WO2020042721A1 - 一种资源配置的预测方法及设备 - Google Patents

一种资源配置的预测方法及设备 Download PDF

Info

Publication number
WO2020042721A1
WO2020042721A1 PCT/CN2019/090892 CN2019090892W WO2020042721A1 WO 2020042721 A1 WO2020042721 A1 WO 2020042721A1 CN 2019090892 W CN2019090892 W CN 2019090892W WO 2020042721 A1 WO2020042721 A1 WO 2020042721A1
Authority
WO
WIPO (PCT)
Prior art keywords
data
resource
resource allocation
prediction model
cloud service
Prior art date
Application number
PCT/CN2019/090892
Other languages
English (en)
French (fr)
Inventor
王新猴
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 EP19855405.7A priority Critical patent/EP3840295A4/en
Publication of WO2020042721A1 publication Critical patent/WO2020042721A1/zh
Priority to US17/186,754 priority patent/US20210182106A1/en

Links

Images

Classifications

    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L41/00—Arrangements for maintenance, administration or management of data switching networks, e.g. of packet switching networks
    • H04L41/16—Arrangements for maintenance, administration or management of data switching networks, e.g. of packet switching networks using machine learning or artificial intelligence
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L41/00—Arrangements for maintenance, administration or management of data switching networks, e.g. of packet switching networks
    • H04L41/14—Network analysis or design
    • H04L41/147—Network analysis or design for predicting network behaviour
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F13/00—Interconnection of, or transfer of information or other signals between, memories, input/output devices or central processing units
    • G06F13/38—Information transfer, e.g. on bus
    • G06F13/40—Bus structure
    • G06F13/4063—Device-to-bus coupling
    • G06F13/409—Mechanical coupling
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
    • G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
    • 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
    • G06F9/44—Arrangements for executing specific programs
    • G06F9/455—Emulation; Interpretation; Software simulation, e.g. virtualisation or emulation of application or operating system execution engines
    • G06F9/45533—Hypervisors; Virtual machine monitors
    • G06F9/45558—Hypervisor-specific management and integration aspects
    • 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
    • G06F9/5011—Allocation of resources, e.g. of the central processing unit [CPU] to service a request the resources being hardware resources other than CPUs, Servers and Terminals
    • G06F9/5016—Allocation of resources, e.g. of the central processing unit [CPU] to service a request the resources being hardware resources other than CPUs, Servers and Terminals the resource being the memory
    • 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
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N20/00—Machine learning
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/045—Combinations of networks
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L41/00—Arrangements for maintenance, administration or management of data switching networks, e.g. of packet switching networks
    • H04L41/08—Configuration management of networks or network elements
    • H04L41/0803—Configuration setting
    • H04L41/0823—Configuration setting characterised by the purposes of a change of settings, e.g. optimising configuration for enhancing reliability
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L41/00—Arrangements for maintenance, administration or management of data switching networks, e.g. of packet switching networks
    • H04L41/08—Configuration management of networks or network elements
    • H04L41/0893—Assignment of logical groups to network elements
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L41/00—Arrangements for maintenance, administration or management of data switching networks, e.g. of packet switching networks
    • H04L41/14—Network analysis or design
    • H04L41/145—Network analysis or design involving simulating, designing, planning or modelling of a network
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L41/00—Arrangements for maintenance, administration or management of data switching networks, e.g. of packet switching networks
    • H04L41/50—Network service management, e.g. ensuring proper service fulfilment according to agreements
    • H04L41/5041—Network service management, e.g. ensuring proper service fulfilment according to agreements characterised by the time relationship between creation and deployment of a service
    • 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
    • G06F9/44—Arrangements for executing specific programs
    • G06F9/455—Emulation; Interpretation; Software simulation, e.g. virtualisation or emulation of application or operating system execution engines
    • G06F9/45533—Hypervisors; Virtual machine monitors
    • G06F9/45558—Hypervisor-specific management and integration aspects
    • G06F2009/45595—Network integration; Enabling network access in virtual machine instances
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2209/00—Indexing scheme relating to G06F9/00
    • G06F2209/50—Indexing scheme relating to G06F9/50
    • G06F2209/5019—Workload prediction
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/048—Activation functions
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods

