Cloud resource prediction methods, devices, equipment and media

Abstract

This application discloses a cloud resource prediction method, apparatus, device, and medium. The method includes: obtaining real-time cloud resource utilization; calling a pre-trained neural network model and loading model parameters; inputting the real-time cloud resource utilization into the neural network model, obtaining time-domain and frequency-domain features of the real-time cloud resource utilization through the neural network model, and obtaining a cloud resource prediction result based on the time-domain and frequency-domain features.

Description

Technical Field

[0001] This application relates to the fields of cloud computing and big data technology, specifically to a cloud resource prediction method, apparatus, device, and medium. Background Technology

[0002] Cloud computing is a model for providing computing resources and services via the internet. It combines computer technology with network technology, storing data and applications on remote servers for access and management over a network. In a cloud computing network, tenants can use cloud resources purchased from cloud service providers.

[0003] To ensure timely fulfillment of application needs and Quality of Service (QoS) requirements for a large number of tenants, cloud service providers need to predict cloud network resources. However, existing cloud resource prediction methods primarily use time-domain features for calculation, which may lead to inaccurate prediction results. Summary of the Invention

[0004] This application provides a cloud resource prediction method, apparatus, device, and medium to address the problem of inaccurate cloud resource prediction results when using time-domain features.

[0005] In a first aspect, embodiments of this application provide a cloud resource prediction method, the method comprising:

[0006] Get real-time cloud resource utilization;

[0007] Call the pre-trained neural network model and load the model parameters;

[0008] The real-time cloud resource utilization rate is input into the neural network model, and the time-domain and frequency-domain features of the real-time cloud resource utilization rate are obtained through the neural network model. Based on the time-domain and frequency-domain features, the cloud resource prediction result is obtained.

[0009] Secondly, embodiments of this application provide a cloud resource prediction device, the device comprising:

[0010] The acquisition module is used to obtain real-time cloud resource utilization.

[0011] The calling module is used to invoke a pre-trained neural network model and load the model parameters;

[0012] The cloud resource prediction module is used to input the real-time cloud resource utilization rate into the neural network model, obtain the time-domain and frequency-domain features of the real-time cloud resource utilization rate through the neural network model, and obtain the cloud resource prediction result based on the time-domain and frequency-domain features.

[0013] Thirdly, embodiments of this application provide an electronic device, including: one or more processors; and a storage device having one or more programs stored thereon, wherein when the one or more programs are executed by the one or more processors, the one or more processors implement the method as described in any embodiment of the first aspect.

[0014] Fourthly, embodiments of this application provide a computer-readable medium having a computer program stored thereon that, when executed by a processor, implements the method as described in any embodiment of the first aspect.

[0015] The cloud resource prediction method, apparatus, device, and medium provided in this application can obtain cloud resource prediction results based on the time-domain and frequency-domain characteristics of real-time cloud resource utilization through a pre-trained neural network model. It fully utilizes the following characteristics of frequency-domain features: frequency-domain features can capture the changing patterns of signals at different frequencies, thus providing more comprehensive time-series information; frequency-domain features can also improve the long-term dependency problem that neural network models often face when processing time-series data by providing information at different time scales, thereby capturing dependencies at different time scales; frequency-domain features can also reflect the energy distribution of signals at different frequencies, and this feature is less affected by outliers or noise, thus making the neural network model more robust to noise and outliers; furthermore, frequency-domain features can provide a finer-grained interpretation of time-series data, thereby improving the comprehensiveness and accuracy of cloud resource prediction results and solving the problem of inaccurate cloud resource prediction results when using time-domain features in existing technologies. Attached Figure Description

[0016] Other features, objects, and advantages of this application will become more apparent from the following detailed description of non-limiting embodiments with reference to the accompanying drawings:

[0017] Figure 1 This is a flowchart of an embodiment of the cloud resource prediction method of this application;

[0018] Figure 2 This is a flowchart of another embodiment of the cloud resource prediction method of this application;

[0019] Figure 3 yes Figure 2 One of the flowcharts for step 104 in the cloud resource prediction method of this application is shown;

[0020] Figure 4 yes Figure 2 The flowchart of step 104 in the cloud resource prediction method of this application is shown as a second one;

[0021] Figure 5 yes Figure 4 The flowchart of step 305 in step 104 is shown;

[0022] Figure 6 yes Figure 5 The diagram shown illustrates the TCN convolution calculation in step 305.

