Multimedia intelligent dispatching machine and method thereof
Through data acquisition and timing correlation analysis of multimedia intelligent scheduler, the problem of unstable resource scheduling of multimedia system under traditional manual management is solved, automated management and real-time monitoring are realized, and system stability and performance are improved.
Patent Information
- Application Number
- CN202310756083.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2023-06-26
- Publication Date
- 2024-11-15
- Estimated Expiration
- 2043-06-26
AI Technical Summary
Traditional manual management methods are difficult to achieve efficient resource scheduling and real-time monitoring of multimedia systems, resulting in unstable system operation and prone to failure or crashes.
Design a multimedia intelligent dispatcher to obtain CPU share, memory share, disk storage and network bandwidth values through the data acquisition module, perform timing correlation analysis, and generate operation status feature vectors to detect the operation status of the multimedia intelligent dispatcher Is it normal?
It realizes the automated resource management and scheduling of the multimedia system, provides real-time monitoring and reporting functions, improves the stability and performance of the system, and avoids errors and inefficiency of manual monitoring.
Smart Images

Figure CN117215766B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of intelligent dispatching, and in particular to a multimedia intelligent dispatching machine and a method thereof. Background Art
[0002] With the continuous development of multimedia technology, multimedia systems play an increasingly important role in our daily lives. For a large multimedia system, how to efficiently schedule and monitor resources is a very critical issue.
[0003] At present, the traditional manual management method mainly monitors and schedules various resources in multimedia systems manually. However, manual monitoring and scheduling methods are prone to errors, which affect the normal operation of the entire multimedia system. At the same time, in large-scale multimedia systems, manual monitoring is difficult to fully grasp the usage of various resources in a timely and accurate manner, and there are failures or even crashes caused by monitoring loopholes. In addition, the types of resources that need to be scheduled in multimedia systems are numerous and cover a wide range, and the efficiency of manual scheduling is usually relatively low. The traditional manual management method cannot achieve real-time monitoring and dynamic scheduling, resulting in the system's inability to respond quickly.
[0004] Therefore, an optimized multimedia intelligent scheduling machine is desired. Summary of the invention
[0005] The embodiment of the present invention provides a multimedia intelligent scheduling machine and method thereof, which obtains the CPU occupancy rate, memory occupancy rate, disk storage capacity and network bandwidth value at multiple predetermined time points within a predetermined time period; performs time series correlation analysis on the CPU occupancy rate, memory occupancy rate, disk storage capacity and network bandwidth value at the multiple predetermined time points to obtain an operation status feature vector; and determines whether the operation status of the multimedia intelligent scheduling machine is normal based on the operation status feature vector. In this way, the use of the multimedia intelligent scheduling machine can be detected through comprehensive analysis, so as to determine whether the operation status of the multimedia intelligent scheduling machine is normal, thereby ensuring the normal operation of the multimedia intelligent scheduling machine and improving the stability and performance of the multimedia system.
[0006] The embodiment of the present invention further provides a multimedia intelligent scheduling machine, which includes:
[0007] A data collection module is used to obtain the CPU occupancy rate, memory occupancy rate, disk storage capacity and network bandwidth values at multiple predetermined time points within a predetermined time period;
[0008] a data processing module, configured to perform time series correlation analysis on the CPU occupancy rate, memory occupancy rate, disk storage capacity, and network bandwidth values at the plurality of predetermined time points to obtain an operation status feature vector; and
[0009] The running state detection module is used to determine whether the running state of the multimedia intelligent dispatching machine is normal based on the running state feature vector.
[0010] The embodiment of the present invention also provides a multimedia intelligent scheduling method, which includes:
[0011] Get the CPU usage, memory usage, disk storage and network bandwidth values at multiple scheduled time points within a scheduled time period;
[0012] Performing time series correlation analysis on the CPU occupancy rate, memory occupancy rate, disk storage capacity and network bandwidth values at the plurality of predetermined time points to obtain an operation status feature vector; and
[0013] Based on the running status feature vector, it is determined whether the running status of the multimedia intelligent dispatching machine is normal. BRIEF DESCRIPTION OF THE DRAWINGS
[0014] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the prior art descriptions. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work. In the drawings:
[0015] Figure 1 The present invention provides a block diagram of a multimedia intelligent scheduling machine in an embodiment of the present invention.
[0016] Figure 2 The present invention provides a block diagram of a data processing module in a multimedia intelligent scheduling machine according to an embodiment of the present invention.
[0017] Figure 3 The present invention provides a block diagram of the operation status detection module in a multimedia intelligent scheduling machine according to an embodiment of the present invention.
[0018] Figure 4 The present invention provides a flowchart of a multimedia intelligent scheduling method in an embodiment of the present invention.
[0019] Figure 5 A schematic diagram of a system architecture of a multimedia intelligent scheduling method provided in an embodiment of the present invention.