Definitions

  • the present application relates to the field of cloud computing, and in particular, to a method and a device for predicting resource allocation.
  • the cloud service system needs to allocate appropriate resources to each service, such as a virtual machine (VM).
  • VM virtual machine
  • the cloud service system provider sets the resource configuration for the cloud service. It is mainly divided into two parts: First, before the cloud service system is set up for the enterprise, the supplier must set a reasonable resource configuration for each service and each component in advance, so that It can operate normally and meet the needs of enterprises. Second, during the operation of the cloud service system, it is necessary to continuously adjust the resource configuration of service components according to the change in load scale.
  • the resource allocation of cloud services using the above traditional methods has the following problems: long iteration cycles, unable to quickly develop resource allocation for new cloud services; for large cloud service systems, the test environment is difficult to build; each new cloud is added Services, it is necessary to test the entire cloud service system, wasting a lot of time and human resources; there is a difference between the test environment and the production environment, and the feasible resource allocation in the test environment is not necessarily feasible in the production environment; for cloud service systems due to changes in scale The physical resources are not scalable enough to make good predictions.
  • the above problems will cause the cloud service system to be unable to accurately allocate and plan resources during the deployment and maintenance process, resulting in uncontrollable cloud service system costs.
  • the application provides a method and a device for predicting resource allocation, which are used for resource allocation prediction of a cloud service system, which improves the efficiency of resource allocation and reduces the cost of the cloud service system.
  • a first aspect of the present application provides a method for predicting resource allocation, including:
  • the raw data includes the operating data of the cloud service system in a production or production-like environment
  • the input data of the resource allocation prediction model is generated according to the resource demand data
  • prediction is performed to obtain the resource allocation data of the cloud service system.
  • the operation data in the production or production-like environment can show the true status of the cloud service system at work.
  • the original data can specifically include various operations, And monitoring log files or database files. Since the goal is to build a resource allocation prediction model through deep learning, then when performing deep learning, you need to extract the training data used in the deep learning framework from the original data, and perform the training based on the training data. Deep learning can build a resource allocation prediction model. If the user needs to set up resource demand data before using the cloud service system, obtain the resource demand data entered by the user, and generate resources based on the resource demand data in accordance with the principles applicable to the resource allocation prediction model. Configure the input data of the prediction model.
  • the output data obtained after the resource allocation prediction model processes the input data is actually the resource allocation of the cloud service system.
  • Setting data Use the raw data of the cloud service system to build a resource allocation prediction model, and then use the resource allocation prediction model to predict the resource allocation data corresponding to the resource demand data. Because the original data is the operating data of the cloud service system in a production or production-like environment, Then, compared to the traditional way to set the resource configuration of the cloud service, the construction of a resource configuration prediction model does not need to build a test environment; each time the resource is adjusted, the entire cloud service system is tested under the test environment, and it is only necessary to predict the resource configuration. The model predicts the resource allocation data. Therefore, the efficiency of resource allocation is improved, and the cost of the cloud service system is reduced.
  • obtaining training data based on the original data and performing deep learning construction based on the training data to obtain a resource allocation prediction model includes:
  • Training data is extracted according to the original data.
  • the training data includes data index items and label index items.
  • the data index items include resource request items and resource usage items, and the label index items include resource supply items.
  • Deep learning training is performed according to the formatted data, and a resource allocation prediction model is constructed.
  • the training data includes data index items and label index items.
  • the data index items include resource request items and resource usage items.
  • the label index items include resource supply items.
  • each piece of training data includes data index items and label indicators. Item, due to the need to predict the resource configuration of the cloud service system, convert the training data into formatted data suitable for deep learning model training, use the formatted data for deep learning training, and build a deep neural network model to deep neural
  • the network model is the resource allocation prediction model.
  • the training data includes data index items and label index items.
  • the data index items include resource request items and resource utilization items.
  • the label index items include resource supply items.
  • the resource allocation prediction model obtained based on the training data is learned in the production environment. The relationship between the resource request item and the resource supply item of the cloud service system, and the relationship between the resource request item and the resource supply item of the cloud service system is the resource usage item.
  • generating input data of a resource allocation prediction model according to resource demand data includes:
  • input data of the resource allocation prediction model is determined, and the input data includes the value of the resource request item and the resource usage item of the resource allocation prediction model.
  • the input data of the resource allocation prediction model needs to be obtained based on these simple resource demand data, and the input data of the resource allocation prediction model is the resource request data corresponding to the resource request item and the resource usage rate corresponding to the resource usage item.
  • Data according to the resource request data and the resource usage rate data, determine the input data of the resource allocation prediction model.
  • the input data includes the value of the resource request item and the resource usage item of the resource allocation prediction model.
  • performing prediction based on the input data and the resource allocation prediction model to obtain the resource allocation data of the cloud service system includes:
  • the resource configuration data of the cloud service system is obtained.
  • the value of the resource supply item can be obtained.
  • the value of the resource supply item is actually the value of the cloud service system.
  • the resource demand data can include system scale and resource requirements, etc., and can be responded to according to the resource demand data.
  • the resource request item and the resource usage item are generated locally, and the resource configuration prediction model, the resource request item, and the resource usage item are used to predict the resource configuration of the cloud service system, thereby obtaining the resource configuration data of the cloud service system.
  • the method further includes:
  • the object storage service stores the original data, training data, and model parameters of the resource allocation prediction model.
  • Object Storage Service is a storage service provided by the cloud service system, which can be used to store various types of data, especially after obtaining the original data of the cloud service system, storing the original data to OBS; according to the original data After the training data is obtained, the training data is stored in OBS; the training data is pre-processed to obtain formatted data suitable for deep learning model training, and the formatted data is stored in OBS; and, after the resource allocation prediction model is constructed, Store model parameters to OBS.
  • OBS Object Storage Service
  • the method before performing prediction based on the input data and the resource allocation prediction model, the method further includes:
  • model parameters of the resource allocation prediction model are stored in OBS, before making predictions based on the input data and the resource allocation prediction model, you also need to obtain the model parameters of the resource allocation prediction model through OBS, and load the resource allocation prediction model according to the model parameters. .
  • a second aspect of the present application provides a resource allocation prediction device, including:
  • An acquisition module which is used to obtain the raw data of the cloud service system, and the raw data includes the operation data of the cloud service system in a production or production-like environment;
  • a model training module configured to obtain training data based on the original data, and perform deep learning construction based on the training data to obtain a resource allocation prediction model
  • a processing module configured to generate input data of a resource configuration prediction model according to the resource demand data when the resource demand data input by the user is obtained;
  • the processing module is further configured to perform prediction according to the input data and the resource allocation prediction model to obtain the resource allocation data of the cloud service system.
  • the acquisition module collects the operation data of the cloud service system in a production or production-like environment through a data collector.
  • the operation data in the production or production-like environment can show the true status of the cloud service system at work.
  • the original data can include various types of data. Operations, and monitoring log files or database files. Since the goal of the model training module is to build a resource allocation prediction model through deep learning, then when performing deep learning, the model training module needs to extract the original data for the deep learning framework.
  • the training data in the training data can be used to build a resource allocation prediction model by performing deep learning. If the user needs to use the cloud service system, first set the resource demand data and obtain the resource demand data entered by the user.
  • the resource configuration forecasting model Configure the forecasting model principle, and generate input data for the resource configuration forecasting model based on the resource demand data. Since the input data is generated based on the resource demand data, the resource configuration forecasting model outputs the output data after processing the input data. It is, in fact, the output data resource configuration data cloud service system. Use the raw data of the cloud service system to build a resource allocation prediction model, and then use the resource allocation prediction model to predict the resource allocation data corresponding to the resource demand data.
  • the original data is the operating data of the cloud service system in a production or production-like environment. Then, compared to the traditional way to set the resource configuration of the cloud service, the construction of a resource configuration prediction model does not need to build a test environment; each time the resource is adjusted, the entire cloud service system is tested under the test environment, and it is only necessary to predict the resource configuration. The model predicts the resource allocation data. Therefore, the efficiency of resource allocation is improved, and the cost of the cloud service system is reduced.
  • the model training module is used to extract training data based on the original data.
  • the training data includes data index items and label index items.
  • the data index items include resource request items and resource usage items.
  • the label index items include resource supply items.
  • the model training module is also used to preprocess the training data to obtain formatted data suitable for deep learning model training, and the deep learning model is a resource allocation prediction model;
  • the model training module is also used for deep learning training according to the formatted data to construct a resource allocation prediction model.
  • the training data includes data index items and label index items.
  • the data index items include resource request items and resource usage items.
  • the label index items include resource supply items.
  • each piece of training data includes data index items.
  • label indicators because the resource configuration of the cloud service system needs to be predicted, the model training module converts the training data into formatted data suitable for deep learning model training, uses the formatted data for deep learning training, and builds depth Neural network model, deep neural network model is the resource allocation prediction model.
  • the training data includes data index items and label index items.
  • the data index items include resource request items and resource utilization items.
  • the label index items include resource supply items.
  • the resource allocation prediction model obtained based on the training data is learned in the production environment. The relationship between the resource request item and the resource supply item of the cloud service system, and the relationship between the resource request item and the resource supply item of the cloud service system is the resource usage item.
  • the processing module is further used for processing according to the resource demand data to obtain resource request data and resource usage data;
  • the processing module is further configured to determine input data of the resource allocation prediction model according to the resource request data and the resource usage rate data, and the input data includes values of the resource request item and the resource usage item of the resource allocation prediction model.
  • the processing module needs to obtain the input data of the resource allocation prediction model based on these simple resource demand data, and the input data of the resource allocation prediction model is the resource request data corresponding to the resource request item and the resource corresponding to the resource usage item.
  • the usage rate data is used to determine the input data of the resource allocation prediction model according to the resource request data and the resource usage rate data.
  • the input data includes the value of the resource request item and the resource usage item of the resource allocation prediction model.
  • the processing module is further configured to obtain values of a resource request item and a resource usage item of the resource allocation prediction model according to the input data;
  • the processing module is further configured to substitute the value of the resource request item and the value of the resource usage item into the resource allocation prediction model, and calculate the value of the resource supply item;
  • the processing module is further configured to obtain the resource configuration data of the cloud service system according to the value of the resource supply item.
  • the processing module can obtain the value of the resource supply item by using the constructed resource allocation prediction model.
  • the value of the resource supply item is actually a cloud service
  • the resource demand data can include the system size and resource requirements.
  • the processing module can follow the resource requirements. The data generates a resource request item and a resource usage item accordingly, and uses the resource configuration prediction model, the resource request item, and the resource usage item to predict the resource configuration of the cloud service system, thereby obtaining the resource configuration data of the cloud service system.
  • the prediction device for resource allocation further includes: a storage module;
  • the storage module is configured to store the original data, the training data, and the model parameters of the resource allocation prediction model through the object storage service.
  • OBS is a storage service provided by the cloud service system, which can be used to store various types of data, especially after the acquisition module obtains the original data of the cloud service system, the storage module stores the original data to OBS; the model training module is trained based on the original data After the data, the storage module stores the training data to OBS; the model training module preprocesses the training data to obtain formatted data suitable for deep learning model training, and then the storage module stores the formatted data to OBS; and, the model training module After the resource allocation prediction model is constructed, the storage module stores the model parameters to OBS.
  • the processing module is further configured to obtain model parameters of the resource allocation prediction model through an object storage service of the storage module;
  • the processing module is further configured to obtain a resource allocation prediction model according to the model parameter loading.
  • the processing module also needs to obtain the model parameters of the resource allocation prediction model through the OBS of the storage module and load them according to the model parameters Get the resource allocation prediction model.
  • a third aspect of the present application provides a server, including:
  • the server interface circuit is used to obtain the original data of the cloud service system, and the original data includes the operation data of the cloud service system in the production or production-like environment;
  • a processor configured to obtain training data according to the original data, and perform deep learning construction according to the training data to obtain a resource allocation prediction model
  • the processor is further configured to generate input data of a resource allocation prediction model according to the resource demand data when the resource demand data input by the user is obtained;
  • the processor is further configured to perform prediction according to the input data and the resource configuration prediction model to obtain the resource configuration data of the cloud service system.
  • the server interface circuit collects the operation data of the cloud service system in a production or production-like environment through a data collector.
  • the operation data in the production or production-like environment can show the true status of the cloud service system at work.
  • the original data can specifically include Class operations, and monitoring log files or database files. Since the goal of the processor is to build a resource allocation prediction model through deep learning, then when performing deep learning, the processor needs to extract the original data for use in the deep learning framework. According to the training data, deep learning based on the training data can be used to build a resource allocation prediction model.
  • the processor will apply the resource configuration
  • the principle of the prediction model is to generate the input data of the resource allocation prediction model based on the resource demand data. Because the input data is generated based on the resource demand data, the output data obtained after the resource allocation prediction model processes the input data is output.
  • Data resource configuration data is actually a cloud service system. Use the raw data of the cloud service system to build a resource allocation prediction model, and then use the resource allocation prediction model to predict the resource allocation data corresponding to the resource demand data.
  • the original data is the operating data of the cloud service system in a production or production-like environment. Then, compared to the traditional way to set the resource configuration of the cloud service, the construction of a resource configuration prediction model does not need to build a test environment; each time the resource is adjusted, the entire cloud service system is tested under the test environment, and it is only necessary to predict the resource configuration. The model predicts the resource allocation data. Therefore, the efficiency of resource allocation is improved, and the cost of the cloud service system is reduced.
  • a fourth aspect of the present application provides a chip system, which includes: a prediction system applied to resource allocation, the chip system includes at least one processor, and an interface circuit, and the interface circuit and the at least one processor are interconnected through a line, and the at least one processor executes The operation of the resource allocation prediction device in each of the above-mentioned facilities.
  • a fifth aspect of the present application provides a computer-readable storage medium, which includes: a prediction device applied to a resource configuration, the computer-readable storage medium stores instructions, and when the computer-readable storage medium runs on the computer, causes the computer to execute the second aspect. Operation of prediction equipment for resource allocation in each facility mode.
  • a sixth aspect of the present application provides a computer program product containing instructions that, when run on a computer, causes the computer to perform an operation of a prediction device for resource allocation in each facility mode of the second aspect.
  • FIG. 1 is a schematic flowchart of an embodiment of a resource allocation prediction method provided by this application
  • FIG. 2 is a schematic flowchart of constructing a resource allocation prediction model provided by the present application
  • FIG. 3 is a schematic diagram of a structure of training data provided by the present application.
  • FIG. 4 is a schematic flowchart of another embodiment of a resource allocation prediction method provided by this application.
  • FIG. 5 is a schematic structural diagram of an embodiment of a resource allocation prediction device provided by this application.
  • FIG. 6 is a schematic structural diagram of another embodiment of a resource configuration prediction device provided by this application.
  • FIG. 7 is a schematic structural diagram of a server provided by this application.
  • FIG. 8 is a schematic structural diagram of a chip system provided by the present application.
  • the present application provides a method and a device for predicting resource allocation, which are used for resource allocation prediction of a cloud service system, improve the efficiency of resource allocation, and reduce the cost of the cloud service system.
  • uplink and downlink appearing in this application are used to describe the direction of data / information transmission in some scenarios, for example, the “uplink” direction is the direction in which the data / information is transmitted from the terminal device to the network side, " The “downlink” direction is the direction in which the data / information is transmitted from the network-side device to the terminal device.
  • uplink and downlink are only used to describe the direction, and the specific devices for which the data / information starts and stops are not limited.
  • the naming or numbering of steps in this application does not mean that the steps in the method flow must be executed in the time / logical order indicated by the naming or numbering.
  • the named or numbered process steps can be implemented according to the Technical purposes change the execution order, as long as the same or similar technical effects can be achieved.
  • the division of modules appearing in this application is a logical division. In actual applications, there can be other divisions. For example, multiple modules can be combined or integrated in another system, or some features can be ignored. , Or not executed.
  • the displayed or discussed mutual coupling or direct coupling or communication connection may be through some interfaces.
  • the indirect coupling or communication connection between the modules may be electrical or other similar forms. There are no restrictions in the application.
  • modules or sub-modules described as separate components may or may not be physically separated, may or may not be physical modules, or may be distributed into multiple circuit modules, and some or all of them may be selected according to actual needs. Module to achieve the purpose of the solution of this application.
  • This application is applied to resource allocation of a cloud service system.
  • the cloud service system needs to allocate appropriate resources to each service, such as a VM.
  • the cloud service system provider sets the resource configuration for the cloud service. It is mainly divided into two parts: First, before the cloud service system is set up for the enterprise, the supplier must set a reasonable resource configuration for each service and each component in advance, so that It can operate normally and meet the needs of enterprises.
  • the iteration cycle is long, and resource allocation cannot be made quickly for new cloud services; for large cloud service systems, the test environment is difficult to build; each new cloud service requires the entire cloud service system Testing wastes a lot of time and human resources; there is a difference between the test environment and the production environment, and the feasible resource allocation in the test environment is not necessarily feasible in the production environment; the physical resource scaling of the cloud service system due to scale changes cannot be achieved very Good prediction.
  • the above problems will cause the cloud service system to be unable to accurately allocate and plan resources during the deployment and maintenance process, resulting in uncontrollable cloud service system costs.
  • an embodiment of the present application provides a method for predicting resource allocation, including:
  • the original data of the cloud service system includes the operation data of the cloud service system in a production or production-like environment
  • the cloud service system is a cloud system provided by a cloud service provider for users to use cloud services.
  • the user may be an enterprise.
  • the cloud service system is also applicable to other cloud systems, such as private clouds, hybrids, etc. Cloud, edge cloud, satellite cloud, etc.
  • equipment that can collect data through sensors and other devices can be used as data collectors to collect the operating data of the cloud service system in the production or production-like environment.
  • the operating data in the production or production-like environment can show the trueness of the cloud service system at work.
  • the raw data can include various operations, as well as monitoring log files or database files.
  • the resource allocation prediction model After the resource allocation prediction model has been constructed, if the user needs to use the cloud service system, first set the resource demand data, obtain the resource demand data input by the user, and follow the principles applicable to the resource allocation prediction model according to the resources. Demand data generates input data for the resource allocation prediction model.
  • the output data obtained after the resource allocation prediction model processes the input data is actually the resource configuration data of the cloud service system.