[0023] Figure 7 This is a schematic diagram of the structure of one embodiment of the cloud resource prediction device of this application;

[0024] Figure 8 This is a schematic diagram of the structure of an electronic device used to implement the embodiments of this application. Detailed Implementation

[0025] All actions involving the acquisition of signals, information, or data in this application are carried out in accordance with the relevant data protection laws and policies of the country where the application is located, and with the authorization of the owner of the relevant device.

[0026] The present application will now be described in further detail with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and not intended to limit it. Furthermore, it should be noted that, for ease of description, only the parts relevant to the invention are shown in the accompanying drawings.

[0027] It should be noted that, unless otherwise specified, the embodiments and features described in this application can be combined with each other. This application will now be described in detail with reference to the accompanying drawings and embodiments.

[0028] Please refer to Figure 1 The diagram illustrates a flow 100 of an embodiment of the cloud resource prediction method according to this application. This cloud resource prediction method can be applied to various electronic devices with data processing capabilities. For example, the aforementioned electronic devices may include, but are not limited to, cloud servers, physical servers, etc. The executing entity of the cloud resource prediction method may be a processor in the aforementioned electronic device.

[0029] like Figure 1 As shown, this cloud resource prediction method includes the following steps:

[0030] Step 101: Obtain real-time cloud resource utilization.

[0031] Step 102: Call the pre-trained neural network model and load the model parameters.

[0032] Step 103: Input the real-time cloud resource utilization rate into the neural network model, obtain the time domain and frequency domain features of the real-time cloud resource utilization rate through the neural network model, and obtain the cloud resource prediction results based on the time domain and frequency domain features.

[0033] The cloud resource prediction method provided in this application can obtain cloud resource prediction results based on the time-domain and frequency-domain characteristics of real-time cloud resource utilization through a pre-trained neural network model. It fully utilizes the following characteristics of frequency-domain characteristics: frequency-domain characteristics can capture the changing patterns of signals at different frequencies, thereby providing more comprehensive time-series information; frequency-domain characteristics can also improve the long-term dependency problem that neural network models often face when processing time-series data by providing information at different time scales, thereby capturing dependencies at different time scales; frequency-domain characteristics can also reflect the energy distribution of signals at different frequencies, and this feature is less affected by outliers or noise, thus making the neural network model more robust to noise and outliers; furthermore, frequency-domain characteristics can provide a finer-grained interpretation of time-series data, thereby improving the comprehensiveness and accuracy of cloud resource prediction results and solving the problem of inaccurate cloud resource prediction results when using time-domain characteristics in the prior art.

[0034] Please refer to Figure 2 The diagram illustrates a flow 100 of an embodiment of the cloud resource prediction method according to this application. This cloud resource prediction method can be applied to various electronic devices with data processing capabilities. For example, the aforementioned electronic devices may include, but are not limited to, cloud servers, physical servers, etc. The executing entity of the cloud resource prediction method may be a processor in the aforementioned electronic device.

[0035] like Figure 2 As shown, this cloud resource prediction method, and Figure 1 The two methods are essentially the same, except that, before step 101, the following may also be included:

[0036] Step 104: Construct a neural network model based on historical cloud resource utilization rates.

[0037] For example, such as Figure 3 As shown, step 104 may include:

[0038] Step 301: Obtain the set of historical cloud resource utilization rates within the first time period. The set of historical cloud resource utilization rates includes multiple historical cloud resource utilization rates.

[0039] Step 302: Construct a neural network model based on the historical cloud resource utilization rate set within the first time period.

[0040] Optional, such as Figure 4 As shown, after step 301, the following may also be included:

[0041] Step 303: Normalize the historical cloud resource utilization rate set to obtain the normalized historical cloud resource utilization rate set.

