，该目标云系统资源集合包括该至少一个功能组件，推荐装置还可以将该目标云系统资源集合反馈给用户。
通过上述方法，用户并不需要从有限的几种固定配置的资源集合中，选择所需要的云系统资源集合，只需将该云系统资源集合所需支持的目标云服务以及需要满足的性能需求告知推荐装置，有效的简化了用户的操作。由推荐装置从已有的、能够支持目标云服务的云系统资源集合中进行选择，选择范围不再是有限的几种云系统资源集合，选择范围更大，有利于选择出更加贴合用户的需求的云系统资源集合。
在一种可能的实现方式中，性能需求可以表现为对该目标云系统资源集合整体的性能需求以及该目标云系统资源集合中各个功能组件的性能需求。示例性的，性能需求包括目标云系统资源集合需支持的并发度，除了目标云系统资源集合需支持的并发度还可以包括下列的一种或多种：目标云系统资源集合需达到的中央处理器CPU的平均利用率、目标云系统资源集合需达到的内存平均利用率、目标云系统资源集合需达到的磁盘的平均读写速率、以及目标云系统资源集合中任一功能组件需达到的CPU的平均利用率、目标云系统资源集合中任一功能组件需达到的内存平均利用率以及目标云系统资源集合中任一功能组件需达到的磁盘的平均读写速率。
通过上述方法，性能需求中可以包括信息种类较多，便于推荐装置选择出能够真实贴近用户需求的目标云系统资源集合。
在一种可能的实现方式中，云系统资源集合推荐请求还可以包括目标云系统资源集合需满足的租金范围，推荐装置可以从已有的支持所述目标云服务的云系统资源集合中选择满足租金处于租金范围内目标云系统资源集合。
通过上述方法，云系统资源集合推荐请求携带有租金范围，能够使得选择出的目标云系统资源集合能够满足用户对租金的需求。
在一种可能的实现方式中，推荐装置在基于已有的支持目标云服务的云系统资源集合中选择满足性能需求的云系统资源集合时，可以利用已有的用于支持目标云服务的云系统资源集合对应的系统负载分配模型以及该至少一个功能组件对应的组件负载模型，根据系统负载分配模型以及组件负载模型所确定的数据，从已有的支持目标云服务的云系统资源集合中选择满足性能需求的目标云系统资源集合。
其中，每个云系统资源集合对应的系统负载分配模型用于根据云系统资源集合的负载情况确定云系统集合中每个功能组件的负载情况，每个功能组件对应的组件负载模型用于根据功能组件的负载情况确定功能组件的性能数据。
通过上述方法，推荐装置借助系统负载分配模型以及组件负载模型可以较为快捷的确定目标云系统资源集合，能够有效加快目标云系统资源集合的选择效率。
在一种可能的实现方式中，推荐装置在借助系统负载分配模型以及组件负载模型确定目标云系统资源集合之前，可以预先生成系统负载分配模型以及组件负载模型。示例性的，推荐装置可以先收集已有的支持目标云服务的云系统资源集合的结构数据以及运行数据，其中，任一支持目标云服务的云系统资源集合的结构数据指示该云系统资源集合中功能组件的组成以及每个功能组件的资源配置；任一支持目标云服务的云系统资源集合的运行数据用于指示云系统资源集合在运行过程中云系统资源集合的性能以及云系统资源集合中各个功能组件的性能；之后，再根据结构数据以及运行数据生成已有的支持目标云服务的云系统资源集合对应的系统负载分配模型以及功能组件对应的组件负载模型。上述说明中仅是以推荐装置生成系统负载分配模型以及组件负载模型为例，也可以使其他装置生成系统负载分配模型以及组件负载模型，将系统负载分配模型以及组件负载模型配置在推荐装置中。
通过上述方法，系统负载分配模型以及组件负载模型的生成过程中是基于预先收集的已有的支持目标云服务的云系统资源集合的结构数据以及运行数据进行的，能够保证生成有效的系统负载分配模型以及组件负载模型，系统负载分配模型以及组件负 载模型所产生的数据更精确。
在一种可能的实现方式中，推荐装置在根据结构数据以及运行数据生成已有的支持目标云服务的云系统资源集合对应的系统负载分配模型以及该至少一个功能组件对应的组件负载模型时，可以将结构数据以及运行数据整合为多个组件负载模型的训练集以及多个系统负载分配模型的训练集；基于组件负载模型的训练集，生成该至少一个功能组件对应的组件负载模型；基于系统负载分配模型的训练集，生成已有的支持目标云服务的云系统资源集合对应的系统负载分配模型。
通过上述方法，推荐装置可以对结构数据以及运行数据进行整理，产生多个组件负载模型的训练集以及多个系统负载分配模型的训练集，便于生成较为有效的系统负载分配模型以及组件负载模型。
第二方面，本申请实施例还提供了一种推荐装置，有益效果可以参见第一方面的描述此处不再赘述。该装置具有实现上述第一方面的方法实例中行为的功能。所述功能可以通过硬件实现，也可以通过硬件执行相应的软件实现。所述硬件或软件包括一个或多个与上述功能相对应的模块。在一个可能的设计中，所述装置的结构中包括接收模块、资源选择模块、发送模块、组件负载模型、系统负载分配模型以及模型训练模块，这些模块或模型可以执行上述第一方面方法示例中的相应功能，具体参见方法示例中的详细描述，此处不做赘述。
第三方面，本申请实施例还提供了一种计算设备，所述计算设备包括处理器和存储器，还可以包括通信接口。可选的，还包括显示屏。所述处理器执行所述存储器中的程序指令执行上述第一方面或第一方面任一可能的实现方式提供的方法。所述存储器与所述处理器耦合，其保存确定交通流量的过程中必要的程序指令和数据。所述通信接口，用于与其他设备进行通信，如接收云系统资源集合推荐请求。所述显示屏用于在所述处理器的触发下向用户显示信息，如显示目标云系统资源集合。
第四方面，本申请提供了一种计算设备集群，该计算设备集群包括至少一个计算设备。每个计算设备包括存储器和处理器。至少一个计算设备的处理器用于访问所述存储器中的代码以执行第一方面或第一方面的任意一种可能的实现方式提供的方法。
第五方面，本申请提供了一种非瞬态的可读存储介质，所述非瞬态的可读存储介质被计算设备执行时，所述计算设备执行前述第一方面或第一方面的任意可能的实现方式中提供的方法。该存储介质中存储了程序。该存储介质包括但不限于易失性存储器，例如随机访问存储器，非易失性存储器，例如快闪存储器、硬盘(hard disk drive，HDD)、固态硬盘(solid state drive，SSD)。
第六方面，本申请提供了一种计算机程序产品，所述计算机程序产品包括计算机指令，在被计算设备执行时，所述计算设备执行前述第一方面或第一方面的任意可能的实现方式中提供的方法。该计算机程序产品可以为一个软件安装包，在需要使用前述第一方面或第一方面的任意可能的实现方式中提供的方法的情况下，可以下载该计算机程序产品并在计算设备上执行该计算机程序产品。
第七方面，本申请还提供一种计算机芯片，芯片与存储器相连，芯片用于读取并执行存储器中存储的软件程序，执行前述第一方面或第一方面的任意可能的实现方式中提供的方法。
附图说明
图1为本申请提供的一种推荐装置的结构示意图；
图2A为本申请提供的一种计算设备集群的结构示意图；
图2B为本申请提供的一种系统架构示意图；
图3为本申请提供的组件负载模型140以及系统负载分配模型150的训练方法示意图；
图4为本申请提供的系统推荐方法的示意图；
图5为本申请提供的一种向用户显示的选择界面的示意图；
图6为本申请提供的另一种向用户显示的选择界面的示意图；