  • a resource allocation prediction model is constructed using the original data of the cloud service system, and then the resource allocation prediction model is used to predict the resource allocation data corresponding to the resource demand data. Since the original data is the cloud service system in production or similar production Running data in the environment, then constructing a resource allocation prediction model compared to the traditional way of setting up the cloud service's resource allocation, there is no need to build a test environment; each time the resource is adjusted, the entire cloud service system is tested in the test environment. Only the resource allocation data need to be predicted through the resource allocation prediction model. Therefore, the efficiency of resource allocation is improved, and the cost of the cloud service system is reduced.
  • training data is obtained based on the original data, and deep learning is constructed based on the training data to obtain a resource allocation prediction model, including:
  • step 101 in the embodiment shown in FIG. 1.
  • the training data includes a data index item and a label index item
  • the data index item includes a resource request item and a resource usage item
  • the label index item includes a resource supply item.
  • the training data data The index item and the label index item are respectively three types: resource supply S (resource supply), resource request D (resource demand), and resource utilization rate U (resource Utilization);
  • the resource supply S includes two types of data center specifications (Datacenter specifications) and service deployment specifications (Service specifications).
  • the data center specifications include the management resource scale (Manage resource pool) and the tenant resource scale (Tenant resource pool).
  • the size of the resources on the side and the scale of the resources on the tenant side include the corresponding computing pool specifications (such as the number of computing nodes, the number of central processing units (CPUs), the number of nodes, the memory, and the disk), and the storage pool specifications (such as storage Number of nodes, total storage space, available space, etc.) and network pool size (such as the number of switches, network ports, etc.).
  • the service deployment specifications include an overall deployment specification (Overall Flavor) and each service 1-n deployment specification (Service Flavor).
  • Each service deployment specification includes a 1-m deployment specification (Flavor) of all components of the service, and the deployment specifications of each component include t-type resources such as CPU, memory, and disks (for example, the deployment of Huawei Cloud Universal Virtual Machine s3.small (Specification is 1 vCPU and 1G memory);
  • Resource request D includes three types: user size (Consumer Specification), service resource usage (Service Resource Allocation), and application programming interface (Application Programming Interface (API) request specification (API) Request Specification) .
  • the user size includes cloud service system account numbers. Number, number of users, etc.
  • the resource usage of each service includes the total resource usage of each service by the tenant and the 1-n resource demand usage of each service. For example, for computing services, such as the Huawei Elastic Cloud Server (ECS), its resource requirements include the number of virtual machines, storage volumes, and networks requested by all tenants.
  • the API request volume includes the number of requests for all p-type APIs in each service, and the types and number of APIs included in each service are different.
  • the resource usage rate U is used to measure the service resource consumption of each service (Service utilization) of the cloud service system when the resource request of the tenant is D and the resource supply of the cloud service system is S.
  • the matching degree between supply S and resource request D is evaluated (supply & demand matching), where the resource consumption of each service 1-n includes the usage of r-type resources by all components 1-m of the service, such as CPU usage, Memory usage, network bandwidth usage, etc., and the matching evaluation of each service is evaluated based on resource usage.
  • Huawei Cloud expects the average virtual machine CPU usage to be 30% to 40%, and the peak value to be 50% to 60%;
  • the average memory usage is 60% to 70%, and the peak is 80% to 90%.
  • each training data contains a total of 682,849 index items, and the specific number of each type of index items is shown in Figure 3.
  • Preprocess the training data to obtain formatted data suitable for deep learning model training, and the deep learning model is a resource allocation prediction model;
  • each piece of training data includes data index items and label index items. Because the resource configuration of the cloud service system needs to be predicted, that is, in the training data index items proposed in step 202, the resource supply S is a label item, and resource request D and resource usage rate U are data items. Therefore, it is necessary to use resource request D and resource usage rate U in the training data as data items and resource supply S as a label item to convert the training data into depth.
  • the type of data required for the training of the learning framework such as the TFRecord type data of Google's artificial intelligence learning system Tensorflow, can be used as formatted data.
  • the resource allocation prediction model is trained in detail.
  • the training data includes data index items and label index items.
  • the data index items include resource request items and resource usage items.
  • the label index items include The resource supply item, and thus the resource allocation prediction model obtained according to the training data, learns the relationship between the resource request D and the resource supply S of the cloud service system in the production environment, and the resource request D and the resource supply S of the cloud service system.
  • the relationship is the resource utilization rate U.
  • the resource allocation prediction model can be used to predict the resource allocation of the cloud service system.
  • the specific implementation is described in the following embodiments.
  • an embodiment of the present application provides a method for predicting resource allocation, including:
  • the number of the resource request D and the resource usage rate U are 270640 and 270609 items, respectively.
  • an enterprise that needs to deploy a cloud service system cannot give such a large and specific index item requirement.
  • Users (such as enterprises) will only enter some simple system scale and service requirements data, such as the account size, user size, virtual machine size, cloud service system, and cloud service system that the cloud service system needs to support.
  • Overall resource utilization That is, the input data of the resource allocation prediction model needs to be obtained based on these simple resource demand data, and the input data of the resource allocation prediction model is the resource request data corresponding to the resource request item and the resource usage rate corresponding to the resource usage item. data.
  • the 402. Determine input data of the resource allocation prediction model according to the resource request data and the resource usage rate data.
  • the input data includes the value of the resource request item and the resource usage item of the resource allocation prediction model.
  • the value of all CPU usage in the resource usage item can be obtained as 60%, and the memory usage The value is 70%.
  • the resource usage item is not clear and can be set to the default value.
  • the value of the resource supply item can be obtained.
  • the value of the resource supply item is actually the resource configuration of the service component of the cloud service system, so the resource configuration data of the cloud service system can be obtained according to the value of the resource supply item.
  • the resource allocation prediction model since the resource allocation prediction model has been trained and constructed according to the training data, users who need to build a cloud service system only need to submit simple resource demand data, and the resource demand data may include system scale and resource requirements, etc.
  • the resource demand data may include system scale and resource requirements, etc.
  • the original data, training data, and resource configuration prediction model model parameters can also be stored through OBS. Before performing prediction based on the input data and resource configuration prediction model, you also need to obtain the model of the resource configuration prediction model through OBS. The parameters are loaded according to the model parameters to obtain a resource allocation prediction model.
  • the foregoing embodiment describes the method for predicting resource configuration.
  • the following describes the device for predicting resource configuration to which the method is applied through an embodiment.
  • an embodiment of the present application provides a device for predicting resource configuration, including:
  • the obtaining module 501 is configured to obtain raw data of a cloud service system, where the raw data includes operating data of the cloud service system in a production or production-like environment;
  • a model training module 502 configured to obtain training data according to the original data, and perform deep learning construction according to the training data to obtain a resource allocation prediction model;
  • a processing module 503 configured to generate input data of a resource allocation prediction model according to the resource demand data when the resource demand data input by the user is obtained;
  • the processing module 503 is further configured to perform prediction according to the input data and the resource configuration prediction model to obtain the resource configuration data of the cloud service system.
  • the acquisition module 501 collects the operation data of the cloud service system in a production or production-like environment through a data collector, and the operation data in the production or production-like environment can show the true status of the cloud service system at work.
  • the raw data can specifically include various operations, as well as monitoring log files or database files. Since the goal of the model training module 502 is to build a resource allocation prediction model through deep learning, the model training module 502 needs to learn from the original
  • the training data used in the deep learning framework is extracted from the data, and the resource allocation prediction model can be constructed by performing deep learning based on the training data.
  • the processing module 503 If the user needs to use the cloud service system, first set the resource requirement data to obtain the resources entered by the user Demand data, the processing module 503 generates input data of the resource allocation prediction model based on the resource demand data according to the principles applicable to the resource allocation prediction model. Since the input data is generated based on the resource demand data, the resource allocation prediction model pairs the number of inputs. Output data obtained after processing, the output data is actually a resource configuration data cloud service system. Use the raw data of the cloud service system to build a resource allocation prediction model, and then use the resource allocation prediction model to predict the resource allocation data corresponding to the resource demand data.
  • the original data is the operating data of the cloud service system in a production or production-like environment. Then, compared to the traditional way to set the resource configuration of the cloud service, the construction of a resource configuration prediction model does not need to build a test environment; each time the resource is adjusted, the entire cloud service system is tested under the test environment, and it is only necessary to predict the resource configuration. The model predicts the resource allocation data. Therefore, the efficiency of resource allocation is improved, and the cost of the cloud service system is reduced.
  • the model training module 502 is configured to obtain training data according to the original data extraction.
  • the training data includes data index items and label index items, the data index items include resource request items and resource usage items, and the label index items include resource supply items.
  • the model training module 502 is further used to preprocess the training data to obtain formatted data suitable for deep learning model training, and the deep learning model is a resource allocation prediction model;
  • the model training module 502 is further configured to perform deep learning training according to the formatted data, and construct a resource allocation prediction model.
  • the training data includes a data index item and a label index item
  • the data index item includes a resource request item and a resource usage item
  • the label index item includes a resource supply item.
  • the model training module 502 A piece of training data includes data index items and label index items. Because the resource configuration of the cloud service system needs to be predicted, the model training module 502 converts the training data into formatted data suitable for deep learning model training, and uses the formatted data. Deep learning training is performed to build a deep neural network model, which is a resource allocation prediction model.
  • the training data includes data index items and label index items.
  • the data index items include resource request items and resource utilization items.
  • the label index items include resource supply items.
  • the resource allocation prediction model obtained based on the training data is learned in the production environment. The relationship between the resource request item and the resource supply item of the cloud service system, and the relationship between the resource request item and the resource supply item of the cloud service system is the resource usage item.
  • the processing module 503 is further configured to perform processing according to the resource demand data to obtain resource request data and resource usage data;
  • the processing module 503 is further configured to determine input data of the resource allocation prediction model according to the resource request data and the resource usage rate data, and the input data includes values of the resource request item and the resource usage item of the resource allocation prediction model.
  • the processing module 503 needs to obtain input data of the resource allocation prediction model based on these simple resource demand data, and the input data of the resource allocation prediction model is the resource request data corresponding to the resource request item and the The resource utilization data is used to determine the input data of the resource allocation prediction model according to the resource request data and the resource usage data.
  • the input data includes the value of the resource request item and the resource usage item of the resource allocation prediction model.
  • the processing module 503 is further configured to obtain values of a resource request item and a resource usage item of the resource configuration prediction model according to the input data;
  • the processing module 503 is further configured to substitute the value of the resource request item and the value of the resource usage item into the resource allocation prediction model, and calculate the value of the resource supply item;
  • the processing module 503 is further configured to obtain resource configuration data of the cloud service system according to the value of the resource supply item.
  • the processing module 503 can obtain the value of the resource supply item by using the constructed resource allocation prediction model.
  • the value is actually the resource configuration of the service components of the cloud service system. Therefore, the resource configuration data of the cloud service system can be obtained according to the value of the resource supply item. Since the resource allocation prediction model has been trained and constructed according to the training data, users who need to build a cloud service system only need to submit simple resource demand data.
  • the resource demand data can include system scale and resource requirements
  • the processing module 503 can The demand data generates a resource request item and a resource usage item accordingly, and uses the resource configuration prediction model, the resource request item, and the resource usage item to predict the resource configuration of the cloud service system, thereby obtaining the resource configuration data of the cloud service system.
  • the resource allocation prediction device further includes: a storage module 601;
  • the storage module 601 is configured to store original data, training data, and model parameters of a resource allocation prediction model through an object storage service.
  • OBS is a storage service provided by the cloud service system, and can be used to store various types of data, especially after the acquisition module 501 obtains the original data of the cloud service system, the storage module 601 stores the original data to the OBS; After the model training module 502 obtains training data according to the original data, the storage module 601 stores the training data in OBS; the model training module 502 preprocesses the training data to obtain formatted data suitable for deep learning model training, and the storage module 601 The formatted data is stored in the OBS; and after the model training module 502 constructs a resource allocation prediction model, the storage module 601 stores the model parameters in the OBS.
  • the processing module 503 is further configured to obtain the model parameters of the resource allocation prediction model through the object storage service of the storage module 601;
  • the processing module 503 is further configured to load and obtain a resource allocation prediction model according to the model parameters.
  • the processing module 503 since the model parameters of the resource allocation prediction model are stored in OBS, before performing prediction based on the input data and the resource allocation prediction model, the processing module 503 also needs to obtain the resource allocation prediction model through the OBS of the storage module 601 Based on the model parameters, the processing module 503 loads the resource allocation prediction model according to the model parameters.
  • the resource allocation prediction device was described by means of a modular device structure.
  • the resource allocation prediction device may be a server in a cloud service system during specific implementation, as follows:
  • an embodiment of the present application provides a server 700, including:
  • the server interface circuit 710 is configured to obtain raw data of the cloud service system, and the raw data includes operating data of the cloud service system in a production or production-like environment;
  • a processor 720 configured to obtain training data according to the original data, and perform deep learning construction according to the training data to obtain a resource allocation prediction model
  • the processor 720 is further configured to generate input data of a resource allocation prediction model according to the resource demand data when the resource demand data input by the user is obtained;
  • the processor 720 is further configured to perform prediction according to the input data and the resource configuration prediction model to obtain resource configuration data of the cloud service system.
  • the server interface circuit 710 collects operation data of the cloud service system in a production or production-like environment through a data collector, and the operation data in the production or production-like environment can show the true status of the cloud service system at work.
  • the raw data can specifically include various operations, as well as monitoring log files or database files. Since the goal of the processor 720 is to build a resource allocation prediction model through deep learning, then the processor 720 needs to learn from the raw data when performing deep learning. Extract the training data used in the deep learning framework, and perform deep learning based on the training data to build a resource allocation prediction model.
  • the processor 720 If the user needs to use the cloud service system, first set the resource requirement data to obtain the resource requirements entered by the user Data, the processor 720 generates the input data of the resource allocation prediction model based on the resource demand data according to the principles applicable to the resource allocation prediction model. Since the input data is generated based on the resource demand data, the resource allocation prediction model processes the input data. After obtaining the output data, the output data is actually a resource configuration data cloud service system. Use the raw data of the cloud service system to build a resource allocation prediction model, and then use the resource allocation prediction model to predict the resource allocation data corresponding to the resource demand data.
  • the original data is the operating data of the cloud service system in a production or production-like environment. Then, compared to the traditional way to set the resource configuration of the cloud service, the construction of a resource configuration prediction model does not need to build a test environment; each time the resource is adjusted, the entire cloud service system is tested under the test environment, and it is only necessary to predict the resource configuration. The model predicts the resource allocation data. Therefore, the efficiency of resource allocation is improved, and the cost of the cloud service system is reduced.
  • server 700 may further include a memory 730, which is configured to be coupled to the processor 720 and stores program instructions and data necessary for the server 700.
  • a part of the memory 730 may further include a non-volatile random access memory (NVRAM).
  • NVRAM non-volatile random access memory
  • the memory 730 stores the following elements, executable modules or data structures, or a subset of them, or their extended set:
  • the processor 720 controls operations of the server 700, and the processor 720 may also be referred to as a CPU.
  • the memory 730 may include a read-only memory and a random access memory, and provide instructions and data to the processor 720. A part of the memory 730 may further include a non-volatile random access memory (NVRAM).
  • NVRAM non-volatile random access memory
  • various components of the server 700 are coupled together through a bus system 740.
  • the bus system 740 may include a power bus, a control bus, and a status signal bus in addition to a data bus. However, for the sake of clarity, various buses are marked as the bus system 740 in the figure.
  • FIG. 8 is a schematic structural diagram of a chip system 800 according to an embodiment of the present application.
  • the chip system 800 includes at least one processor 810 and an interface circuit 830, and the interface circuit 830 and the at least one processor 810 are interconnected through a line.
  • the chip system 800 further includes: a memory 850; the memory 850 may include a read-only memory and a random access memory, and provide the processor 810 with operation instructions and data. A part of the memory 850 may further include a non-volatile random access memory (NVRAM).
  • NVRAM non-volatile random access memory
  • the memory 850 stores the following elements, executable modules or data structures, or their subsets, or their extended sets:
  • a corresponding operation is performed by calling an operation instruction stored in the memory 850 (the operation instruction may be stored in an operating system).
  • the processor 810 controls operations of the network element device, and the processor 810 may also be referred to as a CPU.
  • the memory 850 may include a read-only memory and a random access memory, and provide instructions and data to the processor 810. A part of the memory 850 may further include a non-volatile random access memory (NVRAM).
  • NVRAM non-volatile random access memory
  • various components are coupled together through a bus system 820.
  • the bus system 820 may include a power bus, a control bus, and a status signal bus in addition to a data bus. However, for the sake of clarity, various buses are marked as the bus system 820 in the figure.
  • the methods disclosed in the embodiments of the present application may be applied to the processor 810, or implemented by the processor 810.
  • the processor 810 may be an integrated circuit chip with signal processing capabilities. In the implementation process, each step of the above method may be completed by using hardware integrated logic circuits or instructions in the form of software in the processor 810.
  • the processor 810 may be a general-purpose processor, a digital signal processor (DSP), an application specific integrated circuit (ASIC), an off-the-shelf programmable gate array (FPGA) or other programmable logic device, a discrete gate or transistor logic device, or discrete hardware Components.
  • DSP digital signal processor
  • ASIC application specific integrated circuit
  • FPGA off-the-shelf programmable gate array
  • Various methods, steps, and logical block diagrams disclosed in the embodiments of the present application can be implemented or executed.
  • a general-purpose processor may be a microprocessor or the processor may be any conventional processor or the like.
  • the steps of the method disclosed in combination with the embodiments of the present application may be directly implemented by a hardware decoding processor, or may be performed by using a combination of hardware and software modules in the decoding processor.
  • the software module may be located in a mature storage medium such as a random access memory, a flash memory, a read-only memory, a programmable read-only memory, or an electrically erasable programmable memory, a register, and the like.
  • the storage medium is located in the memory 850, and the processor 810 reads the information in the memory 850 and completes the steps of the foregoing method in combination with its hardware.
  • the instructions stored in the memory for execution by the processor may be implemented in the form of a computer program product.
  • the computer program product may be written in the memory in advance, or may be downloaded and installed in the memory in the form of software.
  • the present application also provides a computer-readable storage medium.
  • the computer-readable storage medium has instructions stored therein, which when run on a computer, cause the computer to execute the prediction method of resource configuration described in the above embodiments.
  • the present application also provides a computer program product containing instructions, which when executed on a computer, causes the computer to execute the method for predicting resource configuration described in the above embodiments.
  • the computer program product includes one or more computer instructions.
  • the computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices.
  • the computer instructions may be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium, for example, the computer instructions may be from a website site, computer, server, or data center Transmission by wire (for example, coaxial cable, optical fiber, digital subscriber line (DSL)) or wireless (for example, infrared, wireless, microwave, etc.) to another website site, computer, server, or data center.
  • wire for example, coaxial cable, optical fiber, digital subscriber line (DSL)
  • wireless for example, infrared, wireless, microwave, etc.
  • the computer-readable storage medium may be any available medium that can be stored by a computer or a data storage device such as a server, a data center, and the like that includes one or more available medium integration.
  • the available medium may be a magnetic medium (for example, a floppy disk, a hard disk, a magnetic tape), an optical medium (for example, a DVD), or a semiconductor medium (for example, a solid state disk (Solid State Disk (SSD)), and the like.
  • the size of the sequence numbers of the above processes does not mean the order of execution.
  • the execution order of each process should be determined by its function and internal logic, and should not deal with the embodiments of the present application.
  • the implementation process constitutes any limitation.