[0042] Specifically, step 303 can be expressed as: For the set of historical cloud resource utilization rates within the first time period T, M = {m1, m2, ..., m...} T The data is then normalized to obtain a set of normalized historical cloud resource utilization rates.

[0043] Where, m T Let T be the historical cloud resource utilization rate at time T; the normalization formula is: 1≤t≤T, m max Let M = {m1, m2, ..., m} T The maximum historical cloud resource utilization rate included in}.

[0044] Step 304: Using a sliding window mechanism, the normalized historical cloud resource utilization set is segmented to obtain multiple subsets of historical cloud resource utilization within the second time period.

[0045] In this embodiment, step 304 can use the second time period h as a window and adopt a sliding window mechanism to set the historical cloud resource utilization rate M = {m1, m2, ..., m}. T The data is divided into multiple subsets of historical cloud resource utilization rates: {m1, m2, ..., m} h},{m2,m3,…,m h+1},{m3,m4,…,m h+2},…,{m T-h+1 ,m T-h+2 ,…,m T The split data is denoted as dataset D.

[0046] At this point, step 302 can be replaced with: step 305, constructing a neural network model based on each subset of historical cloud resource utilization.

[0047] For example, such as Figure 5 As shown, step 305 may include the following steps:

[0048] Step 501: Perform a Fast Fourier Transform on the subset of historical cloud resource usage to obtain the frequency domain characteristics of the time-series data of the subset of historical cloud resource usage.

[0049] For example, step 501 can be achieved through the following steps:

[0050] First, for the second time period h, the one-dimensional time series D with channel dimension C. 1D ∈R h×c Time-series data of a subset of historical cloud resource utilization {m1,m2,…,m h Using the Fast Fourier Transform, obtain the Fourier Transform result {M}. k} is: {Mk}:=M1,M1,…,M h ,in,

[0051] Secondly, the Fourier transform result is decomposed into a sequence of real parts and a sequence of imaginary parts; M k =M kr +j·M ki ;

[0052] Third, perform conjugation processing on the decomposed Fourier transform results to obtain the conjugation processing result:

[0053] {M k} * ={M k} * :=M1 * M2 * ,…,M h *

[0054] in,

[0055] Fourth, by multiplying the decomposed Fourier transform result with the conjugate processing result, the Fourier features are obtained:

[0056]

[0057] Fifth, perform two repeated one-dimensional convolutions on the Fourier features, and then process them with the Rectified Linear Unit (ReLU) activation function, i.e.: F′=ReLU[G conv1d (F)],F″=ReLU[G conv1d [F′], obtain the frequency domain features F″ of the time-series data of the historical cloud resource usage subset.

[0058] Step 502: Use a Temporal Convolutional Network (TCN) to process the historical cloud resource usage subset and obtain the temporal domain features of the time series data of the historical cloud resource usage subset.

[0059] like Figure 6 The image shows a schematic diagram of TCN convolution calculation. According to... Figure 6 As shown, for example, step 502 can be implemented through the following steps:

[0060] First, based on the time-series data of a subset of historical cloud resource utilization {m1,m2,…,m h} and convolution kernel f:{0,...,k-1}, after dilated convolution, obtain the output of l neurons:

[0061]

[0062] Where k is the kernel size, d is the dilation factor, f(i) is the i-th element in the kernel, and x l-d·i For time-series data that are multiplied with the corresponding elements in the convolution kernel;

[0063] Secondly, the neuron outputs are processed through three consecutive TCN structures for deep feature extraction to obtain the temporal features of the time-series data of the historical cloud resource usage subset.

[0064] In this embodiment, the TCN feature extraction expressions for the three stages can be denoted as: G 1 TCN G 2 TCN G 3 TCN .

[0065] Step 503: Perform tensor concatenation operations on the frequency features and time-domain features in the channel dimension, using skip concatenation, to fuse the frequency features and time-domain features and obtain the fused time-domain and frequency-domain features.