[0020] Figure 6 This is an application scenario diagram of a multimedia intelligent scheduling machine provided in an embodiment of the present invention. DETAILED DESCRIPTION
[0021] To make the purpose, technical solution and advantages of the embodiments of the present invention more clear, the embodiments of the present invention are further described in detail below in conjunction with the accompanying drawings. Here, the exemplary embodiments of the present invention and their descriptions are used to explain the present invention, but are not intended to limit the present invention.
[0022] Unless otherwise specified, all technical and scientific terms used in the embodiments of the present application have the same meaning as those commonly understood by those skilled in the art of the present application. The terms used in this application are only for the purpose of describing specific embodiments and are not intended to limit the scope of this application.
[0023] In the embodiments of the present application, it should be noted that, unless otherwise specified and limited, the term "connection" should be understood in a broad sense. For example, it can be an electrical connection or a connection between two components. It can be a direct connection or an indirect connection through an intermediate medium. For ordinary technicians in this field, the specific meanings of the above terms can be understood according to the specific circumstances.
[0024] It should be noted that the terms "first\second\third" involved in the embodiments of the present application are only used to distinguish similar objects, and do not represent a specific order for the objects. It is understandable that the specific order or sequence of "first\second\third" can be interchanged where permitted. It should be understood that the objects distinguished by "first\second\third" can be interchanged where appropriate, so that the embodiments of the present application described herein can be implemented in an order other than those illustrated or described herein.
[0025] In one embodiment of the present invention, Figure 1 FIG. 1 is a block diagram of a multimedia intelligent scheduling machine provided in an embodiment of the present invention. Figure 1 As shown, the multimedia intelligent scheduling machine 100 according to an embodiment of the present invention includes: a data acquisition module 110, which is used to obtain the CPU occupancy rate, memory occupancy rate, disk storage capacity and network bandwidth values at multiple predetermined time points within a predetermined time period; a data processing module 120, which is used to perform time series correlation analysis on the CPU occupancy rate, memory occupancy rate, disk storage capacity and network bandwidth values at the multiple predetermined time points to obtain an operation status feature vector; and an operation status detection module 130, which is used to determine whether the operation status of the multimedia intelligent scheduling machine is normal based on the operation status feature vector.
[0026] Among them, the data acquisition module 110 can monitor and analyze the operating status of the multimedia intelligent dispatcher by collecting these data, so as to find and solve problems in time. It can quickly collect a large amount of data and pass it to the data processing module. The data processing module 120 can quickly process a large amount of data and extract useful information. By analyzing the data, the operating status of the multimedia intelligent dispatcher can be understood, and problems can be found and solved in time, thereby improving the reliability and stability of the system. The operating status detection module 130 can find and solve problems in time. By detecting the operating status, the normal operation of the multimedia intelligent dispatcher can be guaranteed, and the reliability and stability of the system can be improved.
[0027] The multimedia intelligent scheduling machine 100 monitors and analyzes the CPU occupancy, memory occupancy, disk storage capacity and network bandwidth values at multiple predetermined time points through the above three modules, so as to timely discover and solve problems and improve the reliability and stability of the system.
[0028] Specifically, the data acquisition module 110 is used to obtain the CPU occupancy rate, memory occupancy rate, disk storage capacity and network bandwidth value at multiple predetermined time points within a predetermined time period. With the continuous development of multimedia technology, multimedia systems play an increasingly important role in our daily lives. For a large multimedia system, how to perform efficient resource scheduling and monitoring is a very critical issue.
[0029] At present, the traditional manual management method mainly monitors and schedules various resources in multimedia systems manually. However, the manual monitoring and scheduling methods are prone to errors, which affect the normal operation of the entire multimedia system. At the same time, in large-scale multimedia systems, manual monitoring is difficult to fully grasp the usage of various resources in a timely and accurate manner, and there are failures or even crashes caused by monitoring loopholes. In addition, the types of resources that need to be scheduled in multimedia systems are numerous and cover a wide range, and the efficiency of manual scheduling is usually relatively low. The traditional manual management method cannot achieve real-time monitoring and dynamic scheduling, resulting in the system's inability to respond quickly. Therefore, an optimized multimedia intelligent scheduling machine is desired.
[0030] The multimedia intelligent dispatcher is a system for managing and dispatching multiple media resources. It can automatically manage and dispatch different types of media resources, such as audio, video, images, etc. Moreover, the system can collect media resources from multiple sources and store them in a central media library. Then, it can automatically select and allocate appropriate media resources according to user needs to meet various needs, such as advertising playback, information release, entertainment, etc. In this way, various resources can be efficiently managed through resource allocation and scheduling, improving system efficiency and resource utilization.
[0031] Not only that, the multimedia intelligent dispatcher can also provide real-time monitoring and reporting functions so that administrators can understand the system's operation and resource usage. It should be understood that system resource utilization refers to the usage of various resources in the system (such as CPU, memory, disk, network bandwidth, etc.). If the utilization of certain resources is too high, it may cause a bottleneck in the system and need to be adjusted.