图7为本申请提供的另一种向用户显示的选择界面的示意图；
图8为本申请提供的向用户显示目标云系统资源集合的界面示意图。
具体实施方式
通常，一个云系统资源集合可以用于支持一种云服务(一种云服务也可以称为一种云业务)，一个云系统资源集合所能支持的云服务取决于该云系统资源集合中每个功能组件所运行的应用程序。从云服务角度来说，一种云服务可以部署在不同的云系统资源集合上，每个云系统资源集合的性能影响了所提供的云服务的质量。
任一云系统资源集合的性能与该云系统资源集合中包括的功能组件的种类以及每个功能组件的资源配置有关。云系统资源集合可以包括一个或多个功能组件，每个功能组件用于实现一种功能，多个功能组件配合支持一种云服务。举例来说，对于支持通话服务的一种云系统资源集合，可以包括三种功能组件，分为用于接收语音的功能组件、用于处理语音(如去噪)的功能组件、以及用于输出语音的功能组件。对于提供通话服务的另一种云系统资源集合，可以包括两种功能组件，分为用于接收语音的功能组件、以及用于输出语音的功能组件。这两种云系统资源集合均可以用于提供通话服务，但包括的功能组件不同，所提供的云服务的质量也存在差别。例如，对于前一种支持通话服务的云系统资源集合中由于包括用于处理语音的功能组件，对接收的语音进行了处理，可以使得最终输出的语音较为清晰，这种云系统资源集合能够对处于嘈杂环境的用户提供较好的通话服务；而后一种提供通话服务的云系统资源集合中并不包括用于处理语音的功能组件，省略了语音处理的操作，能够加快语音传输的效率，虽然不能较好的保证输出的语音的清晰度，却可以较好的保证语音传输的及时性，这种云系统资源集合适用于对语音传输速度有较高要求的场景中。
需要说明的是，若两个云系统资源集合中包括的功能组件的种类相同，且这两个云系统资源集合中能够实现相同功能的功能组件的资源配置不同，这两个云系统资源集合也属于不同的云系统资源集合。
任一功能组件的资源配置包括以下的任意一个或多个：该功能组件中包括的虚拟机(virtual machine，VM)的数量、该功能组件中包括的容器的数量、该功能组件中包括的物理机(physical machine，PM)的数量、每个虚拟机的虚拟中央处理器(virtual central processing unit，VCPU)的数量、每个虚拟机的内存容量、每个虚拟机的磁 盘容量、每个物理机的中央处理器(central processing unit，CPU)的数量、每个物理机的内存容量、每个物理机的磁盘容量。为方便说明将功能组件中一个虚拟机或物理机称为该功能组件的一个实例。
有上可知，不同的云系统资源集合所支持的云服务的质量不同；租户在租用云系统资源集合时，可以考虑该云服务的受众用户以及应用场景，选择适宜的云系统资源集合；仅是靠云端提供的几种有限的云系统资源集合，很难满足租户的需求。为此本申请实施例提供了一种云系统资源集合推荐方法，在对该系统推荐方法说明之前，先对本申请实施例适用的系统架构进行说明。
本申请实施例提供一种云系统资源推荐方法，该方法由推荐装置执行。如图1所示，为本申请实施例提供的一种推荐装置的结构示意图，该推荐装置100包括接收模块110、资源选择模块120、发送模块130、该推荐装置100可以包括组件负载模型140以及系统负载分配模型150，本申请实施例并不限定组件负载模型140以及系统负载分配模型150的数量，这系统负载分配模型150的数量与已有的云系统资源集合的数量有关，组件负载模型140的数量与已有的云系统资源集合中所可能包括的功能组件的种类的数量有关。
接收模块110，用于接收用户触发的云系统资源集合推荐请求，云系统资源集合推荐请求用于请求获取能够支持某一类型的云服务的云系统资源集合。
系统负载分配模型150，一种云系统资源集合对应一个系统负载分配模型，一种云系统资源集合对应一个系统负载分配模型用于根据输入的云系统资源集合的负载情况确定该云资源集合中各个功能组件的负载情况。
组件负载模型140，在本申请实施例中一个功能组件对应一个组件负载模型，一个功能组件对应的组件负载模型用于根据输入的该功能组件的负载情况确定相应的该功能组件的性能数据。
资源选择模块120，用于从已有的支持该类型的云服务的云系统资源集合中选择的云系统资源集合。具体的，资源选择模块可以利用组件负载模型140以及系统负载分配模型150，从已有的支持该类型的云服务的云系统资源集合中选择云系统资源集合。
如图1所示，示例性的绘制了资源选择模块120所选择的两种云系统资源集合，分别为云系统资源集合170和云系统资源集合180。其中，云系统资源集合170中包括功能组件171、功能组件172以及功能组件173，功能组件171中包括一个PM以及一个VM，功能组件171中的PM包括两个CPU、内存以及磁盘，功能组件171中的PM的内存以及磁盘的容量分别为128吉字节(Gigabyte，GB)和500GB。功能组件172中包括一个PM，功能组件172中的PM包括一个CPU、内存以及磁盘，功能组件172中VM的内存以及磁盘的容量分别为256GB和1太字节(Terabyte，T)。功能组件173中包括一个VM以及三个容器，功能组件173中的VM包括一个VCPU、内存以及磁盘，功能组件173中VM的内存以及磁盘的容量分别为256GB和1T。
云系统资源集合180中包括功能组件181、功能组件182以及功能组件183，功能组件181中包括一个PM，功能组件181中的PM包括两个CPU、内存以及磁盘，功能组件181中内存以及磁盘的容量分别为256GB和500GB。功能组件182中包括一个VM， 功能组件182中的VM包括两个VCPU、内存以及磁盘，功能组件182中VM的内存以及磁盘的容量分别为256GB和1T。功能组件183中包括两个VM，功能组件183中的每个VM包括一个VCPU、内存以及磁盘，功能组件183中每个VM的内存以及磁盘的容量分别为128GB和500GB。
发送模块130，用于将资源选择模块120选择的目标云系统资源集合发送给用户。
推荐装置100中的组件负载模型140和系统负载分配模型150可以是推荐装置100之外的装置训练完成并配置在推荐装置中的。作为一种可能的实施方式，推荐装置100中可以包括模型训练模块160，模型训练模块160用于对组件负载模型和系统负载分配模型进行训练，并将训练完成的组件负载模型和系统负载分配模型配置在推荐装置100中。
下面对本申请实施例所适用于的系统架构进行说明，如图2A所示，为本申请实施例提供的一种计算机集群，该计算机集群中包括至少一个计算设备10，每个计算设备10间通过通信网络建立通信通路。
每个计算设备10中包括总线101、处理器102、通信接口103以及存储器104。可选的，计算设备10中还可以包括显示屏105。处理器102、存储器104和通信接口103之间通过总线101通信。
其中，处理器102可以由一个或者多个通用处理器构成，例如中央处理器(central processing unit，CPU)，或者CPU和硬件芯片的组合。上述硬件芯片可以是专用集成电路(application-specific integrated circuit，ASIC)、可编程逻辑器件(programmable logic device，PLD)或其组合。上述PLD可以是复杂可编程逻辑器件(complex programmable logic device，CPLD)、现场可编程逻辑门阵列(field-programmable gate array，FPGA)、通用阵列逻辑(generic array logic，GAL)或其任意组合。