[0066] For example, step 503 can use the temporal features G of the first-stage TCN. 1 TCN The temporal characteristics G of the third-stage TCN 3 TCN The time-domain and frequency-domain fused features H = Concat(G) are then fused with the frequency domain feature F″ to obtain the time-domain and frequency-domain fused features. 1 TCN G 3 TCN ,F″), where Concat represents concatenating tensors along the channel dimension.

[0067] Step 504: Perform TCN calculation on the time-domain and frequency-domain fused features to obtain the prediction results corresponding to the subset of historical cloud resource utilization rates.

[0068] Optional, such as Figure 2 As shown, after step 104, the following may also be included:

[0069] Step 105: Train the neural network model and obtain the model parameters.

[0070] For example, step 105 may include:

[0071] First, the target loss function for the backpropagation direction is obtained as the root mean square error:

[0072]

[0073] in, Let i be the predicted value corresponding to the i-th cloud resource utilization sample. is the actual value corresponding to the i-th cloud resource utilization sample, and n is the number of samples;

[0074] Secondly, the model parameters in the application network model are updated using the Stochastic Gradient Descent (SGD) method to minimize the objective loss function and obtain the global optimum. Once the neural network model has converged, the training of the neural network model is stopped, and the model parameters are obtained.

[0075] The cloud resource prediction method provided in this application can obtain cloud resource prediction results based on the time-domain and frequency-domain characteristics of real-time cloud resource utilization through a pre-trained neural network model. It fully utilizes the following characteristics of frequency-domain characteristics: frequency-domain characteristics can capture the changing patterns of signals at different frequencies, thereby providing more comprehensive time-series information; frequency-domain characteristics can also improve the long-term dependency problem that neural network models often face when processing time-series data by providing information at different time scales, thereby capturing dependencies at different time scales; frequency-domain characteristics can also reflect the energy distribution of signals at different frequencies, and this feature is less affected by outliers or noise, thus making the neural network model more robust to noise and outliers; furthermore, frequency-domain characteristics can provide a finer-grained interpretation of time-series data, thereby improving the comprehensiveness and accuracy of cloud resource prediction results and solving the problem of inaccurate cloud resource prediction results when using time-domain characteristics in the prior art.

[0076] Please refer to Figure 7 As an implementation of the methods shown in the figures, this application provides an embodiment of a cloud resource prediction device, which corresponds to the methods shown in the above embodiments and can be specifically applied to various electronic devices.

[0077] like Figure 7 As shown, the cloud resource prediction device 700 of this embodiment includes:

[0078] Module 701 is used to obtain real-time cloud resource utilization.

[0079] Module 702 is used to call a pre-trained neural network model and load the model parameters;

[0080] The cloud resource prediction module 703 is used to input the real-time cloud resource utilization rate into the neural network model, obtain the time-domain and frequency-domain features of the real-time cloud resource utilization rate through the neural network model, and obtain the cloud resource prediction result based on the time-domain and frequency-domain features.

[0081] Optionally, the cloud resource prediction device 700 in this embodiment may further include:

[0082] The neural network model building module is used to obtain a set of historical cloud resource utilization rates within a first time period, wherein the set of historical cloud resource utilization rates includes multiple historical cloud resource utilization rates; and to build the neural network model based on the set of historical cloud resource utilization rates within the first time period.

[0083] Optionally, the neural network model construction module is further configured to normalize the historical cloud resource utilization rate set to obtain a normalized historical cloud resource utilization rate set; use a sliding window mechanism to segment the normalized historical cloud resource utilization rate set to obtain multiple subsets of historical cloud resource utilization rates within a second time period; and construct the neural network model based on each subset of historical cloud resource utilization rates.

[0084] In this embodiment, the normalization process performed by the neural network model construction module can be specifically represented as follows:

[0085] For the first time period T, the set of historical cloud resource utilization rates is M = {m1, m2, ..., m}. T The data is then normalized to obtain a set of normalized historical cloud resource utilization rates.