Abstract

本申请提供了一种资源配置的预测方法及设备，用于云服务系统的资源配置预测，提高了资源分配的效率，降低了云服务系统的成本。本申请实施例方法包括：获取云服务系统的原始数据，原始数据包括云服务系统在生产或类生产环境中的运行数据，根据原始数据得到训练数据，根据训练数据进行深度学习构建得到资源配置预测模型，当获取到用户输入的资源需求数据时，根据资源需求数据生成资源配置预测模型的输入数据，根据输入数据及资源配置预测模型进行预测，得到云服务系统的资源配置数据。

Description

一种资源配置的预测方法及设备 技术领域
本申请涉及云计算领域，尤其涉及一种资源配置的预测方法及设备。
背景技术
随着云计算技术的不断成熟，越来越多的企业搭建云服务系统来支撑自己的业务和服务客户。为满足不断增加的需求，各式各样的服务应运而生。云服务系统需要给每一个服务分配合适的资源，比如虚拟机(Virtual Machine，VM)。然而由于不同企业所需要云服务系统的规模不同，云服务的种类和资源需求也存在差异。云服务系统供应商给云服务设置资源配置主要分为两个部分：第一，供应商在给企业搭建云服务系统上线之前，必须预先给每一个服务每一个组件设置合理的资源配置，使其能够正常运行并满足企业的需求；第二，云服务系统在运行期间，需要根据负载规模的变化，不断对服务组件的资源配置进行调整。
传统云服务供应商在云服务系统上线之前，采用测试迭代的方法为所有服务组件设置合理的资源配置，具体为：首先给云服务系统搭建测试环境，然后给所有云服务一个保守的或基于已有经验的资源配置，接着对整个云服务系统进行压力测试，最后根据测试结果调整各云服务的资源配置，以此循环。
但是，采用以上传统方式设置云服务的资源配置存在如下问题：迭代周期长，无法快速地为新的云服务制定资源配置；对于大型的云服务系统，测试环境难构建；每增加一个新的云服务，需要对整个云服务系统进行测试，浪费大量时间和人力资源；测试环境与生产环境存在差异，测试环境下可行的资源配置在生产环境下不一定可行；对于云服务系统因为规模变化而进行的物理资源伸缩，无法做到很好的预测。以上问题会造成云服务系统在部署和维护过程中，不能精准的进行资源分配和规划，造成云服务系统成本不可控。
发明内容
本申请提供了一种资源配置的预测方法及设备，用于云服务系统的资源配置预测，提高了资源分配的效率，降低了云服务系统的成本。
本申请第一方面提供一种资源配置的预测方法，包括：
获取云服务系统的原始数据，原始数据包括云服务系统在生产或类生产环境中的运行数据；
根据原始数据得到训练数据，根据训练数据进行深度学习构建得到资源配置预测模型；
当获取到用户输入的资源需求数据时，根据资源需求数据生成资源配置预测模型的输入数据；
根据输入数据及资源配置预测模型进行预测，得到云服务系统的资源配置数据。
通过数据采集器采集云服务系统在生产或类生产环境下的运行数据，生产或类生 产环境下的运行数据能够表现出云服务系统在工作时候的真实状况，原始数据具体可以包括各类操作、以及监控日志文件或者数据库文件等，由于目标是需要通过深度学习构建得到资源配置预测模型，那么进行深度学习时，需要从原始数据中提取出用于深度学习框架中的训练数据，根据训练数据进行深度学习就能构建得到资源配置预测模型，如果用户需要使用云服务系统之前，先设置资源需求数据，获取到用户输入的资源需求数据，按照适用于资源配置预测模型原则，根据资源需求数据生成资源配置预测模型的输入数据，由于输入数据是根据资源需求数据生成的，那么资源配置预测模型对输入数据进行处理之后得到的输出数据，输出数据实际上就是云服务系统的资源配置数据。先使用云服务系统的原始数据构建出资源配置预测模型，再使用资源配置预测模型来预测资源需求数据对应的资源配置数据，由于原始数据是云服务系统在生产或类生产环境中的运行数据，那么构建资源配置预测模型相比于传统方式设置云服务的资源配置，无需构建测试环境；每次资源调整时在测试环境下对整个云服务系统进行测试也无需进行了，只需要通过资源配置预测模型预测得到资源配置数据。因此，提高了资源分配的效率，降低了云服务系统的成本。
结合本申请第一方面，第一实施方式中，根据原始数据得到训练数据，根据训练数据进行深度学习构建得到资源配置预测模型，包括：
根据原始数据提取得到训练数据，训练数据包括数据指标项及标签指标项，数据指标项包括资源请求项及资源使用率项，标签指标项包括资源供给项；
对训练数据进行预处理，得到适用于深度学习模型训练的格式化数据，深度学习模型为资源配置预测模型；
根据格式化数据进行深度学习训练，构建得到资源配置预测模型。
训练数据中包括了数据指标项及标签指标项，数据指标项包括资源请求项及资源使用率项，标签指标项包括资源供给项，在深度学习中，每一条训练数据包含数据指标项和标签指标项，由于需要对云服务系统的资源配置进行预测，将训练数据转换为适用于深度学习模型训练所需的格式化数据，使用格式化数据进行深度学习训练，构建得到深度神经网络模型，深度神经网络模型即资源配置预测模型。训练数据中包括了数据指标项及标签指标项，数据指标项包括资源请求项及资源使用率项，标签指标项包括资源供给项，从而依据训练数据得到的资源配置预测模型是学习了生产环境中资源请求项和云服务系统的资源供给项之间的关系的，资源请求项和云服务系统的资源供给项之间的关系就是资源使用率项。
结合本申请第一方面第一实施方式，第二实施方式中，根据资源需求数据生成资源配置预测模型的输入数据，包括：
根据资源需求数据进行处理，得到资源请求数据及资源使用率数据；
根据资源请求数据及资源使用率数据，确定资源配置预测模型的输入数据，输入数据包括资源配置预测模型的资源请求项及资源使用率项的取值。
云服务系统部署时，需要部署云服务系统的企业无法给出如此庞大的、具体的指标项要求。而用户(例如企业)只会输入一些简单的系统规模和服务要求数据，例如云服务系统需要支撑的账户规模、用户规模、发放和管理虚拟机规模、并发度、部署的服务种类以及云服务系统资源整体使用率等。也就是说，需要根据这些简单的资源 需求数据,得到资源配置预测模型的输入数据，而资源配置预测模型的输入数据就是对应资源请求项的资源请求数据，以及对应资源使用率项的资源使用率数据，据根据资源请求数据及资源使用率数据，确定资源配置预测模型的输入数据，输入数据包括资源配置预测模型的资源请求项及资源使用率项的取值。
结合本申请第一方面第二实施方式，第三实施方式中，根据输入数据及资源配置预测模型进行预测，得到云服务系统的资源配置数据，包括：
根据输入数据得到资源配置预测模型的资源请求项及资源使用率项的取值；
将资源请求项的取值及资源使用率项的取值，代入资源配置预测模型，计算得到资源供给项的取值；
根据资源供给项的取值，得到云服务系统的资源配置数据。
在得到了资源请求项的取值及资源使用率项的取值之后，利用构建的资源配置预测模型，就能得到资源供给项的取值，资源供给项的取值实际上是云服务系统的服务组件的资源配置，因而，根据资源供给项的取值就能够得到云服务系统的资源配置数据。由于资源配置预测模型是已经根据训练数据训练构建好的，那么需要搭建云服务系统的用户只需要提出简单的资源需求数据，资源需求数据可以包括系统规模和资源要求等，能够按照资源需求数据相应地生成资源请求项和资源使用率项，并使用资源配置预测模型、资源请求项和资源使用率项，对云服务系统的资源配置进行预测，从而得到云服务系统的资源配置数据。
结合本申请第一方面、第一实施方式、第二实施方式或第三实施方式，第四实施方式中，方法还包括：
通过对象存储服务存储原始数据、训练数据及资源配置预测模型的模型参数。
对象存储服务(Object Storage Service，OBS)是云服务系统提供的一个存储服务，能够用于存储各类型的数据，尤其是在获取云服务系统的原始数据之后，存储原始数据至OBS；根据原始数据得到训练数据之后，将训练数据存储至OBS；对训练数据进行预处理，得到适用于深度学习模型训练的格式化数据后，将格式化数据存储至OBS；以及，构建得到资源配置预测模型后，将模型参数存储至OBS。
结合本申请第一方面第四实施方式，第五实施方式中，根据输入数据及资源配置预测模型进行预测之前，还包括：
通过对象存储服务获取资源配置预测模型的模型参数；
根据模型参数加载得到资源配置预测模型。
由于资源配置预测模型的模型参数存储在OBS中，那么在根据输入数据及资源配置预测模型进行预测之前，还需要先通过OBS获取资源配置预测模型的模型参数，根据模型参数加载得到资源配置预测模型。
本申请第二方面提供一种资源配置的预测设备，包括：
获取模块，用于获取云服务系统的原始数据，原始数据包括云服务系统在生产或类生产环境中的运行数据；
模型训练模块，用于根据原始数据得到训练数据，根据训练数据进行深度学习构建得到资源配置预测模型；
处理模块，用于当获取到用户输入的资源需求数据时，根据资源需求数据生成资 源配置预测模型的输入数据；
处理模块，还用于根据输入数据及资源配置预测模型进行预测，得到云服务系统的资源配置数据。
获取模块通过数据采集器采集云服务系统在生产或类生产环境下的运行数据，生产或类生产环境下的运行数据能够表现出云服务系统在工作时候的真实状况，原始数据具体可以包括各类操作、以及监控日志文件或者数据库文件等，由于模型训练模块的目标是需要通过深度学习构建得到资源配置预测模型，那么进行深度学习时，模型训练模块需要从原始数据中提取出用于深度学习框架中的训练数据，根据训练数据进行深度学习就能构建得到资源配置预测模型，如果用户需要使用云服务系统之前，先设置资源需求数据，获取到用户输入的资源需求数据，处理模块按照适用于资源配置预测模型原则，根据资源需求数据生成资源配置预测模型的输入数据，由于输入数据是根据资源需求数据生成的，那么资源配置预测模型对输入数据进行处理之后得到的输出数据，输出数据实际上就是云服务系统的资源配置数据。先使用云服务系统的原始数据构建出资源配置预测模型，再使用资源配置预测模型来预测资源需求数据对应的资源配置数据，由于原始数据是云服务系统在生产或类生产环境中的运行数据，那么构建资源配置预测模型相比于传统方式设置云服务的资源配置，无需构建测试环境；每次资源调整时在测试环境下对整个云服务系统进行测试也无需进行了，只需要通过资源配置预测模型预测得到资源配置数据。因此，提高了资源分配的效率，降低了云服务系统的成本。
结合本申请第二方面，第一实施方式中，
模型训练模块，用于根据原始数据提取得到训练数据，训练数据包括数据指标项及标签指标项，数据指标项包括资源请求项及资源使用率项，标签指标项包括资源供给项；
模型训练模块，还用于对训练数据进行预处理，得到适用于深度学习模型训练的格式化数据，深度学习模型为资源配置预测模型；
模型训练模块，还用于根据格式化数据进行深度学习训练，构建得到资源配置预测模型。
训练数据中包括了数据指标项及标签指标项，数据指标项包括资源请求项及资源使用率项，标签指标项包括资源供给项，模型训练模块在深度学习中，每一条训练数据包含数据指标项和标签指标项，由于需要对云服务系统的资源配置进行预测，模型训练模块将训练数据转换为适用于深度学习模型训练所需的格式化数据，使用格式化数据进行深度学习训练，构建得到深度神经网络模型，深度神经网络模型即资源配置预测模型。训练数据中包括了数据指标项及标签指标项，数据指标项包括资源请求项及资源使用率项，标签指标项包括资源供给项，从而依据训练数据得到的资源配置预测模型是学习了生产环境中资源请求项和云服务系统的资源供给项之间的关系的，资源请求项和云服务系统的资源供给项之间的关系就是资源使用率项。
结合本申请第二方面第一实施方式，第二实施方式中，
处理模块，还用于根据资源需求数据进行处理，得到资源请求数据及资源使用率数据；
处理模块，还用于根据资源请求数据及资源使用率数据，确定资源配置预测模型的输入数据，输入数据包括资源配置预测模型的资源请求项及资源使用率项的取值。
云服务系统部署时，需要部署云服务系统的企业无法给出如此庞大的、具体的指标项要求。而用户(例如企业)只会输入一些简单的系统规模和服务要求数据，例如云服务系统需要支撑的账户规模、用户规模、发放和管理虚拟机规模、并发度、部署的服务种类以及云服务系统资源整体使用率等。也就是说，处理模块需要根据这些简单的资源需求数据,得到资源配置预测模型的输入数据，而资源配置预测模型的输入数据就是对应资源请求项的资源请求数据，以及对应资源使用率项的资源使用率数据，据根据资源请求数据及资源使用率数据，确定资源配置预测模型的输入数据，输入数据包括资源配置预测模型的资源请求项及资源使用率项的取值。
结合本申请第二方面第二实施方式，第三实施方式中，
处理模块，还用于根据输入数据得到资源配置预测模型的资源请求项及资源使用率项的取值；
处理模块，还用于将资源请求项的取值及资源使用率项的取值，代入资源配置预测模型，计算得到资源供给项的取值；
处理模块，还用于根据资源供给项的取值，得到云服务系统的资源配置数据。
处理模块在得到了资源请求项的取值及资源使用率项的取值之后，利用构建的资源配置预测模型，就能得到资源供给项的取值，资源供给项的取值实际上是云服务系统的服务组件的资源配置，因而，根据资源供给项的取值就能够得到云服务系统的资源配置数据。由于资源配置预测模型是已经根据训练数据训练构建好的，那么需要搭建云服务系统的用户只需要提出简单的资源需求数据，资源需求数据可以包括系统规模和资源要求等，处理模块能够按照资源需求数据相应地生成资源请求项和资源使用率项，并使用资源配置预测模型、资源请求项和资源使用率项，对云服务系统的资源配置进行预测，从而得到云服务系统的资源配置数据。
结合本申请第二方面、第一实施方式、第二实施方式或第三实施方式，第四实施方式中，资源配置的预测设备还包括：存储模块；
存储模块，用于通过对象存储服务存储原始数据、训练数据及资源配置预测模型的模型参数。
OBS是云服务系统提供的一个存储服务，能够用于存储各类型的数据，尤其是在获取模块获取云服务系统的原始数据之后，存储模块存储原始数据至OBS；模型训练模块根据原始数据得到训练数据之后，存储模块将训练数据存储至OBS；模型训练模块对训练数据进行预处理，得到适用于深度学习模型训练的格式化数据后，存储模块将格式化数据存储至OBS；以及，模型训练模块构建得到资源配置预测模型后，存储模块将模型参数存储至OBS。
结合本申请第二方面第四实施方式，第五实施方式中，
处理模块，还用于通过存储模块的对象存储服务获取资源配置预测模型的模型参数；
处理模块，还用于根据模型参数加载得到资源配置预测模型。
由于资源配置预测模型的模型参数存储在OBS中，那么在根据输入数据及资源配置预测模型进行预测之前，处理模块还需要先通过存储模块的OBS获取资源配置预测模型的模型参数，根据模型参数加载得到资源配置预测模型。
本申请第三方面提供一种服务器，包括：
服务器接口电路，用于获取云服务系统的原始数据，原始数据包括云服务系统在生产或类生产环境中的运行数据；
处理器，用于根据原始数据得到训练数据，根据训练数据进行深度学习构建得到资源配置预测模型；
处理器，还用于当获取到用户输入的资源需求数据时，根据资源需求数据生成资源配置预测模型的输入数据；
处理器，还用于根据输入数据及资源配置预测模型进行预测，得到云服务系统的资源配置数据。