[0032] Based on this, in the technical solution of the present application, it is expected to detect the usage of the multimedia intelligent scheduling machine based on a comprehensive analysis of various resources in the system resource utilization, such as CPU, memory, disk and network bandwidth, so as to determine whether the operating status of the multimedia intelligent scheduling machine is normal, thereby ensuring the normal operation of the multimedia intelligent scheduling machine and improving the stability and performance of the multimedia system.
[0033] Specifically, in the technical solution of the present application, first, the CPU occupancy rate, memory occupancy rate, disk storage capacity and network bandwidth values at multiple predetermined time points within a predetermined time period are obtained. It should be understood that the system resource utilization rate refers to the usage of various resources (such as CPU, memory, disk, network bandwidth, etc.) in the system. If the utilization rate of certain resources is too high, it may cause bottlenecks in the system and need to be adjusted. Therefore, it is expected to judge the operating load and operating state of the system based on the timing analysis of the CPU occupancy rate, memory occupancy rate, disk storage capacity and network bandwidth value. In other words, when the system operating state is unstable, the current resource usage of the system can be better evaluated by collecting various resource indicators in the system to determine whether adaptive adjustments are needed to optimize system performance. For example, when performing high CPU occupancy tasks such as video encoding and decoding, monitoring the changes in CPU occupancy rate can help the intelligent scheduling machine determine whether the current system has bottlenecks and take corresponding measures to avoid system crashes caused by excessive load; at the same time, monitoring data such as disk storage capacity and network bandwidth value can help the intelligent scheduling machine estimate the running time.
[0034] Specifically, the data processing module 120 is used to perform time series correlation analysis on the CPU occupancy rate, memory occupancy rate, disk storage capacity and network bandwidth values at the plurality of predetermined time points to obtain an operation status feature vector. Figure 2 A block diagram of a data processing module in a multimedia intelligent scheduling machine provided in an embodiment of the present invention, such as Figure 2As shown, the data processing module 120 includes: a data timing arrangement unit 121, which is used to arrange the CPU occupancy, memory occupancy, disk storage capacity and network bandwidth values of the multiple predetermined time points into a resource multi-parameter timing matrix according to the time dimension and the sample dimension; a context resource multi-parameter timing association feature extraction unit 122, which is used to extract features from the resource multi-parameter timing matrix to obtain multiple context resource multi-parameter timing association feature vectors; and a local association enhancement unit 123, which is used to perform local association feature enhancement on the multiple context resource multi-parameter timing association feature vectors to obtain the operating status feature vector. In the data timing arrangement unit 121, considering that the CPU occupancy rate, memory occupancy rate, disk storage capacity and network bandwidth values have a dynamic change law of timing in the time dimension, the CPU occupancy rate, memory occupancy rate, disk storage capacity and network bandwidth values at the multiple predetermined time points are further arranged into a resource multi-parameter timing matrix according to the time dimension and the sample dimension, so as to respectively integrate the timing distribution information of the CPU occupancy rate, memory occupancy rate, disk storage capacity and network bandwidth value in the time dimension, so as to facilitate the subsequent timing change correlation feature analysis of various resources in the system and the operation status detection of the multimedia intelligent scheduling machine.
[0035] In the resource multi-parameter time series matrix, the time dimension refers to a column of data formed by arranging the CPU occupancy rate, memory occupancy rate, disk storage capacity, and network bandwidth values at multiple predetermined time points in chronological order. For example, if it is scheduled to collect these data once every minute, the time dimension is the data collected every minute, forming a time series. The arrangement of the time dimension facilitates the analysis of the time series changes of various resources in the system.
[0036] The sample dimension refers to a row of data formed by arranging the CPU occupancy, memory occupancy, disk storage, and network bandwidth values at multiple predetermined time points in the order of samples. For example, if these data are collected from multiple servers, the sample dimension is the data collected from each server, forming a sample sequence. The arrangement of the sample dimension facilitates the analysis of the correlation characteristics between different resources.
[0037] In one embodiment of the present application, 1. The CPU occupancy rate, memory occupancy rate, disk storage capacity and network bandwidth values at multiple predetermined time points are arranged according to the time dimension to form a time series. 2. The CPU occupancy rate, memory occupancy rate, disk storage capacity and network bandwidth values at multiple predetermined time points are arranged according to the sample dimension to form a sample sequence. 3. The time series and the sample sequence are combined to form a resource multi-parameter time series matrix.
[0038] In another embodiment of the present application, 1. The CPU occupancy rate, memory occupancy rate, disk storage capacity and network bandwidth values at multiple predetermined time points are arranged according to the time dimension to form a time series. 2. The CPU occupancy rate, memory occupancy rate, disk storage capacity and network bandwidth values at multiple predetermined time points are arranged according to the sample dimension to form a sample sequence. 3. The time series and the sample sequence are respectively combined according to the resource type to form a timing matrix of multiple resource types. 4. The timing matrices of multiple resource types are combined to form a resource multi-parameter timing matrix.