[0086] Where, m T Let T be the historical cloud resource utilization rate at time T; the normalization formula is: 1≤t≤T, m max Let M = {m1, m2, ..., m} T The maximum historical cloud resource utilization rate included in}.

[0087] In this embodiment, the neural network model building module can use a sliding window mechanism, taking the second time period h as a window, to construct a set of historical cloud resource utilization rates M = {m1, m2, ..., m}. T The data is divided into multiple subsets of historical cloud resource utilization rates: {m1, m2, ..., m} h},{m2,m3,…,m h+1},{m3,m4,…,m h+2},…,{m T-h+1 ,m T-h+2 ,…,m T The split data is denoted as dataset D.

[0088] Optionally, the neural network model construction module is further configured to perform a Fast Fourier Transform on the historical cloud resource utilization subset to obtain the frequency domain features of the time-series data of the historical cloud resource utilization subset; process the historical cloud resource utilization subset using a Temporal Convolutional Network (TCN) to obtain the temporal domain features of the time-series data of the historical cloud resource utilization subset; perform a tensor concatenation operation on the frequency features and the temporal features in the channel dimension, using skip connections, to fuse the frequency features and the temporal features to obtain the time-domain and frequency-domain fused features; and perform TCN calculation on the time-domain and frequency-domain fused features to obtain the prediction result corresponding to the historical cloud resource utilization subset.

[0089] Optionally, the neural network model building module is further used to process the second time period h and the one-dimensional time series D with channel dimension C. 1D ∈R h×c The time-series data of the historical cloud resource utilization subset {m1,m2,…,m h Using the Fast Fourier Transform, obtain the Fourier Transform result {M}. k} is: {M k}:=M1,M1,…,M h ,in,

[0090] The Fourier transform result is decomposed into a sequence of real parts and a sequence of imaginary parts; M k =M kr +j·M ki ;

[0091] Perform conjugate processing on the decomposed Fourier transform results to obtain the conjugate processing result:

[0092]

[0093] in,

[0094] By multiplying the decomposed Fourier transform result with the conjugate processing result, Fourier features are obtained:

[0095]

[0096] The Fourier features are subjected to two repeated one-dimensional convolutions, and then processed with a modified linear unit (ReLU) activation function, i.e.: F′=ReLU[G conv1d (F)],F″=ReLU[G conv1d (F′)], obtain the frequency domain features F″ of the time-series data of the historical cloud resource usage subset.

[0097] Optionally, the neural network model building module is further configured to, based on the time-series data {m1, m2, ..., m} of the historical cloud resource utilization subset, h} and convolution kernel f:{0,...,k-1}, after dilated convolution, obtain the output of l neurons:

[0098]

[0099] Where k is the kernel size, d is the dilation factor, f(i) is the i-th element in the kernel, and x l-d·i For time-series data that are multiplied with the corresponding elements in the convolution kernel;

[0100] The neuron output is processed through three consecutive TCN structures for deep feature extraction to obtain the temporal features of the time-series data of the historical cloud resource usage subset.

[0101] In this embodiment, the TCN feature extraction expressions for the three stages can be denoted as: G 1 TCN G 2 TCN G 3 TCN .

[0102] For example, the neural network model building module can use the temporal features G of the first-stage TCN. 1 TCN The temporal characteristics G of the third-stage TCN 3 TCN The time-domain and frequency-domain fused features H = Concat(G) are then fused with the frequency domain feature F″ to obtain the time-domain and frequency-domain fused features. 1 TCN G 3 TCN ,F″), where Concat represents concatenating tensors along the channel dimension.

[0103] Optionally, the cloud resource prediction device 700 in this embodiment may further include:

[0104] A neural network model training module, used to train the neural network model and obtain the model parameters, including:

[0105] The target loss function for obtaining the backpropagation direction is the root mean square error:

[0106]

[0107] in, Let i be the predicted value corresponding to the i-th cloud resource utilization sample. is the actual value corresponding to the i-th cloud resource utilization sample, and n is the number of samples;

[0108] The model parameters in the proposed network model are updated using the gradient descent (SGD) method to minimize the objective loss function and obtain the global optimum. Once the neural network model has converged, training is stopped, and the model parameters are retrieved.