服务器接口电路通过数据采集器采集云服务系统在生产或类生产环境下的运行数据，生产或类生产环境下的运行数据能够表现出云服务系统在工作时候的真实状况，原始数据具体可以包括各类操作、以及监控日志文件或者数据库文件等，由于处理器的目标是需要通过深度学习构建得到资源配置预测模型，那么进行深度学习时，处理器需要从原始数据中提取出用于深度学习框架中的训练数据，根据训练数据进行深度学习就能构建得到资源配置预测模型，如果用户需要使用云服务系统之前，先设置资源需求数据，获取到用户输入的资源需求数据，处理器按照适用于资源配置预测模型原则，根据资源需求数据生成资源配置预测模型的输入数据，由于输入数据是根据资源需求数据生成的，那么资源配置预测模型对输入数据进行处理之后得到的输出数据，输出数据实际上就是云服务系统的资源配置数据。先使用云服务系统的原始数据构建出资源配置预测模型，再使用资源配置预测模型来预测资源需求数据对应的资源配置数据，由于原始数据是云服务系统在生产或类生产环境中的运行数据，那么构建资源配置预测模型相比于传统方式设置云服务的资源配置，无需构建测试环境；每次资源调整时在测试环境下对整个云服务系统进行测试也无需进行了，只需要通过资源配置预测模型预测得到资源配置数据。因此，提高了资源分配的效率，降低了云服务系统的成本。
本申请第四方面提供一种芯片系统，包括：应用于资源配置的预测设备中，芯片系统包括至少一个处理器，和接口电路，接口电路和至少一个处理器通过线路互联，至少一个处理器执行上述第二方面各设施方式中资源配置的预测设备的操作。
本申请第五方面提供一种计算机可读存储介质，包括：应用于资源配置的预测设备中，计算机可读存储介质中存储有指令，当其在计算机上运行时，使得计算机执行上述第二方面各设施方式中资源配置的预测设备的操作。
本申请第六方面提供一种包含指令的计算机程序产品，当其在计算机上运行时，使得所述计算机执行上述第二方面各设施方式中资源配置的预测设备的操作。
附图说明
图1为本申请提供的资源配置的预测方法的一个实施例流程示意图；
图2为本申请提供的构建资源配置预测模型的流程示意图；
图3为本申请提供的训练数据的构成示意图；
图4为本申请提供的资源配置的预测方法的另一个实施例流程示意图；
图5为本申请提供的资源配置的预测设备的一个实施例结构示意图；
图6为本申请提供的资源配置的预测设备的另一个实施例结构示意图；
图7为本申请提供的服务器的结构示意图；
图8为本申请提供的芯片系统的结构示意图。
具体实施方式
本申请提供了一种资源配置的预测方法及设备，用于云服务系统的资源配置预测，提高资源分配的效率，降低了云服务系统的成本。
本申请中出现的术语“上行”和“下行”，在某些场景用于描述数据/信息传输的方向，比如，“上行”方向为该数据/信息从终端设备向网络侧传输的方向，“下行”方向为该数据/信息从网络侧设备向终端设备传输的方向，“上行”和“下行”仅用于描述方向，该数据/信息传输起止的具体设备都不作限定。
本申请中出现的术语“和/或”，可以是一种描述关联对象的关联关系，表示可以存在三种关系，例如，A和/或B，可以表示：单独存在A，同时存在A和B，单独存在B这三种情况。另外，本申请中字符“/”，一般表示前后关联对象是一种“或”的关系。
本申请中可能出现的对各种消息/信息/设备/网元/系统/装置/动作/操作/流程/概念等各类客体进行了赋名，但这些具体的名称并不构成对相关客体的限定，所赋名称可随着场景，语境或者使用习惯等因素而变更，对相关客体的技术含义的理解，应主要从其在技术方案中所体现/执行的功能和技术效果来确定。
本申请的说明书和权利要求书及上述附图中的术语“第一”、“第二”等是用于区别类似的对象，而不必用于描述特定的顺序或先后次序。应该理解这样使用的数据在适当情况下可以互换，以便这里描述的实施例能够以除了在这里图示或描述的内容以外的顺序实施。此外，术语“包括”和“具有”以及他们的任何变形，意图在于覆盖不排他的包含，例如，包含了一系列步骤或模块的过程、方法、系统、产品或设备不必限于清楚地列出的那些步骤或模块，而是可包括没有清楚地列出的或对于这些过程、方法、产品或设备固有的其它步骤或模块。在本申请中出现的对步骤进行的命名或者编号，并不意味着必须按照命名或者编号所指示的时间/逻辑先后顺序执行方法流程中的步骤，已经命名或者编号的流程步骤可以根据要实现的技术目的变更执行次序，只要能达到相同或者相类似的技术效果即可。本申请中所出现的模块的划分，是一种逻辑上的划分，实际应用中实现时可以有另外的划分方式，例如多个模块可以结合成或集成在另一个系统中，或一些特征可以忽略，或不执行，另外，所显示的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，模块之间的间接耦合或通信连接可以是电性或其他类似的形式，本申请中均不作限定。并且，作为分离部件说明的模块或子模块可以是也可以不是物理上的分离，可以是也可以不是物理模块，或者可以分布到多个电路模块中，可以根据实际的需要选择其中的部分或全部模块来 实现本申请方案的目的。
首先简单介绍本申请应用的系统构架或场景。
本申请应用于云服务系统的资源配置。随着云计算技术的不断成熟，越来越多的企业搭建云服务系统来支撑自己的业务和服务客户。为满足不断增加的需求，各式各样的服务应运而生。云服务系统需要给每一个服务分配合适的资源，比如VM。然而由于不同企业所需要云服务系统的规模不同，云服务的种类和资源需求也存在差异。云服务系统供应商给云服务设置资源配置主要分为两个部分：第一，供应商在给企业搭建云服务系统上线之前，必须预先给每一个服务每一个组件设置合理的资源配置，使其能够正常运行并满足企业的需求；第二，云服务系统在运行期间，需要根据负载规模的变化，不断对服务组件的资源配置进行调整。传统云服务供应商在云服务系统上线之前，采用测试迭代的方法为所有服务组件设置合理的资源配置，具体为：首先给云服务系统搭建测试环境，然后给所有云服务一个保守的或基于已有经验的资源配置，接着对整个云服务系统进行压力测试，最后根据测试结果调整各云服务的资源配置，以此循环。
但是以上传统的方式中，迭代周期长，无法快速地为新的云服务制定资源配置；对于大型的云服务系统，测试环境难构建；每增加一个新的云服务，需要对整个云服务系统进行测试，浪费大量时间和人力资源；测试环境与生产环境存在差异，测试环境下可行的资源配置在生产环境下不一定可行；对于云服务系统因为规模变化而进行的物理资源伸缩，无法做到很好的预测。以上问题会造成云服务系统在部署和维护过程中，不能精准的进行资源分配和规划，造成云服务系统成本不可控。
为了解决以上问题，如图1所示，本申请实施例提供了一种资源配置的预测方法，包括：
101、获取云服务系统的原始数据，原始数据包括云服务系统在生产或类生产环境中的运行数据；
本实施例中，云服务系统是由云服务商提供给用户使用云服务的云系统，用户可以是企业，云服务系统除了公有云之外，还适用于其他云系统，例如，私有云、混合云、边缘云、卫星云等。一般可以通过感应器等采集数据的装备作为数据采集器，采集云服务系统在生产或类生产环境下的运行数据，生产或类生产环境下的运行数据能够表现出云服务系统在工作时候的真实状况，原始数据具体可以包括各类操作、以及监控日志文件或者数据库文件等。
102、根据原始数据得到训练数据，根据训练数据进行深度学习构建得到资源配置预测模型；
本实施例中，由于目标是需要通过深度学习构建得到资源配置预测模型，那么进行深度学习时，使用原始数据是无法进行的，需要从原始数据中提取出用于深度学习框架中的训练数据，根据训练数据进行深度学习就能构建得到资源配置预测模型。
103、当获取到用户输入的资源需求数据时，根据资源需求数据生成资源配置预测模型的输入数据；
本实施例中，在资源配置预测模型已经构建了之后，如果用户需要使用云服务系统之前，先设置资源需求数据，获取到用户输入的资源需求数据，按照适用于资源配 置预测模型原则，根据资源需求数据生成资源配置预测模型的输入数据。
104、根据输入数据及资源配置预测模型进行预测，得到云服务系统的资源配置数据。
本实施例中，由于输入数据是根据资源需求数据生成的，那么资源配置预测模型对输入数据进行处理之后得到的输出数据，输出数据实际上就是云服务系统的资源配置数据。
本申请实施例中，先使用云服务系统的原始数据构建出资源配置预测模型，再使用资源配置预测模型来预测资源需求数据对应的资源配置数据，由于原始数据是云服务系统在生产或类生产环境中的运行数据，那么构建资源配置预测模型相比于传统方式设置云服务的资源配置，无需构建测试环境；每次资源调整时在测试环境下对整个云服务系统进行测试也无需进行了，只需要通过资源配置预测模型预测得到资源配置数据。因此，提高了资源分配的效率，降低了云服务系统的成本。
以上图1所示的实施例中，对于资源配置预测模型的构建并未进行具体说明，下面通过具体的实施例进行详细说明。
请参阅图2，基于图1所示的实施例，本申请的一些实施例中，根据原始数据得到训练数据，根据训练数据进行深度学习构建得到资源配置预测模型，包括：
201、获取云服务系统的原始数据；
具体参考图1所示的实施例中的步骤101。
202、根据原始数据提取得到训练数据；
本实施例中，训练数据中包括了数据指标项及标签指标项，数据指标项包括资源请求项及资源使用率项，标签指标项包括资源供给项，具体如图3所示，训练数据的数据指标项和标签指标项分别为，资源供给S(resource Supply)、资源请求D(resource Demand)和资源使用率U(resource Utilization)三种；
其中，资源供给S包括数据中心规格(Datacenter specification)和服务部署规格(Service specification)两种，其中数据中心规格包括管理侧资源规模(Manage resource pool)和租户侧资源规模(Tenant resource pool),管理侧资源规模和租户侧资源规模都包含相应的计算池规格(例如计算节点数、中央处理器(Central Processing Unit，CPU)核数、节点赫兹数、内存、磁盘等)、存储池规格(例如存储节点数、总存储空间、可用空间等)和网络池规模(例如交换机数量、网口数等)。服务部署规格包含总体部署规格(Overall Flavor)和各服务1-n部署规格(Service Flavor)。其中各服务部署规格包括该服务所有组件1-m的部署规格(Flavor)，每个组件的部署规格包括CPU、内存、磁盘等t类资源(例如，华为云通用型虚机s3.small的部署规格为1个vCPU和1G内存)；
资源请求D包括用户规模(Consumer Specification)、各服务资源用量(Service Resource Allocation)和应用程序编程接口(Application Programming Interface，API)请求量(API Request Specification)三种,其中用户规模包含云服务系统账号数量、用户数量等。各服务资源用量包含租户对各服务的总体资源使用量以及分别对各服务1-n资源需求使用量。比如对计算服务来说，例如华为云弹性云服务器服务(Elastic Cloud Server，ECS)，其资源需求量包括所有租户请求的虚拟机数量、存 储卷数量和网络数量等。API请求量包含各服务中所有p类API的请求数，各服务包含的API类型、数量不同。
资源使用率U用来衡量云服务系统在租户资源请求为D，云服务系统资源供给为S的情形下，各服务的服务资源消耗(Service utilization),同时可以根据各服务的资源消耗，对资源供给S和资源请求D的匹配度进行评价(supply & demand matching Evaluation),其中各服务1-n的资源消耗，包括该服务所有组件1-m对r类资源的使用率，例如CPU使用率、内存使用率、网络带宽使用量等,而各服务的匹配度评价根据资源使用率进行评价,例如,华为云期望虚拟机CPU平均使用率在30％～40％，峰值在50％～60％；内存平均使用率60％～70％，峰值80％～90％等。
在本实施例中，假设训练数据中云服务类型数为n＝256，每个服务包含组件类型数为m＝32，每个组件部署规格包含的资源类型数为t＝16，每个服务包含API类型数为p＝1024，每个服务消耗资源类型为r＝32。根据上述配置，每一条训练数据总共包含的指标项682849项，每一类指标项包含具体数量如图3所示。另外，本实施例中每条训练数据时间跨度为5分钟，因此通过数据采集器每天可以从一套生产环境中获取24*12＝288条训练数据。
203、对训练数据进行预处理，得到适用于深度学习模型训练的格式化数据，深度学习模型为资源配置预测模型；
本实施例中，在深度学习中，每一条训练数据包含数据指标项和标签指标项，由于需要对云服务系统的资源配置进行预测，也就是说步骤202提出的训练数据指标项中，资源供给S是标签项，资源请求D和资源使用率U是数据项，因此，需要以训练数据中资源请求D和资源使用率U作为数据项，以资源供给S作为标签项，将训练数据转换为深度学习框架训练所需数据类型，例如谷歌人工智能学习系统Tensorflow的TFRecord类型数据，就可以作为格式化数据。
204、根据格式化数据进行深度学习训练，构建得到资源配置预测模型。
本实施例中，可以采用深度神经网络(Deep Neutral Network，DNN)或卷积神经网络(Convolutional Neutral Network，CNN)等来进行深度学习训练，包括神经网络层数L、每一层神经元数量li(0<i<＝L)、代价函数均分误差(Mean Square Error，MSE)、激活函数线性整流函数(Rectified Linear Unit，ReLu)等，使用格式化数据进行深度学习训练，构建得到深度神经网络模型，深度神经网络模型即资源配置预测模型。
本申请实施例中，详细的介绍了资源配置预测模型是如何训练得到的，训练数据中包括了数据指标项及标签指标项，数据指标项包括资源请求项及资源使用率项，标签指标项包括资源供给项，从而依据训练数据得到的资源配置预测模型是学习了生产环境中资源请求D和云服务系统的资源供给S之间的关系的，资源请求D和云服务系统的资源供给S之间的关系就是资源使用率U。
根据如图2所示的实施例中的描述，在资源配置预测模型构建完成之后，便可以通过资源配置预测模型来进行云服务系统的资源配置的预测，具体的实施通过以下实施例进行说明。
请参阅图4，基于以上图1和图2所示的实施例，本申请实施例提供一种资源配置的预测方法，包括：
401、根据资源需求数据进行处理，得到资源请求数据及资源使用率数据；
本实施例中，依据如图3中所示的指标项，资源请求D和资源使用率U的个数分别为270640和270609项。但是云服务系统部署时，需要部署云服务系统的企业无法给出如此庞大的、具体的指标项要求。而用户(例如企业)只会输入一些简单的系统规模和服务要求数据，例如云服务系统需要支撑的账户规模、用户规模、发放和管理虚拟机规模、并发度、部署的服务种类以及云服务系统资源整体使用率等。也就是说，需要根据这些简单的资源需求数据,得到资源配置预测模型的输入数据，而资源配置预测模型的输入数据就是对应资源请求项的资源请求数据，以及对应资源使用率项的资源使用率数据。
402、根据资源请求数据及资源使用率数据，确定资源配置预测模型的输入数据，输入数据包括资源配置预测模型的资源请求项及资源使用率项的取值；
本实施例中，假设用户输入的资源使用率数据为CPU使用率为60％，内存使用率为70％，则可以得到资源使用率项中所有CPU使用率的取值为60％，内存使用率的取值为70％。而资源使用率项中没有明确的，可以设置为默认值。对于资源请求项中各API请求量，可以根据并发度设置为C档(例如C＝10)，然后根据用户输入的并发度选择相应档位的API请求量，从而得到资源请求项的取值。