[0039] By arranging the CPU occupancy, memory occupancy, disk storage capacity and network bandwidth values at multiple predetermined time points according to the time dimension and sample dimension into a resource multi-parameter timing matrix, the timing distribution information of various resources in the time dimension can be easily integrated, thereby analyzing the correlation characteristics of the timing changes of various resources in the system and detecting the operating status of the multimedia intelligent scheduling machine.
[0040] Furthermore, in the context resource multi-parameter timing association feature extraction unit 122, it includes: a matrix division sub-unit, used to perform matrix segmentation on the resource multi-parameter timing matrix to obtain multiple resource multi-parameter local timing matrices; a global association feature extraction sub-unit, used to pass the multiple resource multi-parameter local timing matrices through a converter module including an embedding layer to obtain multiple context resource multi-parameter timing association feature vectors.
[0041] Among them, considering that there is time series feature distribution information about various resource parameters in the resource multi-parameter time series matrix, a convolutional neural network model with excellent performance in extracting local implicit correlation features is used to extract the time series correlation features of the various resource parameters. In particular, considering that due to the inherent limitations of convolution operations, the pure CNN method can only capture local implicit correlation features, its receptive field is limited, and it is difficult to learn clear global and long-range semantic information interactions.
[0042] Therefore, in order to capture the global correlation characteristic information of the various resource parameters, the resource multi-parameter timing matrix is further matrix-sliced to obtain multiple resource multi-parameter local timing matrices. It should be understood that the matrix segmentation of the resource multi-parameter timing matrix involved can cut the original data into multiple small local matrices according to certain rules and methods. In this way, the independent and fine-grained features of various resource parameters can be better extracted, avoiding the complexity and excessive noise interference caused by calculation on the overall matrix.
[0043] In this application, matrix segmentation is the process of dividing a large matrix into multiple small local matrices, which can enable large-scale computing tasks to be processed in parallel, thereby improving computing efficiency. Matrix segmentation includes two methods: horizontal segmentation and vertical segmentation.
[0044] Horizontal splitting is to split the rows of a matrix into multiple small local matrices, each of which contains a portion of the rows. For example, a m×n matrix is horizontally split into k local matrices, each of which contains m / k rows. This splitting method is suitable for independent computing tasks between rows.
[0045] Vertical slicing is to split the columns of a matrix into multiple small local matrices, each of which contains a part of the columns. For example, a m×n matrix is vertically split into k local matrices, each of which contains
[0046] n / k columns. This splitting method is suitable for independent computing tasks between columns.
[0047] In addition to horizontal and vertical splitting, it also includes splitting in blocks, that is, splitting the matrix into multiple block matrices of equal size. This splitting method is suitable for tasks that require parallel calculation of the entire matrix.
[0048] Then, the multiple resource multi-parameter local timing matrices are encoded in a converter module including an embedding layer to obtain multiple context resource multi-parameter timing correlation feature vectors. It is worth mentioning that here, the converter module including an embedding layer can extract the local timing correlation features of the various resource parameters based on the global context resource multi-parameter timing correlation feature information to aggregate the local implicit timing correlation features of the various resource parameters, so as to facilitate the subsequent operation status detection of the multimedia intelligent scheduling machine.
[0049] Specifically, the converter module including the embedding layer includes: an embedding layer, which maps each element in each local time series matrix into a vector space to obtain an embedding matrix. An encoder, which takes the embedding matrix as input and processes it through multiple encoding layers to obtain a context feature vector. A decoder, which takes the context feature vector as input and processes it through multiple decoding layers to obtain multiple context resource multi-parameter time series associated feature vectors.
[0050] In the encoder, each encoding layer contains a self-attention layer and a feedforward neural network layer. The self-attention layer is used to calculate the degree of association between each element in the input matrix and other elements, thereby capturing contextual information. In the decoder, each decoding layer also contains a self-attention layer and a feedforward neural network layer. The difference is that the self-attention layer of the decoder not only considers the degree of association between each element in the input matrix and other elements, but also considers the degree of association between the context feature vector output by the encoder and each element in the input matrix, thereby better capturing contextual information.
[0051] Through the converter module including the embedding layer, multiple resource multi-parameter local timing matrices can be encoded into multiple context resource multi-parameter timing correlation feature vectors, so as to better analyze the resource timing change correlation characteristics and detect the operating status of the multimedia intelligent scheduling machine.
[0052] In one embodiment of the present application, first, an embedding layer is defined in the converter module, and each element in each local timing matrix is mapped to a vector space to obtain an embedding matrix. Then, each element in each local timing matrix is mapped to the vector space through the embedding layer to obtain an embedded local timing matrix. Next, the embedded local timing matrix is encoded through an encoder module to obtain a context vector, which contains the context information of the local timing matrix. Finally, multiple context resource multi-parameter timing association feature vectors are obtained: multiple local timing matrices are embedded and encoded in turn to obtain multiple context vectors, which constitute multiple resource multi-parameter timing association feature vectors.
[0053] Furthermore, in the local association enhancement unit 123, it is used to: arrange the multiple context resource multi-parameter temporal association feature vectors into a one-dimensional feature vector and then pass it through a local association feature enhancement module including a first convolution layer and a second convolution layer to obtain an operation status feature vector.