[0109] For a detailed implementation method of the cloud resource prediction device provided in this application embodiment, please refer to the above. Figure 1-6 The cloud resource prediction method provided in the embodiments of this application shown herein will not be described in detail here.

[0110] The cloud resource prediction device provided in this application can obtain cloud resource prediction results based on the time-domain and frequency-domain characteristics of real-time cloud resource utilization through a pre-trained neural network model. It fully utilizes the following characteristics of frequency-domain characteristics: frequency-domain characteristics can capture the changing patterns of signals at different frequencies, thereby providing more comprehensive time-series information; frequency-domain characteristics can also improve the long-term dependency problem that neural network models often face when processing time-series data by providing information at different time scales, thereby capturing the dependency relationship at different time scales; frequency-domain characteristics can also reflect the energy distribution of signals at different frequencies, and this feature is less affected by outliers or noise, thus making the neural network model more robust to noise and outliers; furthermore, frequency-domain characteristics can provide a finer-grained interpretation of time-series data, thereby improving the comprehensiveness and accuracy of cloud resource prediction results and solving the problem of inaccurate cloud resource prediction results when using time-domain characteristics in the prior art.

[0111] The following is for reference. Figure 8 It shows a schematic diagram of the structure of an electronic device used to implement some embodiments of this application. Figure 8 The electronic device shown is merely an example and should not be construed as limiting the functionality and scope of the embodiments of this application.

[0112] like Figure 8 As shown, the electronic device 800 may include a processing device (e.g., a central processing unit, a graphics processor, etc.) 801, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 802 or a program loaded from a storage device 808 into a random access memory (RAM) 803. The RAM 803 also stores various programs and data required for the operation of the electronic device 800. The processing device 801, ROM 802, and RAM 803 are interconnected via a bus 804. An input / output (I / O) interface 805 is also connected to the bus 804.

[0113] Typically, the following devices can be connected to I / O interface 805: input devices 806 including, for example, touchscreens, touchpads, keyboards, mice, cameras, microphones, accelerometers, gyroscopes, etc.; output devices 807 including, for example, liquid crystal displays (LCDs), speakers, vibrators, etc.; storage devices 808 including, for example, disks, hard disks, etc.; and communication devices 809. Communication device 809 allows electronic device 800 to communicate wirelessly or wiredly with other devices to exchange data. Although Figure 7 An electronic device 800 with various devices is shown; however, it should be understood that it is not required to implement or possess all of the devices shown. More or fewer devices may be implemented or possessed alternatively. Figure 4 Each box shown can represent a device or multiple devices as needed.

[0114] In particular, according to some embodiments of this application, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, some embodiments of this application include a computer program product comprising a computer program carried on a computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via a communication device 809, or installed from a storage device 808, or installed from a ROM 802. When the computer program is executed by the processing device 801, it performs the functions defined in the methods of some embodiments of this application.

[0115] It should be noted that the computer-readable medium described in some embodiments of this application may be a computer-readable signal medium or a computer-readable storage medium, or any combination thereof. A computer-readable storage medium may be, for example,—but not limited to—an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of a computer-readable storage medium may include, but are not limited to: an electrical connection having one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof. In some embodiments of this application, a computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device. In some embodiments of this application, a computer-readable signal medium may include a data signal propagated in baseband or as part of a carrier wave, carrying computer-readable program code. Such propagated data signals may take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. A computer-readable signal medium can be any computer-readable medium other than a computer-readable storage medium, which can send, propagate, or transmit a program for use by or in connection with an instruction execution system, apparatus, or device. The program code contained on the computer-readable medium can be transmitted using any suitable medium, including but not limited to: wires, optical fibers, RF (radio frequency), etc., or any suitable combination thereof.