存储器104可以包括易失性存储器(volatile memory)，例如随机存取存储器(random access memory，RAM)。存储器104还可以包括非易失性存储器(non-volatile memory，NVM)，例如只读存储器(read-only memory，ROM)，快闪存储器，硬盘(hard disk drive，HDD)或固态硬盘(solid-state drive，SSD)。存储器104还可以包括上述种类的组合。
存储器104中存储存可执行代码，处理器102可以读取存储器104中的该可执行代码实现功能，还可以通过通信接口103与其他计算设备进行通信，处理器102还可以触发显示屏105向用户显示信息，如本申请实施例中如图5～图8所示的界面中包括的信息。
在本申请实施例中，处理器102可以实现推荐装置100的一个或多个模块(如接收模块110、资源选择模块120、发送模块130、组件负载模型140、系统负载分配模型150以及模型训练模块160中的一个或多个模块)的功能，这种情况下，存储器104中存储有推荐装置100中的接收模块110、资源选择模块120、发送模块130、组件负载模型140以及系统负载分配模型150。
在本申请实施例中，多个计算设备10中的处理器102可以协调工作，执行本申请实施例提供的云系统资源集合推荐方法。
如图2B所示，为本申请实施例提供的一种系统架构，该系统架构中包括客户端200和部署有推荐装置的云端设备300，客户端200与云端设备300通过网络连接，该云端设备300位于云环境中，可以是部署在云数据中心中的服务器或者虚拟机，图2B中，仅是以该推荐装置部署在一个云端设备300为例，作为一种可能的实施方式，该推荐装置可以分布式地部署在多个云端设备300上。
如图2B所示，客户端200包括总线201、处理器202、通信接口203、存储器204。可选的包括客户端200还包括显示屏205。处理器202、存储器204和通信接口203之间通过总线201通信。其中，处理器202和存储器204的类型可以参见处理器102以及存储器104的相关说明，此处不再赘述。存储器204中存储存可执行代码，处理器202可以读取存储器204中的该可执行代码实现功能。处理器202还可以触发显示屏205向用户显示信息，如本申请实施例中如图5～图8所示的界面中包括的信息。处理器202还可以通过通信接口203与云端设备进行通信。例如处理器202可以通过显示屏提示用户输入信息，该信息可以为云系统资源集合推荐请求中携带的信息，如云系统资源集合所需支持的云服务的类型以及云系统资源集合的性能需求；通过通信接口203将携带有云系统资源集合所需支持的云服务的类型以及云系统资源集合的性能需求等信息的云系统资源集合推荐请求发送给云端设备300。
如图2B所示，云端设备300包括总线301、处理器302、通信接口303以及存储器304。处理器302、存储器304和通信接口303之间通过总线301通信。其中，处理器302和存储器304的类型可以参见处理器102以及存储器104的相关说明，此处不再赘述。存储器304中存储存可执行代码，处理器302可以读取存储器304中的该可执行代码实现功能，还可以通过通信接口303与客户端200进行通信。在本申请实施例中，参考图2B，可以由多个云端设备300的处理器302配合实现推荐装置100的功能，这种情况下，存储器304中存储有接收模块110、资源选择模块120、发送模块130、组件负载模型140、系统负载分配模型150以及模型训练模块160中的一个或多个。
处理器302通过通信接口303从客户端200接收到云系统资源集合推荐请求后，可以调用存储器304中存储的模块实现本申请实施例提供的云系统资源集合推荐方法，为用户选择云系统资源集合，处理器302可以通过通信接口303向客户端200发送选择的云系统资源集合。
客户端200中处理器202通过通信接口203在接收到云端设备300选择的云系统资源集合后，处理器202可以触发显示屏205向用户显示该选择的云系统资源集合。
无论推荐装置采用哪种系统架构，在本申请实施例中推荐装置中可以配置有两类不同的模型。
一类为组件负载模型140，每个功能组件对应一个组件负载模型140，一个功能组件对应的组件负载模型140是基于已有的各类云系统资源集合中该功能组件的负载情况与该功能组件的性能数据生成的，这样，该功能组件对应的组件负载模型140可以根据输入的该功能组件的负载情况确定该功能组件的性能数据。
其中，功能组件的负载情况可以用功能组件的并发度(也即在某个时刻该功能组件接收的内部请求数量)，也可以用其他用于反映功能组件的负载情况的参数表征。
功能组件的性能数据包括以下的任意一个或多个：该功能组件的CPU(包括VCPU)平均利用率、该功能组件的内存平均利用率、该功能组件中磁盘的平均读写速率(input/output operations per second，IOPS)。在本申请实施例中仅是以功能组件的性能数据包括上述数据为例进行说明，应需理解的是功能组件的性能数据还可以包括其他能够反映功能组件的性能的数据。
该功能组件的CPU(包括VCPU)平均利用率等于该功能组件中所有实例(一个实例为一个虚拟机或一个物理机)的CPU利用率的平均值，一个实例的CPU的利用率反映一段时间内CPU运行时长占总时长的比重。
功能组件的内存平均利用率该功能组件中被占用的内存与该功能组件总的内存的比重。该功能组件中磁盘的平均IOPS等于该功能组件中所有实例的磁盘的IOPS的平均值，一个实例的磁盘的IOPS为磁盘每秒的读写次数。
上述三种功能组件的性能数据仅是举例，本申请实施例并不限定功能组件的性能数据的具体数据类型，凡是可以反映功能组件的性能的数据均适用于本申请实施例。
另一类为系统负载分配模型150，一种云系统资源集合对应一个系统负载分配模型150，一种云系统资源集合对应一个系统负载分配模型150是基于已有的该云系统资源集合的负载情况与该云系统资源集合中各个功能组件的负载情况生成的，这样，该云系统资源集合对应的系统负载分配模型150可以根据云系统资源集合的负载情况确定该云系统资源集合中各个功能组件的负载情况。
其中，云系统资源集合的负载情况可以用云系统资源集合的并发度(也即在某时刻云系统接收的外部请求的数量)，还可以用其他用于反映云系统资源集合的负载情况的参数表征。在本申请实施例中，以云系统资源集合的并发度表征云系统资源集合的负载情况。功能组件的负载情况的表征方式可参见前述说明，此处不再赘述。