403、将资源请求项的取值及资源使用率项的取值，代入资源配置预测模型，计算得到资源供给项的取值；
本实施例中，在得到了资源请求项的取值及资源使用率项的取值之后，利用图2所示的实施例中构建的资源配置预测模型，就能得到资源供给项的取值。
404、根据资源供给项的取值，得到云服务系统的资源配置数据。
本实施例中，资源供给项的取值实际上是云服务系统的服务组件的资源配置，因而根据资源供给项的取值就能够得到云服务系统的资源配置数据。
本申请实施例中，由于资源配置预测模型是已经根据训练数据训练构建好的，那么需要搭建云服务系统的用户只需要提出简单的资源需求数据，资源需求数据可以包括系统规模和资源要求等，能够按照资源需求数据相应地生成资源请求项和资源使用率项，并使用资源配置预测模型、资源请求项和资源使用率项，对云服务系统的资源配置进行预测，从而得到云服务系统的资源配置数据。
需要说明的是，还可以通过OBS存储原始数据、训练数据及资源配置预测模型的模型参数，那么在根据输入数据及资源配置预测模型进行预测之前，还需要先通过OBS获取资源配置预测模型的模型参数，根据模型参数加载得到资源配置预测模型。
以上实施例描述的是资源配置的预测方法，下面通过实施例对应用该方法的资源配置的预测设备进行说明。
请参阅图5，本申请实施例提供一种资源配置的预测设备，包括：
获取模块501，用于获取云服务系统的原始数据，原始数据包括云服务系统在生产或类生产环境中的运行数据；
模型训练模块502，用于根据原始数据得到训练数据，根据训练数据进行深度学习构建得到资源配置预测模型；
处理模块503，用于当获取到用户输入的资源需求数据时，根据资源需求数据生成资源配置预测模型的输入数据；
处理模块503，还用于根据输入数据及资源配置预测模型进行预测，得到云服务系统的资源配置数据。
本申请实施例中，获取模块501通过数据采集器采集云服务系统在生产或类生产环境下的运行数据，生产或类生产环境下的运行数据能够表现出云服务系统在工作时候的真实状况，原始数据具体可以包括各类操作、以及监控日志文件或者数据库文件等，由于模型训练模块502的目标是需要通过深度学习构建得到资源配置预测模型，那么进行深度学习时，模型训练模块502需要从原始数据中提取出用于深度学习框架中的训练数据，根据训练数据进行深度学习就能构建得到资源配置预测模型，如果用户需要使用云服务系统之前，先设置资源需求数据，获取到用户输入的资源需求数据，处理模块503按照适用于资源配置预测模型原则，根据资源需求数据生成资源配置预测模型的输入数据，由于输入数据是根据资源需求数据生成的，那么资源配置预测模型对输入数据进行处理之后得到的输出数据，输出数据实际上就是云服务系统的资源配置数据。先使用云服务系统的原始数据构建出资源配置预测模型，再使用资源配置预测模型来预测资源需求数据对应的资源配置数据，由于原始数据是云服务系统在生产或类生产环境中的运行数据，那么构建资源配置预测模型相比于传统方式设置云服务的资源配置，无需构建测试环境；每次资源调整时在测试环境下对整个云服务系统进行测试也无需进行了，只需要通过资源配置预测模型预测得到资源配置数据。因此，提高了资源分配的效率，降低了云服务系统的成本。
可选的，结合图5所示的实施例，本申请的一些实施例中，
模型训练模块502，用于根据原始数据提取得到训练数据，训练数据包括数据指标项及标签指标项，数据指标项包括资源请求项及资源使用率项，标签指标项包括资源供给项；
模型训练模块502，还用于对训练数据进行预处理，得到适用于深度学习模型训练的格式化数据，深度学习模型为资源配置预测模型；
模型训练模块502，还用于根据格式化数据进行深度学习训练，构建得到资源配置预测模型。
本申请实施例中，训练数据中包括了数据指标项及标签指标项，数据指标项包括资源请求项及资源使用率项，标签指标项包括资源供给项，模型训练模块502在深度学习中，每一条训练数据包含数据指标项和标签指标项，由于需要对云服务系统的资源配置进行预测，模型训练模块502将训练数据转换为适用于深度学习模型训练所需的格式化数据，使用格式化数据进行深度学习训练，构建得到深度神经网络模型，深度神经网络模型即资源配置预测模型。训练数据中包括了数据指标项及标签指标项，数据指标项包括资源请求项及资源使用率项，标签指标项包括资源供给项，从而依据训练数据得到的资源配置预测模型是学习了生产环境中资源请求项和云服务系统的资源供给项之间的关系的，资源请求项和云服务系统的资源供给项之间的关系就是资源 使用率项。
可选的，结合图5所示的实施例，本申请的一些实施例中，
处理模块503，还用于根据资源需求数据进行处理，得到资源请求数据及资源使用率数据；
处理模块503，还用于根据资源请求数据及资源使用率数据，确定资源配置预测模型的输入数据，输入数据包括资源配置预测模型的资源请求项及资源使用率项的取值。
本申请实施例中，云服务系统部署时，需要部署云服务系统的企业无法给出如此庞大的、具体的指标项要求。而用户(例如企业)只会输入一些简单的系统规模和服务要求数据，例如云服务系统需要支撑的账户规模、用户规模、发放和管理虚拟机规模、并发度、部署的服务种类以及云服务系统资源整体使用率等。也就是说，处理模块503需要根据这些简单的资源需求数据,得到资源配置预测模型的输入数据，而资源配置预测模型的输入数据就是对应资源请求项的资源请求数据，以及对应资源使用率项的资源使用率数据，据根据资源请求数据及资源使用率数据，确定资源配置预测模型的输入数据，输入数据包括资源配置预测模型的资源请求项及资源使用率项的取值。
可选的，结合图5所示的实施例，本申请的一些实施例中，
处理模块503，还用于根据输入数据得到资源配置预测模型的资源请求项及资源使用率项的取值；
处理模块503，还用于将资源请求项的取值及资源使用率项的取值，代入资源配置预测模型，计算得到资源供给项的取值；
处理模块503，还用于根据资源供给项的取值，得到云服务系统的资源配置数据。
本申请实施例中，处理模块503在得到了资源请求项的取值及资源使用率项的取值之后，利用构建的资源配置预测模型，就能得到资源供给项的取值，资源供给项的取值实际上是云服务系统的服务组件的资源配置，因而，根据资源供给项的取值就能够得到云服务系统的资源配置数据。由于资源配置预测模型是已经根据训练数据训练构建好的，那么需要搭建云服务系统的用户只需要提出简单的资源需求数据，资源需求数据可以包括系统规模和资源要求等，处理模块503能够按照资源需求数据相应地生成资源请求项和资源使用率项，并使用资源配置预测模型、资源请求项和资源使用率项，对云服务系统的资源配置进行预测，从而得到云服务系统的资源配置数据。
可选的，结合图6所示的实施例，本申请的一些实施例中，资源配置的预测设备还包括：存储模块601；
存储模块601，用于通过对象存储服务存储原始数据、训练数据及资源配置预测模型的模型参数。
本申请实施例中，OBS是云服务系统提供的一个存储服务，能够用于存储各类型的数据，尤其是在获取模块501获取云服务系统的原始数据之后，存储模块601存储原始数据至OBS；模型训练模块502根据原始数据得到训练数据之后，存储模块601将训练数据存储至OBS；模型训练模块502对训练数据进行预处理，得到适用于深度学习模型训练的格式化数据后，存储模块601将格式化数据存储至OBS；以及，模型训练模块502构建得到资源配置预测模型后，存储模块601将模型参数存储至OBS。
可选的，结合图6所示的实施例，本申请的一些实施例中，
处理模块503，还用于通过存储模块601的对象存储服务获取资源配置预测模型的模型参数；
处理模块503，还用于根据模型参数加载得到资源配置预测模型。
本申请实施例中，由于资源配置预测模型的模型参数存储在OBS中，那么在根据输入数据及资源配置预测模型进行预测之前，处理模块503还需要先通过存储模块601的OBS获取资源配置预测模型的模型参数，处理模块503根据模型参数加载得到资源配置预测模型。
以上实施例中通过模块化的装置结构的方式对资源配置的预测设备进行了说明，资源配置的预测设备在具体实施时可以是处于云服务系统中的服务器，具体如下：
请参阅图7，本申请实施例提供一种服务器700，包括：
服务器接口电路710，用于获取云服务系统的原始数据，原始数据包括云服务系统在生产或类生产环境中的运行数据；
处理器720，用于根据原始数据得到训练数据，根据训练数据进行深度学习构建得到资源配置预测模型；
处理器720，还用于当获取到用户输入的资源需求数据时，根据资源需求数据生成资源配置预测模型的输入数据；
处理器720，还用于根据输入数据及资源配置预测模型进行预测，得到云服务系统的资源配置数据。
本申请实施例中，服务器接口电路710通过数据采集器采集云服务系统在生产或类生产环境下的运行数据，生产或类生产环境下的运行数据能够表现出云服务系统在工作时候的真实状况，原始数据具体可以包括各类操作、以及监控日志文件或者数据库文件等，由于处理器720的目标是需要通过深度学习构建得到资源配置预测模型，那么进行深度学习时，处理器720需要从原始数据中提取出用于深度学习框架中的训练数据，根据训练数据进行深度学习就能构建得到资源配置预测模型，如果用户需要使用云服务系统之前，先设置资源需求数据，获取到用户输入的资源需求数据，处理器720按照适用于资源配置预测模型原则，根据资源需求数据生成资源配置预测模型的输入数据，由于输入数据是根据资源需求数据生成的，那么资源配置预测模型对输入数据进行处理之后得到的输出数据，输出数据实际上就是云服务系统的资源配置数据。先使用云服务系统的原始数据构建出资源配置预测模型，再使用资源配置预测模型来预测资源需求数据对应的资源配置数据，由于原始数据是云服务系统在生产或类生产环境中的运行数据，那么构建资源配置预测模型相比于传统方式设置云服务的资源配置，无需构建测试环境；每次资源调整时在测试环境下对整个云服务系统进行测试也无需进行了，只需要通过资源配置预测模型预测得到资源配置数据。因此，提高了资源分配的效率，降低了云服务系统的成本。
需要说明的是，服务器700还可以包括存储器730，存储器730用于与处理器720耦合，其保存服务器700必要的程序指令和数据。
存储器730的一部分还可以包括非易失性随机存取存储器(NVRAM)。存储器730 存储了如下的元素，可执行模块或者数据结构，或者他们的子集，或者他们的扩展集:
通过调用存储器730存储的操作指令(该操作指令可存储在操作系统中)，执行相应的操作。处理器720控制服务器700的操作，处理器720还可以称为CPU。存储器730可以包括只读存储器和随机存取存储器，并向处理器720提供指令和数据。存储器730的一部分还可以包括非易失性随机存取存储器(NVRAM)。具体的应用中服务器700的各个组件通过总线系统740耦合在一起，其中总线系统740除包括数据总线之外，还可以包括电源总线、控制总线和状态信号总线等。但是为了清楚说明起见，在图中将各种总线都标为总线系统740。
图8是本申请实施例提供的芯片系统800的结构示意图。芯片系统800包括至少一个处理器810和接口电路830，接口电路830和至少一个处理器810通过线路互联。
可选的，如图8所示，在本申请的一些实施方式中，芯片系统800还包括：存储器850；存储器850可以包括只读存储器和随机存取存储器，并向处理器810提供操作指令和数据。存储器850的一部分还可以包括非易失性随机存取存储器(NVRAM)。
在一些实施方式中，存储器850存储了如下的元素，可执行模块或者数据结构，或者他们的子集，或者他们的扩展集:
在本申请实施例中，通过调用存储器850存储的操作指令(该操作指令可存储在操作系统中)，执行相应的操作。
处理器810控制网元设备的操作，处理器810还可以称为CPU。存储器850可以包括只读存储器和随机存取存储器，并向处理器810提供指令和数据。存储器850的一部分还可以包括非易失性随机存取存储器(NVRAM)。具体的应用中各个组件通过总线系统820耦合在一起，其中总线系统820除包括数据总线之外，还可以包括电源总线、控制总线和状态信号总线等。但是为了清楚说明起见，在图中将各种总线都标为总线系统820。
上述本申请实施例揭示的方法可以应用于处理器810中，或者由处理器810实现。处理器810可能是一种集成电路芯片，具有信号的处理能力。在实现过程中，上述方法的各步骤可以通过处理器810中的硬件的集成逻辑电路或者软件形式的指令完成。上述的处理器810可以是通用处理器、数字信号处理器(DSP)、专用集成电路(ASIC)、现成可编程门阵列(FPGA)或者其他可编程逻辑器件、分立门或者晶体管逻辑器件、分立硬件组件。可以实现或者执行本申请实施例中的公开的各方法、步骤及逻辑框图。通用处理器可以是微处理器或者该处理器也可以是任何常规的处理器等。结合本申请实施例所公开的方法的步骤可以直接体现为硬件译码处理器执行完成，或者用译码处理器中的硬件及软件模块组合执行完成。软件模块可以位于随机存储器，闪存、只读存储器，可编程只读存储器或者电可擦写可编程存储器、寄存器等本领域成熟的存储介质中。该存储介质位于存储器850，处理器810读取存储器850中的信息，结合其硬件完成上述方法的步骤。
在上述实施例中，存储器存储的供处理器执行的指令可以以计算机程序产品的形式实现。所述计算机程序产品可以是事先写入在存储器中，也可以是以软件形式下载并安装在存储器中。
本申请还提供了一种计算机可读存储介质，计算机可读存储介质中存储有指令， 当其在计算机上运行时，使得计算机执行以上实施例所描述的资源配置的预测方法。
本申请还提供了一种包含指令的计算机程序产品，当其在计算机上运行时，使得计算机执行以上实施例所描述的资源配置的预测方法。
在上述实施例中，可以全部或部分地通过软件、硬件、固件或者其任意组合来实现。当使用软件实现时，可以全部或部分地以计算机程序产品的形式实现。
所述计算机程序产品包括一个或多个计算机指令。在计算机上加载和执行所述计算机程序指令时，全部或部分地产生按照本申请实施例所述的流程或功能。所述计算机可以是通用计算机、专用计算机、计算机网络、或者其他可编程装置。所述计算机指令可以存储在计算机可读存储介质中，或者从一个计算机可读存储介质向另一计算机可读存储介质传输，例如，所述计算机指令可以从一个网站站点、计算机、服务器或数据中心通过有线(例如同轴电缆、光纤、数字用户线(DSL))或无线(例如红外、无线、微波等)方式向另一个网站站点、计算机、服务器或数据中心进行传输。所述计算机可读存储介质可以是计算机能够存储的任何可用介质或者是包含一个或多个可用介质集成的服务器、数据中心等数据存储设备。所述可用介质可以是磁性介质，(例如，软盘、硬盘、磁带)、光介质(例如，DVD)、或者半导体介质(例如固态硬盘Solid State Disk(SSD))等。
应理解，在本申请的各种实施例中，上述各过程的序号的大小并不意味着执行顺序的先后，各过程的执行顺序应以其功能和内在逻辑确定，而不应对本申请实施例的实施过程构成任何限定。
以上所述，以上实施例仅用以说明本申请的技术方案，而非对其限制；尽管参照前述实施例对本申请进行了详细的说明，本领域的普通技术人员应当理解：其依然可以对前述各实施例所记载的技术方案进行修改，或者对其中部分技术特征进行等同替换；而这些修改或者替换，并不使相应技术方案的本质脱离本申请各实施例技术方案的范围。