[0054] Then, it is also considered that in the actual operation status detection process of the multimedia intelligent scheduling machine, the various resource parameters, such as CPU occupancy, memory occupancy, disk storage and network bandwidth values, have different correlation characteristics under different time period spans and different resource parameter sample spans, and these different local collaborative correlation feature information play an important role in the operation status detection of the multimedia intelligent scheduling machine. However, the converter module has excellent performance for feature extraction of long-distance dependency associations, and its ability to extract temporal correlation features of medium- and short-distance dependency associations is relatively poor.
[0055] Therefore, in the technical solution of the present application, the multiple context resource multi-parameter temporal correlation feature vectors are further arranged into a one-dimensional feature vector and then passed through a local correlation feature enhancement module including a first convolutional layer and a second convolutional layer to obtain an operation status feature vector. In particular, here, the first convolutional layer and the second convolutional layer use one-dimensional convolution kernels of different scales to extract the multi-scale local temporal collaborative correlation feature information of the various resource parameters at different times and different sample spans. It should be understood that features of multiple levels of abstraction can be gradually extracted through convolution kernels of different sizes and pooling operations, thereby capturing various details and differences in the operation status of the system. In this way, by learning the dependencies in the sample set, local information can be retained and enhanced while extracting global features, thereby obtaining the operation status feature vector with more sufficient feature information.
[0056] The first convolution layer and the second convolution layer respectively include: a convolution layer, which uses one-dimensional convolution kernels of different scales to perform convolution operations to obtain multiple convolution feature matrices of different scales. A pooling layer: performs pooling operations on the convolution feature matrix, usually using maximum pooling or average pooling, to reduce the dimension of the feature matrix and reduce the amount of calculation of the model. A nonlinear activation layer, which performs nonlinear transformation on the pooled feature matrix, usually using the ReLU activation function or its variants, to enhance the nonlinear expression ability of the model.
[0057] It should be noted that the input of the second convolutional layer is usually the output of the first convolutional layer, so the size and number of convolution kernels of the second convolutional layer can be different from those of the first convolutional layer to extract higher-level features. At the same time, the pooling and nonlinear activation operations of the second convolutional layer can also be different from those of the first convolutional layer to further improve the expressiveness of the model.
[0058] In one embodiment of the present application, multiple context resource multi-parameter time series correlation feature vectors are arranged into a one-dimensional feature vector, which includes: 1. Arranging multiple context resource multi-parameter time series correlation feature vectors according to the time dimension to obtain a time series matrix. 2. Arranging the time series matrix according to the sample dimension to obtain a one-dimensional feature vector. 3. Inputting the one-dimensional feature vector into a local correlation feature enhancement module including a first convolutional layer and a second convolutional layer to obtain an operation state feature vector.
[0059] More specifically, for multiple context resource multi-parameter time series correlation feature vectors, assuming that the dimension of each feature vector is d, the time span is T, and the number of samples is N, they are arranged according to the time dimension to obtain a time series matrix with a shape of (T, N, d); the time series matrix is arranged according to the sample dimension to obtain a one-dimensional feature vector with a shape of (N*T, d); the one-dimensional feature vector is input into a local correlation feature enhancement module including a first convolutional layer and a second convolutional layer to obtain an operation state feature vector. Specifically, the one-dimensional feature vector can be convolved, pooled, and nonlinearly activated according to the steps described above to obtain an operation state feature vector with a dimension of k.
[0060] Specifically, the operation status detection module 130 is used to determine whether the operation status of the multimedia intelligent scheduling machine is normal based on the operation status feature vector. Figure 3 A block diagram of the operation status detection module in a multimedia intelligent scheduling machine provided in an embodiment of the present invention, such as Figure 3 As shown, the operation status detection module 130 includes: a feature optimization unit 131, used to optimize the feature distribution of the operation status feature vector to obtain an optimized operation status feature vector; a scheduling machine operation status evaluation unit 132, used to pass the optimized operation status feature vector through a classifier to obtain a classification result, and the classification result is used to indicate whether the operation status of the multimedia intelligent scheduling machine is normal.
[0061] In particular, in the technical solution of the present application, when the multiple resource multi-parameter local timing matrices are passed through a converter module including an embedding layer to obtain multiple context resource multi-parameter timing correlation feature vectors, the multiple context resource multi-parameter timing correlation feature vectors express the context-related multi-parameter sample-time cross-correlation characteristics of each resource multi-parameter local timing matrix, and after the multiple context resource multi-parameter timing correlation feature vectors are arranged into a one-dimensional feature vector and passed through a local correlation feature enhancement module including a first convolutional layer and a second convolutional layer, the local feature distribution can be further enhanced. However, in order to better utilize the context-related sample-time cross-correlation characteristics of the resource parameters, it is preferred to optimize the operating state feature vector by fusing the one-dimensional feature vector and the operating state feature vector.