应需理解的是，云系统资源集合中的并发度中涉及到的请求为外部请求，所谓外部请求是指由云系统资源集合外的设备或用户发起的请求。例如为用户在需要云服务时所触发的请求。功能组件的并发度涉及的请求为功能组件发起的内部请求，所谓内部请求是指云系统资源集合内部的功能组件发起的请求。举例来说，以云系统资源集合为用户提供的云服务为网页服务，该云系统资源集合可以分为三个功能组件，分为前端组件、后端组件以及数据库组件，用户所见的网页由前端组件提供的，用户点击网页中的图片、链接以及图标等后台的网页的转换操作是由后台组件执行的；而数据库组件中保存了网页的相关信号，如网页中的文字、图片、以及该网页可跳转的网页信息。当用户点击网页中的图片、链接、以及图标时，触发前端组件向后端组件发起应用程序编程接口(application programming interface，API)请求，该API请求是由用户触发的，为外部请求；后端组件收到前端组件的API请求后对数据库组件发起数据查询请求，后端组件将数据库组件反馈的查询结果返回给前端组件，这里数据查询请求为云系统资源集合中功能组件自己触发的API请求，为内部请求。
推荐装置中配置有上述两类不同的模型，资源选择模块120可以基于这两类不同的模型，根据用户对云系统资源集合的需求(如云系统资源集合所需支持的云服务类型以及用户对该云系统资源集合的性能需求)，选取满足该需求的云系统资源集合。 本申请实施例并不限定模型的类型，这两类不同的模型可以神经网络模型，也可以是基于统计的机器学习算法所建立的模型，比如基于多项式回归算法、岭回归算法、弹性网回归算法等算法建立的模型。资源选择模块120选择满足该需求的云系统资源集合的具体细节将在下文中进行说明。
下面先以上述两类模型是神经网络模型为例，对上述两类的模型的训练方式进行说明，之后再对资源选择模块120确定满足该需求的云系统资源集合的具体细节进行说明。
一、神经网络模型的训练方式。
在本申请实施例中并不限定训练这两类神经网络模型的装置，可以是推荐装置中包括模型训练模块160对这两类神经网络模型进行训练。也可以是其他装置对这两类神经网络模型进行训练，并在训练完成后，将训练完成的这两类神经网络模型配置在推荐装置中的，本申请实施例中仅是以模型训练模块160对这两类神经网络模型进行训练为例进行说明。
如图3所示，为组件负载模型140以及系统负载分配模型150的训练方法示意图，该方法包括：
步骤301：模型训练模块160从已有的云系统资源集合中收集数据，收集的数据可以划分为两类，一类为云系统资源集合的结构数据，另一类为云系统资源集合的运行数据。
1)、云系统资源集合的结构数据用于指示该云系统资源集合所支持的云服务、该云系统资源集合中功能组件的组成以及每个功能组件的资源配置，云系统资源集合的结构数据包括以下的任意一个或多个：每个云系统资源集合所支持的云服务、每个云系统资源集合中包括的功能组件、每个云系统资源集合中各个功能组件中的虚拟机的数量、物理机的数量、每个虚拟机/物理机的CPU(包括VCPU)的数量、每个虚拟机/物理机的内存容量、每个虚拟机/物理机的磁盘容量。
2)、云系统资源集合的运行数据为在该云系统资源集合运行过程中收集的数据，该运行数据可以表征该云系统资源集合整体的性能，以及具体到该云系统资源集合内部，各个功能组件的性能，云系统资源集合的运行数据包括以下的任意一个或多个：云系统资源集合的API调用记录，该云系统资源集合中各个功能组件中各实例的CPU利用率、内存利用率、以及磁盘IOPS。
本申请实施例并不限定云系统资源集合的性能数据包括的数据类型，云系统资源集合的性能数据包括的数据类型与需要训练的组件负载模型140有关，当组件负载模型140除了确定该功能组件的CPU(包括VCPU)平均利用率、该功能组件的内存平均利用率、该功能组件中磁盘的平均读写速率以外还可以确定其他性能数据时，模型训练模块160收集的云系统资源集合的性能数据也需要包括相同类型的性能数据，或能够计算出该其他性能数据的数据。本申请实施例仅是以的云系统资源集合的性能数据包括各个功能组件中各实例的CPU利用率、内存利用率、以及磁盘IOPS为例进行说明。
API调用记录中记录了云系统资源集合处理的每个API请求(包括该云系统资源集合执行的任一外部请求以及该云系统资源集合内部触发的任一内部请求)执行信息， 如API请求的标识(identity，ID)、API请求的触发时间、API请求的结束时间、API请求的返回结果等。
步骤302：模型训练模块160整理收集到的数据，分别生成组件负载模型140的训练集以及系统负载分配模型150的训练集。
由于每个组件负载模型140与一个功能组件对应，每个系统负载分配模型150与一个云系统资源集合对应；也就是说，当存在多个功能组件时，模型训练模块160需要生成每个功能组件对应的组件负载模型140的训练集；当存在多个云系统资源集合时，模型训练模块160需要生成每个云系统资源集合对应的组件负载模型140的训练集。
以功能组件A对应的组件负载模型140为例，对组件负载模型140的训练集中包括的数据进行说明，功能组件A对应的组件负载模型140的训练集包括功能组件A的并发度、功能组件A中实例数量、功能组件A中单个实例的CPU(包括VCPU)的数量、功能组件A中单个实例的内存容量、功能组件A中单个实例的磁盘容量；功能组件A中CPU(包括VCPU)的平均利用率、功能组件A中内存的平均利用率、功能组件A中磁盘的平均读写速率。
需要说明的是，本申请实施例以功能组件A的每一个实例的CPU数量、内存容量以及磁盘容量相同为例，如果该功能组件A中存在多种不同的实例，每种实例的CPU数量、内存容量或磁盘容量不同，则组件负载模型140的训练集中需要包括每种实例的CPU数量、内存容量或磁盘容量。
模型训练模块160为了生成组件负载模型140的训练集，需要对收集到的数据进行进一步整理，例如，模型训练模块160从API调用记录中获取时刻1云系统资源集合中功能组件接收的内部请求数量确定时刻1的功能组件A的并发度；模型训练模块160分析云系统资源集合的结构数据确定功能组件A中实例数量、功能组件A中单个实例的CPU(包括VCPU)的数量、功能组件A中单个实例的内存容量、以及功能组件A中单个实例的磁盘容量；模型训练模块160确定时刻1该云系统资源集合中各个功能组件中各实例的CPU利用率、内存利用率、以及磁盘IOPS，通过求解平均值的方式确定时刻1对应的功能组件A中CPU(包括VCPU)的平均利用率、功能组件A中内存的平均利用率、功能组件A中磁盘的平均读写速率。
应需理解的是，组件负载模型140的训练集中可以按照时刻分为多组训练数据，需要保证后续训练组件负载模型140时，输入的数据与预设的输出的数据对应于相同时刻。
作为一种可能的实施方式，模型训练模块160可以对所支持的云服务、功能组件、功能组件中各实例进行编号形成组织树，将收集到的数据整合到该组织树中，这样可以通过相应的编号，查询到对应的数据，能够较快速的生成组件负载模型140的训练集。
以云系统资源集合1对应的系统负载分配模型150为例，对系统负载分配模型150的训练集包括的数据进行说明，云系统资源集合1对应的系统负载分配模型150的训练集包括云系统资源集合1的并发度、云系统资源集合1的并发度下云系统资源集合1中包括的各个功能组件的并发度。
模型训练模块160为了生成系统负载分配模型150的训练集，需要对收集到的数据进行进一步整理，例如，模型训练模块160从API调用记录中获取时刻1云系统资源集合中各个功能组件接收的内部请求数量确定时刻1的各个功能组件A的并发度，获取时刻1云系统资源集合接收的外部请求数量确定时刻1的云系统资源集合的并发度。
应需理解的是，系统负载分配模型150的训练集中可以按照时刻分为多组训练数据，需要保证后续训练系统负载分配模型150时，输入的数据与预设的输出的数据对应于相同时刻。