Claims (16)

  1. 一种资源配置的预测方法，其特征在于，包括：
    获取云服务系统的原始数据，所述原始数据包括所述云服务系统在生产或类生产环境中的运行数据；
    根据所述原始数据得到训练数据，根据所述训练数据进行深度学习构建得到资源配置预测模型；
    当获取到用户输入的资源需求数据时，根据所述资源需求数据生成所述资源配置预测模型的输入数据；
    根据所述输入数据及所述资源配置预测模型进行预测，得到所述云服务系统的资源配置数据。
  2. 根据权利要求1所述的方法，其特征在于，所述根据所述原始数据得到训练数据，根据所述训练数据进行深度学习构建得到资源配置预测模型，包括：
    根据所述原始数据提取得到训练数据，所述训练数据包括数据指标项及标签指标项，所述数据指标项包括资源请求项及资源使用率项，所述标签指标项包括资源供给项；
    对所述训练数据进行预处理，得到适用于深度学习模型训练的格式化数据，所述深度学习模型为资源配置预测模型；
    根据所述格式化数据进行深度学习训练，构建得到资源配置预测模型。
  3. 根据权利要求2所述的方法，其特征在于，所述根据所述资源需求数据生成所述资源配置预测模型的输入数据，包括：
    根据所述资源需求数据进行处理，得到资源请求数据及资源使用率数据；
    根据所述资源请求数据及所述资源使用率数据，确定所述资源配置预测模型的输入数据，所述输入数据包括所述资源配置预测模型的资源请求项及资源使用率项的取值。
  4. 根据权利要求3所述的方法，其特征在于，所述根据所述输入数据及所述资源配置预测模型进行预测，得到所述云服务系统的资源配置数据，包括：
    根据所述输入数据得到所述资源配置预测模型的资源请求项及资源使用率项的取值；
    将所述资源请求项的取值及所述资源使用率项的取值，代入所述资源配置预测模型，计算得到所述资源供给项的取值；
    根据所述资源供给项的取值，得到所述云服务系统的资源配置数据。
  5. 根据权利要求1-4中任一项所述的方法，其特征在于，所述方法还包括：
    通过对象存储服务存储所述原始数据、所述训练数据及所述资源配置预测模型的模型参数。
  6. 根据权利要求5所述的方法，其特征在于，所述根据所述输入数据及所述资源配置预测模型进行预测之前，还包括：
    通过所述对象存储服务获取所述资源配置预测模型的模型参数；
    根据所述模型参数加载得到所述资源配置预测模型。
  7. 一种资源配置的预测设备，其特征在于，包括：
    获取模块，用于获取云服务系统的原始数据，所述原始数据包括所述云服务系统在生产或类生产环境中的运行数据；
    模型训练模块，用于根据所述原始数据得到训练数据，根据所述训练数据进行深度学习构建得到资源配置预测模型；
    处理模块，用于当获取到用户输入的资源需求数据时，根据所述资源需求数据生成所述资源配置预测模型的输入数据；
    所述处理模块，还用于根据所述输入数据及所述资源配置预测模型进行预测，得到所述云服务系统的资源配置数据。
  8. 根据权利要求7所述的设备，其特征在于，
    所述模型训练模块，用于根据所述原始数据提取得到训练数据，所述训练数据包括数据指标项及标签指标项，所述数据指标项包括资源请求项及资源使用率项，所述标签指标项包括资源供给项；
    所述模型训练模块，还用于对所述训练数据进行预处理，得到适用于深度学习模型训练的格式化数据，所述深度学习模型为资源配置预测模型；
    所述模型训练模块，还用于根据所述格式化数据进行深度学习训练，构建得到资源配置预测模型。
  9. 根据权利要求8所述的设备，其特征在于，
    所述处理模块，还用于根据所述资源需求数据进行处理，得到资源请求数据及资源使用率数据；
    所述处理模块，还用于根据所述资源请求数据及所述资源使用率数据，确定所述资源配置预测模型的输入数据，所述输入数据包括所述资源配置预测模型的资源请求项及资源使用率项的取值。
  10. 根据权利要求9所述的设备，其特征在于，
    所述处理模块，还用于根据所述输入数据得到所述资源配置预测模型的资源请求项及资源使用率项的取值；
    所述处理模块，还用于将所述资源请求项的取值及所述资源使用率项的取值，代入所述资源配置预测模型，计算得到所述资源供给项的取值；
    所述处理模块，还用于根据所述资源供给项的取值，得到所述云服务系统的资源配置数据。
  11. 根据权利要求7-10任一项所述的设备，其特征在于，所述资源配置的预测设备还包括：存储模块；
    所述存储模块，用于通过对象存储服务存储所述原始数据、所述训练数据及所述资源配置预测模型的模型参数。
  12. 根据权利要求11所述的设备，其特征在于，
    所述处理模块，还用于通过所述存储模块的所述对象存储服务获取所述资源配置预测模型的模型参数；
    所述处理模块，还用于根据所述模型参数加载得到所述资源配置预测模型。
  13. 一种服务器，其特征在于，包括：
    服务器接口电路，用于获取云服务系统的原始数据，所述原始数据包括所述云服 务系统在生产或类生产环境中的运行数据；
    处理器，用于根据所述原始数据得到训练数据，根据所述训练数据进行深度学习构建得到资源配置预测模型；
    所述处理器，还用于当获取到用户输入的资源需求数据时，根据所述资源需求数据生成所述资源配置预测模型的输入数据；
    所述处理器，还用于根据所述输入数据及所述资源配置预测模型进行预测，得到所述云服务系统的资源配置数据。
  14. 一种芯片系统，其特征在于，包括：应用于资源配置的预测设备中，所述芯片系统包括至少一个处理器，和接口电路，所述接口电路和所述至少一个处理器通过线路互联，所述至少一个处理器执行权利要求7-12中所述资源配置的预测设备的操作。
  15. 一种计算机可读存储介质，其特征在于，包括：应用于资源配置的预测设备中，所述计算机可读存储介质中存储有指令，当其在计算机上运行时，使得所述计算机执行权利要求7-12中所述资源配置的预测设备的操作。
  16. 一种包含指令的计算机程序产品，其特征在于，当其在计算机上运行时，使得所述计算机执行权利要求7-12中所述资源配置的预测设备的操作。
PCT/CN2019/090892 2018-08-30 2019-06-12 一种资源配置的预测方法及设备 WO2020042721A1 (zh)