[0062] Here, considering that the one-dimensional feature vector is obtained by cascading the multiple context resource multi-parameter temporal correlation feature vectors, and the operating status feature vector is further obtained by extracting local correlation features at the local correlation scale based on the one-dimensional feature vector, the one-dimensional feature vector and the operating status feature vector both have serialized distribution properties based on local fragment semantics.
[0063] Therefore, in order to improve the fusion effect of the one-dimensional feature vector and the running state feature vector, the applicant of the present application performs the one-dimensional feature vector, for example, V 1 and the operating state characteristic vector, for example, denoted as V 2 Perform fragment-wise enrichment fusion of local sequence semantics to obtain an optimized running state feature vector, for example, denoted as V 2 ', specifically expressed as: using the following optimization formula to perform fragment-wise enrichment fusion of the local sequence semantics on the one-dimensional feature vector and the running state feature vector to obtain the optimized running state feature vector; wherein the optimization formula is:
[0064]
[0065] Among them, V 1 is the one-dimensional eigenvector, V 2 is the operating state feature vector, D(V 1 ,V 2 ) is the distance matrix between the one-dimensional feature vector and the running state feature vector, V 1 and V 2 are all column vectors, and α is a weight hyperparameter, represents vector multiplication, represents vector addition, V 2 ' is the optimized running state feature vector.
[0066] Here, the fragment-based enrichment of local sequence semantics is based on the encoding effect of the directional semantics of the sequence in the predetermined distribution direction of the sequence based on the segment feature distribution of the sequence, and the similarity between the sequence segments is embedded as a reweighting factor for the association between the sequences, thereby capturing the similarity between the sequences at each segment level based on feature appearance, and realizing the one-dimensional feature vector V 1 and the operating state characteristic vector V 2 The local segment-level semantics of the enriched fusion is used to improve the optimized running state feature vector V 2 'For the one-dimensional feature vector V 1 and the operating state characteristic vector V 2 The fusion effect of the optimized running state feature vector V' 2 In this way, the running status of the multimedia intelligent dispatching machine can be detected in real time and accurately, thereby ensuring the normal operation of the multimedia intelligent dispatching machine and improving the stability and performance of the multimedia system.
[0067] Then, the optimized running state feature vector is classified in a classifier to obtain a classification result, and the classification result is used to indicate whether the running state of the multimedia intelligent scheduling machine is normal. That is, in the technical solution of the present application, the label of the classifier includes the running state of the multimedia intelligent scheduling machine is normal (first label), and the running state of the multimedia intelligent scheduling machine is abnormal (second label), wherein the classifier determines which classification label the running state feature vector belongs to through a soft maximum function.
[0068] It is worth noting that the first label p1 and the second label p2 here do not contain artificially set concepts. In fact, during the training process, the computer model does not have the concept of "whether the operating status of the multimedia intelligent dispatching machine is normal". It only has two classification labels and the probability of the output feature under these two classification labels, that is, the sum of p1 and p2 is one. Therefore, the classification result of whether the operating status of the multimedia intelligent dispatching machine is normal is actually converted from the classification label to a binary classification class probability distribution that conforms to the laws of nature. In essence, the physical meaning of the natural probability distribution of the label is used, rather than the linguistic text meaning of "whether the operating status of the multimedia intelligent dispatching machine is normal".
[0069] It should be understood that in the technical solution of the present application, the classification label of the classifier is a detection evaluation label for whether the operating status of the multimedia intelligent scheduling machine is normal. Therefore, after obtaining the classification result, the operating status of the multimedia intelligent scheduling machine can be detected based on the classification result to ensure the normal operation of the multimedia intelligent scheduling machine.
[0070] A classifier is a machine learning model that maps input data into predefined categories. In this case, the input optimized operating status feature vector is mapped to one of two categories, indicating whether the operating status of the multimedia intelligent scheduling machine is normal.
[0071] The structure of a classifier usually consists of three main components: input layer, hidden layer, and output layer. The input layer receives the optimized running state feature vector as input, the hidden layer is a collection of neurons used to process and extract features from the input data, and the output layer maps the input data to the desired category based on the results of the hidden layer. Common classifiers include support vector machines (SVM), decision trees, random forests, and neural networks. Among them, neural networks are a powerful classifier because they can automatically learn the features in the input data and optimize model parameters during the training process to improve classification performance.
[0072] In one embodiment of the present application, the dispatching machine operation status evaluation unit 132 includes: a fully connected encoding subunit, used to use multiple fully connected layers of the classifier to fully connect the optimized operation status feature vector to obtain an encoded classification feature vector; and a classification subunit, used to pass the encoded classification feature vector through the Softmax classification function of the classifier to obtain the classification result.