模型训练模块160在生成组件负载模型140的训练集以及系统负载分配模型150的训练集之后，可以对各个组件负载模型140和系统负载分配模型150进行训练，如下步骤303为组件负载模型140的训练过程，步骤304为系统负载分配模型150的训练过程。
步骤303：模型训练模块160根据组件负载模型140的训练集，对组件负载模型140进行训练。
以功能组件A对应的组件负载模型140为例，模型训练模块160将功能组件A对应的组件负载模型140的训练集中功能组件A的并发度、功能组件A中的实例数量、每个实例的CPU(包括VCPU)的数量、每个实例的内存容量、每个实例的磁盘容量为输入值，功能组件A中CPU(包括VCPU)的平均利用率、功能组件A中内存的平均利用率、功能组件A中磁盘的平均读写速率为输出值，采用监督学习的方式对功能组件A对应的组件负载模型140进行训练，直至损失函数收敛。
步骤304：模型训练模块160根据系统负载分配模型150的训练集，对系统负载分配模型150进行训练。
以云系统资源集合1对应的系统负载分配模型150为例，模型训练模块160将云系统资源集合1对应的系统负载分配模型150的训练集中的云系统资源集合1的并发度作为输入值，将输入的云系统资源集合1的并发度下云系统资源集合1中各个功能组件的并发度作为输出值，采用监督学习的方式对云系统资源集合1对应的系统负载分配模型150进行训练，直至损失函数收敛。
二、推荐装置确定满足该用户需求的云系统资源集合的方式。
如图4所示，为本申请实施例提供的系统推荐方法的示意图，该方法包括：
步骤401：接收模块110接收由用户触发的云系统资源集合推荐请求，该云系统资源集合推荐请求中包括用户对云系统资源集合的需求(简称为用户需求)，该需求中包括云系统资源集合所需要支持的云服务，还可以包括下列信息的任意一种或多种:云系统资源集合需支持的并发度、云系统资源集合需达到的CPU的平均利用率、云系统资源集合需达到的内存平均利用率、云系统资源集合需达到的磁盘的平均读写速率、云系统资源集合的租金范围以及云系统资源集合中功能组件需达到的CPU的平均利用率、内存平均利用率以及磁盘的平均读写速率。
如图5所示，为客户端200中显示屏或计算设备10中的显示屏显示的一个选择界面，在该界面中，提示用户选择云系统资源集合所需支持的云服务。用户可以通过点 击相应的选项进行选择。
进一步地，如图6为客户端200中显示屏或计算设备10中的显示屏向用户显示的另一个选择界面，用户可以对该云系统集合的负载情况(如云系统资源集合所需要支持的并发度)以及性能(云系统资源集合需达到的CPU的平均利用率、云系统资源集合需达到的内存平均利用率、云系统资源集合需达到的磁盘的平均读写速率)进行设置。如图6的选项中还包括默认，当用户对该云系统集合的负载情况以及性能没有需求时，或无法确定该云系统集合需达到的负载情况以及性能，可以选择默认。该“默认”选项对应的云系统资源集合需支持的并发度、云系统资源集合需达到的CPU的平均利用率、云系统资源集合需达到的内存平均利用率或云系统资源集合需达到的磁盘的平均读写速率为预先设置的值，本申请实施例并不限定预先设置的值的确定方式，可以是经验值，也可以已存在的所有的支持该云服务的云系统资源集合对应数据的平均值。
用户也可以只对云系统资源集合所需要支持的并发度、云系统资源集合需达到的CPU的平均利用率、云系统资源集合需达到的内存平均利用率、云系统资源集合需达到的磁盘的平均读写速率中的一个或多个参数进行设置，对于用户未设置的参数可以默认为无具体需求、或将用户未设置的参数设置为预设值，该预设值可以是经验值，也可以是根据用户已设置的参数匹配的值。
作为一种可能的实施方式，用户除了能够对云系统资源集合所需要支持的并发度、云系统资源集合需达到的CPU的平均利用率、云系统资源集合需达到的内存平均利用率、云系统资源集合需达到的磁盘的平均读写速率中的一个或多个参数进行设置，还可以具体到云系统资源集合的内部，对云系统资源集合中的一个或多个功能组件需达到的CPU的平均利用率，一个或多个功能组件需达到的内存平均利用率以及一个或多个功能组件需达到的磁盘的平均读写速率中的一个或多个参数进行设置。而设置的前提是用户可以先确定云系统资源集合中包括的功能组件的类型，例如，用户可以在客户端200中显示屏或计算设备10中的显示屏向用户显示的界面中配置云系统资源集合推荐请求所述请求的支持网页服务的云系统资源集合中包括前端组件、后端组件以及数据库组件。进一步的，用户对客户端200中显示屏或计算设备10中的显示屏向用户显示的界面中配置该云系统资源集合中一个或多个功能组件需达到的CPU的平均利用率，一个或多个功能组件需达到的内存平均利用率以及一个或多个功能组件需达到的磁盘的平均读写速率的一个或多个参数，例如，用户可以在客户端200中显示屏或计算设备10中的显示屏向用户显示的界面中配置前端组件需达到的CPU的平均利用率、需达到的内存平均利用率以及磁盘的平均读写速率。
如图7所示，为客户端200中显示屏或计算设备10中的显示屏向用户显示的另一个选择界面，用户可以选择该云系统资源集合付费模式(月结、天结以及年结)以及具体费用。
用户通过点击相应的选项，可以触发客户端200向推荐装置发送云系统资源集合推荐请求。
步骤402：接收模块110在接收到该系统资源集合推荐请求后，资源选择模块120 可以基于已有的支持该云服务的云系统资源集合中选择满足用户需求的目标云系统资源集合。
资源选择模块120可以利用预先训练好的用于支持该云服务的云系统资源集合对应的系统负载分配模型150以及各个功能组件对应的组件负载模型140，选择满足用户需求的云系统资源集合。
以系统资源集合推荐请求包括云系统资源集合所需要支持的云服务(以支持网页服务为例)，云系统资源集合需支持的并发度(以300并发为例)、云系统资源集合需达到的CPU的平均利用率(以40％为例)对选择满足用户需求的云系统资源集合的方式进行说明。
第一步：资源选择模块120可以先基于已有的用于支持该云服务的云系统资源集合对应的系统负载分配模型150，根据云系统资源集合需支持的并发度确定已有的用于支持该云服务的云系统资源集合中各个功能组件的并发度。
示例性的，资源选择模块120可以根据云系统资源集合推荐请求中携带的云系统资源集合所需要提供的云服务，查询已有的用于支持该云服务的云系统资源集合对应的系统负载分配模型150。
对于已有的任一个用于支持该云服务的云系统资源集合对应的系统负载分配模型150，资源选择模块120可以将云系统资源集合需支持的并发度输入到该云系统资源集合对应的系统负载模型中，可以确定该云系统资源集合中每个功能组件的并发度。
例如，已有的云系统资源集合中存在三种用于支持网页服务的云系统资源集合，分别为云系统资源集合1、云系统资源集合2以及云系统资源集合3；云系统资源集合1中包括功能组件A、功能组件B以及功能组件C；云系统资源集合2中包括功能组件A、以及功能组件B；云系统资源集合3中包括功能组件A、功能组件B、功能组件C以及功能组件D。
其中，云系统资源集合1对应的系统负载分配模型f1(系统并发度)＝(功能组件A并发度，功能组件B并发度，功能组件C并发度)。