Priority Applications (2)

Application Number Priority Date Filing Date Title
EP19855405.7A EP3840295A4 (en) 2018-08-30 2019-06-12 RESOURCE CONFIGURATION PREDICTION METHOD AND DEVICE
US17/186,754 US20210182106A1 (en) 2018-08-30 2021-02-26 Resource configuration prediction method and device

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN201811004607.1A CN109412829B (zh) 2018-08-30 2018-08-30 一种资源配置的预测方法及设备
CN201811004607.1 2018-08-30

Related Child Applications (1)

Application Number Title Priority Date Filing Date
US17/186,754 Continuation US20210182106A1 (en) 2018-08-30 2021-02-26 Resource configuration prediction method and device

Publications (1)

Publication Number Publication Date
WO2020042721A1 true WO2020042721A1 (zh) 2020-03-05

Family

ID=65463759

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2019/090892 WO2020042721A1 (zh) 2018-08-30 2019-06-12 一种资源配置的预测方法及设备

Country Status (4)

Country Link
US (1) US20210182106A1 (zh)
EP (1) EP3840295A4 (zh)
CN (1) CN109412829B (zh)
WO (1) WO2020042721A1 (zh)

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112073239A (zh) * 2020-09-04 2020-12-11 天津大学 一种云计算环境分布式应用性能预测方法
WO2022246833A1 (en) * 2021-05-28 2022-12-01 Huawei Cloud Computing Technologies Co., Ltd. System, method, and medium for elastic allocation of resources for deep learning jobs

Families Citing this family (32)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
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
CN110019151B (zh) * 2019-04-11 2024-03-15 深圳市腾讯计算机系统有限公司 数据库性能调整方法、装置、设备、系统及存储介质
WO2020206699A1 (en) * 2019-04-12 2020-10-15 Microsoft Technology Licensing, Llc Predicting virtual machine allocation failures on server node clusters
CN110084511B (zh) * 2019-04-25 2021-07-13 广东工业大学 一种无人机配型方法、装置、设备及可读存储介质
CN111858015B (zh) * 2019-04-25 2024-01-12 中国移动通信集团河北有限公司 配置应用程序的运行资源的方法、装置及网关
CN110297713A (zh) * 2019-05-22 2019-10-01 菜鸟智能物流控股有限公司 一种云主机的配置管理系统和方法
CN110213104B (zh) * 2019-06-05 2022-04-08 北京百度网讯科技有限公司 容量预测方法及装置
CN110198244B (zh) 2019-06-19 2022-02-18 北京百度网讯科技有限公司 面向异构云服务的资源配置方法和装置
CN112350878A (zh) * 2019-08-07 2021-02-09 阿里巴巴集团控股有限公司 一种压力测试系统
CN112346850A (zh) * 2019-08-09 2021-02-09 北京国双科技有限公司 一种配置内存的方法及装置
CN112346849A (zh) * 2019-08-09 2021-02-09 北京国双科技有限公司 一种配置cpu的方法及装置
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
CN110888792A (zh) * 2020-01-02 2020-03-17 大汉软件股份有限公司 一种节约政务服务器资源和并发性能提升的有效评估方法
US11561842B2 (en) * 2020-01-31 2023-01-24 Hewlett Packard Enterprise Development Lp Determining and implementing a feasible resource optimization plan for public cloud consumption
US11663505B2 (en) 2020-03-10 2023-05-30 International Business Machines Corporation Estimating performance and required resources from shift-left analysis
CN111431996B (zh) * 2020-03-20 2022-08-09 北京百度网讯科技有限公司 用于资源配置的方法、装置、设备和介质
CN113472552B (zh) * 2020-03-30 2023-04-25 中国移动通信集团浙江有限公司 基于深度学习的vim选择方法、装置、计算设备及存储介质
CN111610994B (zh) * 2020-05-20 2023-10-20 山东汇贸电子口岸有限公司 一种云数据中心的部署方法、装置、设备及存储介质
CN111666197A (zh) * 2020-06-05 2020-09-15 北京百度网讯科技有限公司 压力测试方法和装置、电子设备以及计算机可读介质
JP6943318B1 (ja) * 2020-06-05 2021-09-29 ダイキン工業株式会社 プログラム、情報処理方法、及び情報処理装置
US11750471B2 (en) 2020-06-16 2023-09-05 Siemens Aktiengesellschaft Method and apparatus for determining resource configuration of cloud service system
CN112085563A (zh) * 2020-09-04 2020-12-15 浪潮云信息技术股份公司 一种基于cnn的云硬盘资源管理方法
CN112416588B (zh) * 2020-11-20 2022-06-07 中国电子科技集团公司第二十九研究所 一种基于随机森林算法的资源分配方法
CN112788131A (zh) * 2020-12-31 2021-05-11 平安科技（深圳）有限公司 基于人工智能生成预警图片的方法、系统以及存储介质
CN112883654B (zh) * 2021-03-24 2023-01-31 国家超级计算天津中心 一种基于数据驱动的模型训练系统
CN113296951A (zh) * 2021-05-31 2021-08-24 阿里巴巴新加坡控股有限公司 一种资源配置方案确定方法及设备
CN113268350B (zh) * 2021-06-07 2024-01-26 上海数禾信息科技有限公司 基于云的服务构建的资源分配方法、装置、计算机设备
CN113407422B (zh) * 2021-08-20 2021-11-09 太平金融科技服务(上海)有限公司深圳分公司 数据异常告警处理方法、装置、计算机设备和存储介质
CN113791765B (zh) * 2021-09-15 2024-03-08 平安科技（深圳）有限公司 云服务的资源编排方法、装置、设备及存储介质
CN114283023B (zh) * 2021-12-29 2022-09-27 工业云制造(四川)创新中心有限公司 一种基于云制造支撑技术的制造管理方法及系统
CN115202829B (zh) * 2022-08-15 2023-11-21 中国电信股份有限公司 虚拟机的功耗预测模型训练方法、功耗预测方法及装置
CN117522176B (zh) * 2024-01-08 2024-04-12 中国电子科技集团公司第十五研究所 基于多智能体系统的多项目资源调度评估系统

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2018053717A1 (en) * 2016-09-21 2018-03-29 Accenture Global Solutions Limited Dynamic resource allocation for application containers
US20180203751A1 (en) * 2017-01-13 2018-07-19 Bank Of America Corporation Using a concentration risk of a computing resource to define affinity and anti-affinity workloads
CN108446200A (zh) * 2018-02-07 2018-08-24 福建星瑞格软件有限公司 基于大数据机器学习的服务器智能运维方法及计算机设备
CN109412829A (zh) * 2018-08-30 2019-03-01 华为技术有限公司 一种资源配置的预测方法及设备

Family Cites Families (10)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN1936849A (zh) * 2005-09-19 2007-03-28 国际商业机器公司 资源动态调整方法及设备
CN103699440B (zh) * 2012-09-27 2017-05-24 北京搜狐新媒体信息技术有限公司 一种云计算平台系统为任务分配资源的方法和装置
CN104714845B (zh) * 2013-12-17 2018-06-26 华为技术有限公司 资源动态调整方法、装置及多内核操作系统
WO2016070352A1 (en) * 2014-11-05 2016-05-12 Qualcomm Incorporated Managing resources for cooperative uplink transmission
CN107786616A (zh) * 2016-08-30 2018-03-09 江苏蓝创聚联数据与应用研究院有限公司 基于云端的主机智能监控系统
US10997135B2 (en) * 2016-09-16 2021-05-04 Oracle International Corporation Method and system for performing context-aware prognoses for health analysis of monitored systems
US11048879B2 (en) * 2017-06-19 2021-06-29 Vettd, Inc. Systems and methods to determine and utilize semantic relatedness between multiple natural language sources to determine strengths and weaknesses
US11397887B2 (en) * 2017-09-26 2022-07-26 Amazon Technologies, Inc. Dynamic tuning of training parameters for machine learning algorithms
US11483324B2 (en) * 2017-11-30 2022-10-25 Shape Security, Inc. Detection of malicious activity using behavior data
US10579494B2 (en) * 2018-01-05 2020-03-03 Nec Corporation Methods and systems for machine-learning-based resource prediction for resource allocation and anomaly detection

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2018053717A1 (en) * 2016-09-21 2018-03-29 Accenture Global Solutions Limited Dynamic resource allocation for application containers
US20180203751A1 (en) * 2017-01-13 2018-07-19 Bank Of America Corporation Using a concentration risk of a computing resource to define affinity and anti-affinity workloads
CN108446200A (zh) * 2018-02-07 2018-08-24 福建星瑞格软件有限公司 基于大数据机器学习的服务器智能运维方法及计算机设备
CN109412829A (zh) * 2018-08-30 2019-03-01 华为技术有限公司 一种资源配置的预测方法及设备

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112073239A (zh) * 2020-09-04 2020-12-11 天津大学 一种云计算环境分布式应用性能预测方法
WO2022246833A1 (en) * 2021-05-28 2022-12-01 Huawei Cloud Computing Technologies Co., Ltd. System, method, and medium for elastic allocation of resources for deep learning jobs

Also Published As

Publication number Publication date
EP3840295A1 (en) 2021-06-23
EP3840295A4 (en) 2021-08-11
US20210182106A1 (en) 2021-06-17
CN109412829A (zh) 2019-03-01
CN109412829B (zh) 2020-11-17

Similar Documents

Publication Publication Date Title
WO2020042721A1 (zh) 一种资源配置的预测方法及设备
US10243879B2 (en) Intelligent placement within a data center
US11422859B2 (en) Methods and systems for cloud application optimization
CN110417686B (zh) 云资源动态调度系统
US10680902B2 (en) Virtual agents for facilitation of network based storage reporting
US20200371896A1 (en) Exponential decay real-time capacity planning
US10691500B1 (en) Modeling workloads using micro workloads representing normalized units of resource consumption metrics
US20190095245A1 (en) System and Method for Apportioning Shared Computer Resources
US11381483B2 (en) Maintenance recommendation for containerized services
CN103365727A (zh) 一种云计算环境中的主机负载预测方法
CN111861793A (zh) 基于云边协同计算架构的配用电业务分配方法及装置
US20240020556A1 (en) Information processing method and apparatus, server, and user device
US11669374B2 (en) Using machine-learning methods to facilitate experimental evaluation of modifications to a computational environment within a distributed system
US20190215262A1 (en) System and method for dynamically testing networked target systems
WO2023197453A1 (zh) 一种故障诊断方法、装置、设备及存储介质
CN105786626B (zh) 基于K划分的Storm平台线程分配方法
WO2023050534A1 (zh) 轨道交通站点设备能耗预测方法、装置、设备和存储介质
WO2020206699A1 (en) Predicting virtual machine allocation failures on server node clusters
US20210263718A1 (en) Generating predictive metrics for virtualized deployments
CN108139942A (zh) 数据中心端到端功率效率建模的方法
US20230403210A1 (en) Methods and systems for cloud application optimization
CN113205128A (zh) 基于无服务器计算的分布式深度学习性能保证方法
WO2021127640A1 (en) Modeling cloud inefficiencies using domain-specific templates
Rayan et al. Resource Prediction for Big Data Processing in a Cloud Data Center: A Machine Learning Approach: A Machine Learning Approach
CN113822454B (zh) 切片投诉处理资源的预估方法及装置

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: 19855405

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: 2019855405

Country of ref document: EP

Effective date: 20210316