[0073] In another embodiment of the present application, the optimized running state feature vector is passed through a classifier to obtain a classification result, and it can also include: data preparation, the optimized running state feature vector and the corresponding label (normal or abnormal) are composed of a training set and a test set. Feature extraction, using the fragment enrichment fusion technology of local sequence semantics to extract features and sequence modeling for each time segment to obtain richer local sequence semantic information, and the local sequence semantic features of all time segments are fused to obtain a complete optimized running state feature vector. Classifier training, using the training set to train the classifier. Common classifiers include support vector machines, decision trees, random forests, and neural networks. During the training process, the classifier updates the model parameters through the back propagation algorithm to minimize the classification error and improve the classification performance. Classifier testing, using the test set to test the classifier to evaluate the performance of the classifier, the optimized running state feature vector is input into the classifier, and the classifier will output the corresponding classification result. Running state judgment, judging whether the running state of the multimedia intelligent scheduling machine is normal according to the classification result. If the classification result is normal, it means that the running state of the multimedia intelligent dispatching machine is normal; if the classification result is abnormal, it means that the running state of the multimedia intelligent dispatching machine is abnormal.
[0074] By passing the optimized operating status feature vector through the classifier to obtain the classification result, the operating status of the multimedia intelligent dispatching machine can be monitored and judged in real time, abnormal situations can be handled in time, and the operating efficiency and stability of the multimedia intelligent dispatching machine can be improved.
[0075] In summary, the multimedia intelligent scheduling machine 100 based on the embodiment of the present invention is explained, which detects the usage of the multimedia intelligent scheduling machine through comprehensive analysis, so as to determine whether the operating status of the multimedia intelligent scheduling machine is normal, thereby ensuring the normal operation of the multimedia intelligent scheduling machine and improving the stability and performance of the multimedia system.
[0076] In one embodiment of the present invention, Figure 4 The present invention provides a flowchart of a multimedia intelligent scheduling method in an embodiment of the present invention. Figure 5 FIG. 1 is a schematic diagram of a system architecture of a multimedia intelligent scheduling method provided in an embodiment of the present invention. Figure 4 and Figure 5As shown, a multimedia intelligent scheduling method includes: 210, obtaining CPU occupancy, memory occupancy, disk storage capacity and network bandwidth values at multiple predetermined time points within a predetermined time period; 220, performing time series correlation analysis on the CPU occupancy, memory occupancy, disk storage capacity and network bandwidth values at the multiple predetermined time points to obtain an operation status feature vector; and, 230, determining whether the operation status of the multimedia intelligent scheduling machine is normal based on the operation status feature vector.
[0077] In a specific example of the present invention, in the above-mentioned multimedia intelligent scheduling method, a timing correlation analysis is performed on the CPU occupancy, memory occupancy, disk storage capacity and network bandwidth values at the multiple predetermined time points to obtain an operation status feature vector, including: arranging the CPU occupancy, memory occupancy, disk storage capacity and network bandwidth values at the multiple predetermined time points into a resource multi-parameter timing matrix according to the time dimension and the sample dimension; performing feature extraction on the resource multi-parameter timing matrix to obtain multiple context resource multi-parameter timing correlation feature vectors; and, performing local correlation feature enhancement on the multiple context resource multi-parameter timing correlation feature vectors to obtain the operation status feature vector.
[0078] Those skilled in the art will appreciate that the specific operations of each step in the above-mentioned multimedia intelligent scheduling method have been described in detail above. Figures 1 to 3 The multimedia intelligent scheduling machine has been introduced in detail, and therefore, its repeated description will be omitted.
[0079] Figure 6 FIG. 1 is an application scenario diagram of a multimedia intelligent dispatching machine provided in an embodiment of the present invention. Figure 6 As shown, in this application scenario, first, a multimedia intelligent scheduling machine (for example, Figure 6 The CPU occupancy rate (e.g., Figure 6 C1), memory usage (e.g. Figure 6 C2), disk storage (e.g. Figure 6 C3) and network bandwidth values (e.g., Figure 6 C4 shown in the figure), in a specific example, the model of the multimedia intelligent scheduling machine is PPHO-236DST-V1; then, the obtained CPU occupancy, memory occupancy, disk storage capacity and network bandwidth values are input to a server (for example, Figure 6In the S) shown in , the server can process the CPU occupancy, the memory occupancy, the disk storage capacity and the network bandwidth value based on the multimedia intelligent scheduling algorithm to determine whether the operating status of the multimedia intelligent scheduling machine is normal.
[0080] The specific embodiments described above further illustrate the objectives, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above description is only a specific embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.
Claims
1. A multimedia intelligent dispatching machine, It is characterized in that include: A data collection module is used to obtain the CPU occupancy rate, memory occupancy rate, disk storage capacity and network bandwidth values at multiple predetermined time points within a predetermined time period; A data processing module, used for performing time series correlation analysis on the CPU occupancy rate, memory occupancy rate, disk storage capacity and network bandwidth values at the plurality of predetermined time points to obtain an operation status feature vector; as well as An operation status detection module, used to determine whether the operation status of the multimedia intelligent dispatching machine is normal based on the operation status feature vector; Wherein, the running status detection module includes: A feature optimization unit, used for performing feature distribution optimization on the running state feature vector to obtain an optimized running state feature vector; A dispatcher operation state evaluation unit, used for passing the optimized operation state feature vector through a classifier to obtain a classification result, wherein the classification result is used to indicate whether the operation state of the multimedia intelligent dispatcher is normal; Wherein, the feature optimization unit is used to: The one-dimensional feature vector and the running state feature vector are subjected to fragment-wise enrichment fusion of local sequence semantics using the following optimization formula to obtain the optimized running state feature vector; Wherein, the optimization formula is: , in, is the one-dimensional eigenvector, is the operating state feature vector, is the distance matrix between the one-dimensional feature vector and the running state feature vector, and are column vectors, and is the weight hyperparameter, represents vector multiplication, represents vector addition, is the optimized operating state feature vector.