[0116] In some implementations, clients and servers can communicate using any currently known or future-developed network protocol, such as HTTP (Hypertext Transfer Protocol), and can interconnect with digital data communication (e.g., communication networks) of any form or medium. Examples of communication networks include local area networks (“LANs”), wide area networks (“WANs”), the Internet (e.g., the Internet of Things), and peer-to-peer networks (e.g., ad hoc peer-to-peer networks), as well as any currently known or future-developed networks.

[0117] The aforementioned computer-readable medium may be included in the aforementioned electronic device; or it may exist independently and not assembled into the electronic device. The aforementioned computer-readable medium carries one or more programs that, when executed by the electronic device, cause the electronic device to: acquire real-time cloud resource utilization; invoke a pre-trained neural network model and load model parameters; input the real-time cloud resource utilization into the neural network model; acquire the time-domain and frequency-domain features of the real-time cloud resource utilization through the neural network model; and acquire cloud resource prediction results based on the time-domain and frequency-domain features.

[0118] Computer program code for performing operations of some embodiments of this application can be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, and C++; and conventional procedural programming languages ​​such as the "C" language or similar programming languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving remote computers, the remote computer can be connected to the user's computer via any type of network, or it can be connected to an external computer (e.g., via the Internet using an Internet service provider), including local area networks (LANs) or wide area networks (WANs).

[0119] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of this application. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0120] The units described in some embodiments of this application can be implemented in software or hardware. The described units can also be housed in a processor; for example, a processor may be described as including a first determining unit, a second determining unit, a selecting unit, and a third determining unit. The names of these units do not necessarily limit the specific unit itself.

[0121] The functions described above in this document can be performed, at least in part, by one or more hardware logic components. For example, exemplary types of hardware logic components that can be used, without limitation, include: Field Programmable Gate Arrays (FPGAs), Application-Specific Integrated Circuits (ASICs), Application Standard Products (ASSPs), System-on-Chip (SoCs), Complex Programmable Logic Devices (CPLDs), and so on.

[0122] The above description is merely a selection of preferred embodiments of this application and an explanation of the technical principles employed. Those skilled in the art should understand that the scope of the invention involved in the embodiments of this application is not limited to technical solutions formed by specific combinations of the above-described technical features, but should also cover other technical solutions formed by arbitrary combinations of the above-described technical features or their equivalents without departing from the above-described inventive concept. For example, technical solutions formed by substituting the above-described features with (but not limited to) technical features with similar functions disclosed in the embodiments of this application.

Claims

1. A cloud resource prediction method, characterized in that, The method includes: Get real-time cloud resource utilization; Call the pre-trained neural network model and load the model parameters; The real-time cloud resource utilization rate is input into the neural network model, and the time-domain and frequency-domain features of the real-time cloud resource utilization rate are obtained through the neural network model. Based on the time-domain and frequency-domain features, the cloud resource prediction result is obtained. Before obtaining the real-time cloud resource utilization rate, the following steps are also included: Obtain a set of historical cloud resource utilization rates within a first time period; construct the neural network model based on the set of historical cloud resource utilization rates within the first time period, including: constructing the neural network model based on each subset of historical cloud resource utilization rates; The step of constructing the neural network model based on each subset of historical cloud resource utilization includes: For the second time period h, a one-dimensional time series with channel dimension C The time-series data of the subset of historical cloud resource utilization Use Fast Fourier Transform to obtain the Fourier Transform result. for: ,in, ; The Fourier transform result is decomposed into a sequence of real parts and a sequence of imaginary parts; ; Perform conjugate processing on the decomposed Fourier transform results to obtain the conjugate processing result: in, ; By multiplying the decomposed Fourier transform result with the conjugate processing result, Fourier features are obtained: The Fourier features are subjected to two repeated one-dimensional convolutions, and then processed with a modified linear unit ReLU activation function to obtain the frequency domain features of the time-series data of the historical cloud resource usage subset. The historical cloud resource usage subset is processed using a Temporal Convolutional Network (TCN) to obtain the temporal domain features of the time-series data of the historical cloud resource usage subset. The frequency domain features and the time domain features are joined by tensors in the channel dimension using a skip connection to fuse the frequency domain features and the time domain features, thereby obtaining the time domain and frequency domain fused features. TCN calculation is performed on the time-domain and frequency-domain fused features to obtain the prediction results corresponding to the subset of historical cloud resource utilization rates.