云系统资源集合2对应的系统负载分配模型f2(系统并发度)＝(功能组件A并发度，功能组件B并发度)。
云系统资源集合3对应的系统负载分配模型f3(系统并发度)＝(功能组件A并发度，功能组件B并发度，功能组件C并发度，功能组件D并发度)。
资源选择模块120将云系统资源集合需支持的并发度(300)输入到系统负载分配模型f1(系统并发度)中可以获取功能组件A并发度f1(300)-A、功能组件B并发度f1(300)-B、以及功能组件C并发度f1(300)-C。
资源选择模块120将云系统资源集合需支持的并发度(300)输入到系统负载分配模型f2(300)中可以获取功能组件A并发度f2(300)-A、功能组件B并发度f2(300)-B。
资源选择模块120将云系统资源集合需支持的并发度(300)输入到系统负载分配模型f3(系统并发度)中可以获取功能组件A并发度f3(300)-A、功能组件B并发度f3(300)-B、功能组件C并发度f3(300)-C以及功能组件D并发度f4(300)-D。
第二步、资源选择模块120基于组件负载模型140，根据已有的用于支持该云服务的云系统资源集合中各个功能组件的并发度确定已有的用于支持该云服务的云系统 资源集合中各个功能组件CPU的平均利用率。
对于已有的任一个用于提供该云服务的云系统资源集合，资源选择模块120可以将第一步中计算的该云系统资源集合中每个功能组件的并发度输入至对应的每个功能组件对应的组件负载模型140中，确定每个功能组件CPU的平均利用率。
仍以云系统资源集合推荐请求中携带的云系统资源集合所需要提供的云服务为网页应用为例，已有的云系统资源集合中存在三种用于网页应用的云系统资源集合，共有四种不同的功能组件，分别为功能组件A、功能组件B、功能组件C以及功能组件D。
功能组件A对应的组件负载模型Fa(功能组件A的并发度，功能组件A的单实例CPU用量，功能组件A的单实例内存容量，功能组件A的单实例磁盘容量，功能组件A的实例数)＝(功能组件A的平均CPU利用率，功能组件A的平均内存利用率，功能组件A的平均磁盘读写速度)。
功能组件B对应的组件负载模型Fb(功能组件B的并发度，功能组件B的单实例CPU用量，功能组件B的单实例内存容量，功能组件B的单实例磁盘容量，功能组件B的实例数)＝(功能组件B的平均CPU利用率，功能组件B的平均内存利用率，功能组件B的平均磁盘读写速度)。
功能组件C对应的组件负载模型Fc(功能组件C的并发度，功能组件C的单实例CPU用量，功能组件C的单实例内存容量，功能组件C的单实例磁盘容量，功能组件C的实例数)＝(功能组件C的平均CPU利用率，功能组件C的平均内存利用率，功能组件C的平均磁盘读写速度)。
功能组件D对应的组件负载模型Fd(功能组件D的并发度，功能组件D的单实例CPU用量，功能组件D的单实例内存容量，功能组件D的单实例磁盘容量，功能组件D的实例数)＝(功能组件D的平均CPU利用率，功能组件D的平均内存利用率，功能组件D的平均磁盘读写速度)。
对于云系统资源集合1，资源选择模块120可以将第一步中确定的功能组件A并发度f1(300)-A输入至功能组件A对应的组件负载模型Fa，将功能组件B并发度f1(300)-B输入至功能组件B对应的组件负载模型Fb，将功能组件C并发度f1(300)-C输入至功能组件C对应的组件负载模型Fc。根据功能组件A对应的组件负载模型Fa输出值可以确定功能组件A的平均CPU利用率，根据功能组件B对应的组件负载模型Fb输出值可以确定功能组件B的平均CPU利用率，根据功能组件C对应的组件负载模型Fc输出值可以确定功能组件C的平均CPU利用率。
对于云系统资源集合2，资源选择模块120可以将第一步中确定的功能组件A并发度f2(300)-A输入至功能组件A对应的组件负载模型Fa，将功能组件B并发度f2(300)-B输入至功能组件B对应的组件负载模型Fb。根据功能组件A对应的组件负载模型Fa输出值可以确定功能组件A的平均CPU利用率，根据功能组件B对应的组件负载模型Fb输出值可以确定功能组件B的平均CPU利用率。
对于云系统资源集合3，资源选择模块120可以将第一步中确定的功能组件A并发度f3(300)-A输入至功能组件A对应的组件负载模型Fa，将功能组件B并发度f3(300)-B输入至功能组件B对应的组件负载模型Fb，将功能组件C并发度f3(300)-C输入至功能组件C对应的组件负载模型Fc，将功能组件D并发度f4(300)-D输入至功 能组件D对应的组件负载模型Fd。根据功能组件A对应的组件负载模型Fa输出值可以确定功能组件A的平均CPU利用率，根据功能组件B对应的组件负载模型Fb输出值可以确定功能组件B的平均CPU利用率，根据功能组件C对应的组件负载模型Fc输出值可以确定功能组件C的平均CPU利用率，根据功能组件D对应的组件负载模型Fd输出值可以确定功能组件D的平均CPU利用率。
第三步、资源选择模块120选取云系统资源集合中各个功能组件的CPU的平均利用率均大于云系统资源集合需支持的CPU的平均利用率的云系统资源集合，该云系统资源集合即为目标资源集合。
资源选择模块120也可以选取云系统资源集合中各个功能组件的CPU的平均利用率的均值大于云系统资源集合需支持的CPU的平均利用率的云系统资源集合，该云系统资源集合即为目标资源集合。
若云系统资源集合推荐请求中还携带有租金范围，资源选择模块120还可以从满足云系统资源集合CPU的平均利用率的云系统资源集合中进一步筛选出满足租金范围的目标云系统资源集合。当然，资源选择模块120也可以在执行第一步之前，将不满足租金范围的云系统资源集合剔除，只针对满足租金范围的云系统资源集合执行第一步。
步骤403：发送模块130将选取的目标云系统资源集合反馈给用户。
例如，发送模块130可以将选取的目标云系统资源集合反馈给客户端200，客户端200接收到发送模块130反馈的云系统资源集合，可以将发送模块130反馈的目标云系统资源集合推送给用户，以便用户进行查看。
又例如，发送模块130可以将选取的目标云系统资源集合通过计算设备10中的显示屏推送给用户，以便用户进行查看。
如图8所示，为客户端200中显示屏或计算设备10中的显示屏向用户显示的，客户端200可以将发送模块130反馈的云系统资源集合进行罗列，呈现给用户。客户端200可以采用价格升序或降序的方式对发送模块130反馈的目标云系统资源集合进行排列。
在本申请所提供的几个实施例中，应该理解到，所揭露的系统、装置和方法，可以通过其它的方式实现。例如，以上所描述的装置实施例仅仅是示意性的，例如，所述单元的划分，仅仅为一种逻辑功能划分，实际实现时可以有另外的划分方式，例如多个单元或组件可以结合或者可以集成到另一个系统，或一些特征可以忽略，或不执行。另一点，所显示或讨论的相互之间的耦合或直接耦合或通信连接可以是通过一些接口，装置或单元的间接耦合或通信连接，可以是电性，机械或其它的形式。
所述作为分离部件说明的单元可以是或者也可以不是物理上分开的，作为单元显示的部件可以是或者也可以不是物理单元，即可以位于一个地方，或者也可以分布到多个网络单元上。可以根据实际的需要选择其中的部分或者全部单元来实现本实施例方案的目的。
以上所述，仅为本发明的具体实施方式。熟悉本技术领域的技术人员根据本发明提供的具体实施方式，可想到变化或替换，都应涵盖在本发明的保护范围之内。