2. The multimedia intelligent dispatching machine according to claim 1, It is characterized in that The data processing module comprises: A data time series arrangement unit, used for arranging the CPU occupancy rate, memory occupancy rate, disk storage capacity and network bandwidth values at the plurality of predetermined time points into a resource multi-parameter time series matrix according to the time dimension and the sample dimension; A context resource multi-parameter temporal correlation feature extraction unit, used for extracting features from the resource multi-parameter temporal matrix to obtain a plurality of context resource multi-parameter temporal correlation feature vectors; and The local correlation enhancement unit is used to perform local correlation feature enhancement on the multiple context resource multi-parameter temporal correlation feature vectors to obtain the operating state feature vector.
3. The multimedia intelligent dispatching machine according to claim 2, It is characterized in that The context resource multi-parameter temporal correlation feature extraction unit includes: A matrix division subunit, used for performing matrix division on the resource multi-parameter timing matrix to obtain a plurality of resource multi-parameter local timing matrices; The global correlation feature extraction subunit is used to obtain multiple context resource multi-parameter temporal correlation feature vectors by passing the multiple resource multi-parameter local temporal matrices through a converter module including an embedding layer.
4. The multimedia intelligent dispatching machine according to claim 3, It is characterized in that The local association enhancement unit is used to: arrange the multiple context resource multi-parameter temporal association feature vectors into a one-dimensional feature vector and then pass it through a local association feature enhancement module including a first convolution layer and a second convolution layer to obtain an operation status feature vector.
5. The multimedia intelligent dispatching machine according to claim 4, It is characterized in that The first convolution layer and the second convolution layer of the local correlation feature enhancement module respectively use one-dimensional convolution kernels of different scales.
6. The multimedia intelligent dispatching machine according to claim 5, It is characterized in that The dispatching machine operation status evaluation unit comprises: a fully connected encoding subunit, configured to perform fully connected encoding on the optimized running state feature vector using a plurality of fully connected layers of the classifier to obtain an encoded classification feature vector; and The classification subunit is used to pass the encoded classification feature vector through the Softmax classification function of the classifier to obtain the classification result.
7. A multimedia intelligent scheduling method, It is characterized in that include: Get the CPU usage, memory usage, disk storage and network bandwidth values at multiple scheduled time points within a scheduled time period; Performing time series correlation analysis on the CPU occupancy rate, memory occupancy rate, disk storage capacity and network bandwidth values at the plurality of predetermined time points to obtain an operation status feature vector; as well as Based on the operating status feature vector, determining whether the operating status of the multimedia intelligent dispatching machine is normal; Wherein, based on the running state feature vector, determining whether the running state of the multimedia intelligent dispatching machine is normal includes: Performing feature distribution optimization on the operating state feature vector to obtain an optimized operating state feature vector; Passing the optimized running state feature vector through a classifier to obtain a classification result, wherein the classification result is used to indicate whether the running state of the multimedia intelligent dispatching machine is normal; Among them, performing feature distribution optimization on the running state feature vector to obtain an optimized running state feature vector and performing feature distribution optimization on the running state feature vector to obtain an optimized running state feature vector include: The one-dimensional feature vector and the running state feature vector are subjected to fragment-wise enrichment fusion of local sequence semantics using the following optimization formula to obtain the optimized running state feature vector; Wherein, the optimization formula is: , in, is the one-dimensional eigenvector, is the operating state feature vector, is the distance matrix between the one-dimensional feature vector and the running state feature vector, and are column vectors, and is the weight hyperparameter, represents vector multiplication, represents vector addition, is the optimized operating state feature vector.
8. The multimedia intelligent scheduling method according to claim 7, It is characterized in that Performing time series correlation analysis on the CPU occupancy rate, memory occupancy rate, disk storage capacity, and network bandwidth values at the plurality of predetermined time points to obtain an operation status feature vector includes: Arrange the CPU occupancy rate, memory occupancy rate, disk storage capacity and network bandwidth values at the plurality of predetermined time points into a resource multi-parameter time series matrix according to the time dimension and the sample dimension; Performing feature extraction on the resource multi-parameter timing matrix to obtain a plurality of context resource multi-parameter timing associated feature vectors; and Local correlation feature enhancement is performed on the multiple context resource multi-parameter time series correlation feature vectors to obtain the operating state feature vector.