2. The method according to claim 1, characterized in that, The set of historical cloud resource utilization rates includes multiple historical cloud resource utilization rates.

3. The method according to claim 2, characterized in that, After obtaining the set of historical cloud resource utilization rates for the first time period, the method further includes: The historical cloud resource utilization rate set is normalized to obtain a normalized historical cloud resource utilization rate set. A sliding window mechanism is used to segment the normalized historical cloud resource utilization set to obtain multiple subsets of historical cloud resource utilization within a second time period.

4. The method according to claim 1, characterized in that, The Temporal Convolutional Network (TCN) is used to process the historical cloud resource usage subset to obtain the temporal features of the time-series data of the historical cloud resource usage subset, including: Based on the time-series data of the aforementioned historical cloud resource utilization subset and convolution kernel After dilated convolution, the outputs of l neurons are obtained: Where k is the kernel size and d is the dilation factor. The i-th element in the convolution kernel For time-series data that are multiplied with the corresponding elements in the convolution kernel; The neuron output is processed through three consecutive TCN structures for deep feature extraction to obtain the temporal features of the time-series data of the historical cloud resource usage subset.

5. The method according to claim 1, characterized in that, The method further includes the step of training the neural network model to obtain the model parameters, the step including... The target loss function for obtaining the backpropagation direction is the root mean square error: in, Let i be the predicted value corresponding to the i-th cloud resource utilization sample. is the actual value corresponding to the i-th cloud resource utilization sample, and n is the number of samples; The model parameters in the neural network model are updated using the gradient descent (SGD) method to minimize the objective loss function and obtain the global optimum. Once the neural network model has converged, training is stopped, and the model parameters are retrieved.

6. A cloud resource prediction device, characterized in that, The device includes: The acquisition module is used to obtain real-time cloud resource utilization. The calling module is used to invoke a pre-trained neural network model and load the model parameters; The cloud resource prediction module is used to input the real-time cloud resource utilization rate into the neural network model, obtain the time domain features and frequency domain features of the real-time cloud resource utilization rate through the neural network model, and obtain the cloud resource prediction result based on the time domain features and the frequency domain features. A neural network model building module is used to obtain a set of historical cloud resource utilization rates within a first time period; and to build the neural network model based on the set of historical cloud resource utilization rates within the first time period, including: building the neural network model based on each subset of historical cloud resource utilization rates. The step of constructing the neural network model based on each subset of historical cloud resource utilization includes: For the second time period h, a one-dimensional time series with channel dimension C The time-series data of the subset of historical cloud resource utilization Use Fast Fourier Transform to obtain the Fourier Transform result. for: ,in, ; The Fourier transform result is decomposed into a sequence of real parts and a sequence of imaginary parts; ; Perform conjugate processing on the decomposed Fourier transform results to obtain the conjugate processing result: in, ; By multiplying the decomposed Fourier transform result with the conjugate processing result, Fourier features are obtained: The Fourier features are subjected to two repeated one-dimensional convolutions, and then processed with a modified linear unit ReLU activation function to obtain the frequency domain features of the time-series data of the historical cloud resource usage subset. The historical cloud resource usage subset is processed using a Temporal Convolutional Network (TCN) to obtain the temporal domain features of the time-series data of the historical cloud resource usage subset. The frequency domain features and the time domain features are joined by tensors in the channel dimension using a skip connection to fuse the frequency domain features and the time domain features, thereby obtaining the time domain and frequency domain fused features. TCN calculation is performed on the time-domain and frequency-domain fused features to obtain the prediction results corresponding to the subset of historical cloud resource utilization rates.

7. An electronic device, characterized in that, include: One or more processors; Storage device, on which one or more programs are stored, When the one or more programs are executed by the one or more processors, the one or more processors implement the method as described in any one of claims 1-5.

8. A computer-readable medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the method as described in any one of claims 1-5.

Images