Claims (15)

  1. 一种云系统资源集合推荐方法，其特征在于，该方法包括：
    接收由用户触发的云系统资源集合推荐请求，所述云系统资源集合推荐请求用于请求获取能够支持目标云服务、且满足性能需求的云系统资源集合，其中，所述目标云服务由至少一个功能组件配合提供，每个功能组件实现的功能不同；
    根据所述云系统资源集合推荐请求，从已有的支持所述目标云服务的云系统资源集合中选择满足所述性能需求的目标云系统资源集合，所述目标云系统资源集合包括所述至少一个功能组件；
    反馈所述目标云系统资源集合。
  2. 如权利要求1所述的方法，其特征在于，所述性能需求包括所述目标云系统资源集合需支持的并发度，以及下列的一种或多种：
    所述目标云系统资源集合需达到的中央处理器CPU的平均利用率，所述目标云系统资源集合需达到的内存平均利用率，所述目标云系统资源集合需达到的磁盘的平均读写速率，以及所述目标云系统资源集合中任一功能组件需达到的CPU的平均利用率，所述目标云系统资源集合中任一功能组件需达到的内存平均利用率，以及所述目标云系统资源集合中任一功能组件需达到的磁盘的平均读写速率。
  3. 如权利要求1或2所述的方法，其特征在于，所述云系统资源集合推荐请求还包括所述目标云系统资源集合需满足的租金范围。
  4. 如权利要求1或2所述的方法，其特征在于，所述从已有的支持所述目标云服务的云系统资源集合中选择满足所述性能需求的目标云系统资源集合，包括：
    基于已有的支持所述目标云服务的云系统资源集合对应的系统负载分配模型以及所述至少一个功能组件对应的组件负载模型，从已有的支持所述目标云服务的云系统资源集合中选择满足所述性能需求的所述目标云系统资源集合；
    其中，每个云系统资源集合对应的系统负载分配模型用于根据所述云系统资源集合的负载情况确定所述云系统集合中每个功能组件的负载情况，每个功能组件对应的组件负载模型用于根据所述功能组件的负载情况确定所述功能组件的性能数据。
  5. 如权利要求4所述的方法，其特征在于，所述基于已有的支持所述目标云服务的云系统资源集合对应的系统负载分配模型以及所述至少一个功能组件对应的组件负载模型，从已有的支持所述目标云服务的云系统资源集合中选择满足所述性能需求的所述目标云系统资源集合之前，还包括：
    收集已有的支持所述目标云服务的云系统资源集合的结构数据以及运行数据，其中，任一支持所述目标云服务的云系统资源集合的结构数据指示所述云系统资源集合中功能组件的组成以及每个功能组件的资源配置；任一支持所述目标云服务的云系统资源集合的运行数据用于指示所述云系统资源集合在运行过程中的性能以及所述云系统资源集合中各个功能组件的性能；
    根据所述结构数据以及所述运行数据生成已有的支持所述目标云服务的云系统资源集合对应的系统负载分配模型以及所述至少一个功能组件对应的组件负载模型。
  6. 如权利要求5所述的方法，其特征在于，所述根据所述结构数据以及所述运行数据生成已有的支持所述目标云服务的云系统资源集合对应的系统负载分配模型以及 所述至少一个功能组件对应的组件负载模型，包括：
    将所述结构数据以及所述运行数据整合为多个组件负载模型的训练集以及多个系统负载分配模型的训练集；
    基于所述组件负载模型的训练集，生成所述至少一个功能组件对应的组件负载模型；
    基于所述系统负载分配模型的训练集，生成已有的支持所述目标云服务的云系统资源集合对应的系统负载分配模型。
  7. 一种推荐装置，其特征在于，该推荐装置包括：
    接收模块，用于接收由用户触发的云系统资源集合推荐请求，所述云系统资源集合推荐请求用于请求获取能够支持目标云服务、且满足性能需求的云系统资源集合，其中，所述目标云服务由至少一个功能组件配合提供，每个功能组件实现的功能不同；
    资源选择模块，用于根据所述云系统资源集合推荐请求，从已有的支持所述目标云服务的云系统资源集合中选择满足所述性能需求的目标云系统资源集合，所述目标云系统资源集合包括所述至少一个功能组件；
    发送模块，用于反馈所述目标云系统资源集合。
  8. 如权利要求7所述的装置，其特征在于，所述性能需求包括所述目标云系统资源集合需支持的并发度，以及下列的一种或多种：
    所述目标云系统资源集合需达到的中央处理器CPU的平均利用率，所述目标云系统资源集合需达到的内存平均利用率，所述目标云系统资源集合需达到的磁盘的平均读写速率，以及所述目标云系统资源集合中任一功能组件需达到的CPU的平均利用率，所述目标云系统资源集合中任一功能组件需达到的内存平均利用率，以及所述目标云系统资源集合中任一功能组件需达到的磁盘的平均读写速率。
  9. 如权利要求7或8所述的装置，其特征在于，所述云系统资源集合推荐请求还包括所述目标云系统资源集合需满足的租金范围。
  10. 如权利要求7或8所述的装置，其特征在于，所述推荐装置还包括已有的支持所述目标云服务的云系统资源集合对应的系统负载分配模型以及所述至少一个功能组件对应的组件负载模型；
    其中，每个云系统资源集合对应的系统负载分配模型用于根据所述云系统资源集合的负载情况确定所述云系统集合中每个功能组件的负载情况；
    每个功能组件对应的组件负载模型用于根据所述功能组件的负载情况确定所述功能组件的性能数据；
    所述资源选择模块在从已有的支持所述目标云服务的云系统资源集合中选择满足所述性能需求的目标云系统资源集合时，具体用于：
    基于已有的支持所述目标云服务的云系统资源集合对应的系统负载分配模型以及所述至少一个功能组件对应的组件负载模型，从已有的支持所述目标云服务的云系统资源集合中选择满足所述性能需求的所述目标云系统资源集合。
  11. 如权利要求10所述的装置，其特征在于，所述推荐装置还包括模型训练模块，所述模型训练模块，用于：
    收集已有的支持所述目标云服务的云系统资源集合的结构数据以及运行数据，其 中，任一支持所述目标云服务的云系统资源集合的结构数据指示所述云系统资源集合中功能组件的组成以及每个功能组件的资源配置；任一支持所述目标云服务的云系统资源集合的运行数据用于指示所述云系统资源集合在运行过程中所述云系统资源集合的性能以及所述云系统资源集合中各个功能组件的性能；
    根据所述结构数据以及所述运行数据生成已有的支持所述目标云服务的云系统资源集合对应的系统负载分配模型以及所述至少一个功能组件对应的组件负载模型。
  12. 如权利要求11所述的装置，其特征在于，所述模型训练模型在根据所述结构数据以及所述运行数据生成已有的支持所述目标云服务的云系统资源集合对应的系统负载分配模型以及所述至少一个功能组件对应的组件负载模型时，具体用于：
    将所述结构数据以及所述运行数据整合为多个组件负载模型的训练集以及所述系统负载分配模型的训练集；
    基于所述组件负载模型的训练集，生成所述至少一个功能组件对应的组件负载模型；
    基于系统负载分配模型的训练集，生成所述已有的支持所述目标云服务的云系统资源集合对应的系统负载分配模型。
  13. 一种计算设备集群，其特征在于，所述计算设备集群中包括至少一个计算设备，每个计算设备包括处理器和存储器；所述至少一个计算设备中的存储器，用于存储计算机程序指令；
    所述至少一个计算设备中的处理器调用所述存储器中存储的计算机程序指令执行如权利要求1至6中任一项所述的方法。
  14. 一种非瞬态的可读存储介质，其特征在于，所述非瞬态的可读存储介质被计算设备执行时，所述计算设备执行如权利要求1至6任一所述的方法。
  15. 一种包含指令的计算机程序产品，其特征在于，当所述计算机程序产品在计算设备上运行时，使得所述计算机执行如权利要求1至6任一所述的方法。
PCT/CN2020/117271 2019-09-24 2020-09-24 一种云系统资源集合的推荐方法、装置及计算设备集群 WO2021057825A1 (zh)

Priority Applications (2)

Application Number Priority Date Filing Date Title
EP20870185.4A EP4027240A4 (en) 2019-09-24 2020-09-24 CLOUD SYSTEM RESOURCE SET RECOMMENDATION METHOD AND APPARATUS AND COMPUTER DEVICE CLUSTER
US17/702,224 US20220214927A1 (en) 2019-09-24 2022-03-23 Cloud System Resource Set Recommendation Method and Apparatus, and Computing Device Cluster

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN201910906870.8A CN110837417A (zh) 2019-09-24 2019-09-24 一种云系统资源集合的推荐方法、装置及计算设备集群
CN201910906870.8 2019-09-24

Related Child Applications (1)

Application Number Title Priority Date Filing Date
US17/702,224 Continuation US20220214927A1 (en) 2019-09-24 2022-03-23 Cloud System Resource Set Recommendation Method and Apparatus, and Computing Device Cluster

Publications (1)

Publication Number Publication Date
WO2021057825A1 true WO2021057825A1 (zh) 2021-04-01

Family

ID=69574566

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2020/117271 WO2021057825A1 (zh) 2019-09-24 2020-09-24 一种云系统资源集合的推荐方法、装置及计算设备集群

Country Status (4)

Country Link
US (1) US20220214927A1 (zh)
EP (1) EP4027240A4 (zh)
CN (1) CN110837417A (zh)
WO (1) WO2021057825A1 (zh)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US11811681B1 (en) 2022-07-12 2023-11-07 T-Mobile Usa, Inc. Generating and deploying software architectures using telecommunication resources

Families Citing this family (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110837417A (zh) * 2019-09-24 2020-02-25 华为技术有限公司 一种云系统资源集合的推荐方法、装置及计算设备集群
CN111698326B (zh) * 2020-06-12 2023-01-31 北京百度网讯科技有限公司 用于确定云服务资源的成本归属的方法和装置
CN114896296B (zh) * 2022-02-24 2023-08-11 浙江华通云数据科技有限公司 云服务资源配置方法、装置、电子设备及计算机可读介质

Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20130042003A1 (en) * 2011-08-08 2013-02-14 International Business Machines Corporation Smart cloud workload balancer
CN106453146A (zh) * 2016-11-17 2017-02-22 华胜信泰信息产业发展有限公司 私有云计算资源的分配方法及系统
US20170329978A1 (en) * 2012-09-28 2017-11-16 Intel Corporation Device and methods for management and access of distributed data sources
CN107943423A (zh) * 2017-12-07 2018-04-20 郑州云海信息技术有限公司 云系统中存储资源的管理方法和计算机可读存储介质
CN108241716A (zh) * 2016-12-27 2018-07-03 北京金山云网络技术有限公司 一种资源导入方法及装置
CN110837417A (zh) * 2019-09-24 2020-02-25 华为技术有限公司 一种云系统资源集合的推荐方法、装置及计算设备集群

Family Cites Families (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20140200947A1 (en) * 2013-01-15 2014-07-17 Xerox Corporation Methods and systems for regulating service layer agreements for multiple cloud service requests
US9552232B2 (en) * 2013-03-12 2017-01-24 Xerox Corporation System and process to recommend cloud service cloud configuration based on service similarity
US9871857B2 (en) * 2015-04-29 2018-01-16 Microsoft Technology Licensing, Llc Optimal allocation of dynamic cloud computing platform resources
CN109412829B (zh) * 2018-08-30 2020-11-17 华为技术有限公司 一种资源配置的预测方法及设备

Patent Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20130042003A1 (en) * 2011-08-08 2013-02-14 International Business Machines Corporation Smart cloud workload balancer
US20170329978A1 (en) * 2012-09-28 2017-11-16 Intel Corporation Device and methods for management and access of distributed data sources
CN106453146A (zh) * 2016-11-17 2017-02-22 华胜信泰信息产业发展有限公司 私有云计算资源的分配方法及系统
CN108241716A (zh) * 2016-12-27 2018-07-03 北京金山云网络技术有限公司 一种资源导入方法及装置
CN107943423A (zh) * 2017-12-07 2018-04-20 郑州云海信息技术有限公司 云系统中存储资源的管理方法和计算机可读存储介质
CN110837417A (zh) * 2019-09-24 2020-02-25 华为技术有限公司 一种云系统资源集合的推荐方法、装置及计算设备集群

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
See also references of EP4027240A4 *

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US11811681B1 (en) 2022-07-12 2023-11-07 T-Mobile Usa, Inc. Generating and deploying software architectures using telecommunication resources

Also Published As

Publication number Publication date
EP4027240A1 (en) 2022-07-13
US20220214927A1 (en) 2022-07-07
EP4027240A4 (en) 2022-10-05
CN110837417A (zh) 2020-02-25

Similar Documents

Publication Publication Date Title
WO2021057825A1 (zh) 一种云系统资源集合的推荐方法、装置及计算设备集群
JP6961844B2 (ja) ストレージボリューム作成方法および装置、サーバ、並びに記憶媒体
US11487771B2 (en) Per-node custom code engine for distributed query processing
US9183119B2 (en) Automatically allocating clients for software program testing
CN105893139B (zh) 在云存储环境中用于向租户提供存储服务的方法和装置
US9781020B2 (en) Deploying applications in a networked computing environment
US8789050B2 (en) Systems and methods for transparently optimizing workloads
CN111143286B (zh) 一种云平台日志管理方法及系统
CN102880503A (zh) 数据分析系统及数据分析方法
US9459897B2 (en) System and method for providing data analysis service in cloud environment
WO2019200984A1 (zh) 分布式应用的生命周期管理方法、管理器、设备和介质
CN108683720A (zh) 一种容器集群服务配置方法及装置
Lai et al. Towards a framework for large-scale multimedia data storage and processing on Hadoop platform
CN109614227B (zh) 任务资源调配方法、装置、电子设备及计算机可读介质
CN103019853A (zh) 一种作业任务的调度方法和装置
US11720515B2 (en) Article, device, and techniques for serverless stack for streaming message processing
CN108573029B (zh) 一种获取网络访问关系数据的方法、装置及存储介质
US11321090B2 (en) Serializing and/or deserializing programs with serializable state
KR101765725B1 (ko) 대용량 방송용 빅데이터 분산 병렬처리를 위한 동적 디바이스 연결 시스템 및 방법
CN111177237B (zh) 一种数据处理系统、方法及装置
CN110321364B (zh) 信用卡管理系统的交易数据查询方法、装置及终端
CN107704362A (zh) 一种基于Ambari监控大数据组件的方法及装置
US20230137673A1 (en) Systems and methods for dynamically scaling remote resources
US9432476B1 (en) Proxy data storage system monitoring aggregator for a geographically-distributed environment
US11627193B2 (en) Method and system for tracking application activity data from remote devices and generating a corrective action data structure for the remote devices

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 20870185

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

ENP Entry into the national phase

Ref document number: 2020870185

Country of ref document: EP

Effective date: 20220405