CN116760772B - Control system and method for converging flow divider - Google Patents
Control system and method for converging flow divider Download PDFInfo
- Publication number
- CN116760772B CN116760772B CN202311051172.7A CN202311051172A CN116760772B CN 116760772 B CN116760772 B CN 116760772B CN 202311051172 A CN202311051172 A CN 202311051172A CN 116760772 B CN116760772 B CN 116760772B
- Authority
- CN
- China
- Prior art keywords
- computing resource
- time sequence
- feature vector
- association
- global
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 25
- 238000009826 distribution Methods 0.000 claims abstract description 73
- 238000012545 processing Methods 0.000 claims abstract description 18
- 238000012098 association analyses Methods 0.000 claims abstract description 17
- 239000013598 vector Substances 0.000 claims description 165
- 238000005457 optimization Methods 0.000 claims description 21
- 230000004927 fusion Effects 0.000 claims description 18
- 230000008859 change Effects 0.000 claims description 12
- 238000012546 transfer Methods 0.000 claims description 12
- 238000013527 convolutional neural network Methods 0.000 claims description 10
- 238000010606 normalization Methods 0.000 claims description 7
- 230000007704 transition Effects 0.000 claims description 7
- 238000000605 extraction Methods 0.000 claims description 6
- 238000004364 calculation method Methods 0.000 claims description 5
- 238000010219 correlation analysis Methods 0.000 claims description 5
- 238000003062 neural network model Methods 0.000 claims description 5
- 238000013507 mapping Methods 0.000 claims description 4
- 238000013468 resource allocation Methods 0.000 claims description 4
- 238000004422 calculation algorithm Methods 0.000 abstract description 13
- 238000004458 analytical method Methods 0.000 abstract description 8
- 238000007726 management method Methods 0.000 description 23
- 230000006870 function Effects 0.000 description 13
- 238000010586 diagram Methods 0.000 description 8
- 238000011176 pooling Methods 0.000 description 7
- 230000004913 activation Effects 0.000 description 6
- 239000011159 matrix material Substances 0.000 description 6
- 230000008569 process Effects 0.000 description 5
- 238000013528 artificial neural network Methods 0.000 description 4
- 238000010801 machine learning Methods 0.000 description 4
- 210000002569 neuron Anatomy 0.000 description 4
- 230000003068 static effect Effects 0.000 description 4
- 230000002776 aggregation Effects 0.000 description 3
- 238000004220 aggregation Methods 0.000 description 3
- 238000004140 cleaning Methods 0.000 description 3
- 238000003860 storage Methods 0.000 description 3
- 230000006978 adaptation Effects 0.000 description 2
- 238000002790 cross-validation Methods 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 239000004973 liquid crystal related substance Substances 0.000 description 2
- 238000012423 maintenance Methods 0.000 description 2
- 238000012544 monitoring process Methods 0.000 description 2
- 238000007781 pre-processing Methods 0.000 description 2
- 239000002699 waste material Substances 0.000 description 2
- 230000008901 benefit Effects 0.000 description 1
- 230000001149 cognitive effect Effects 0.000 description 1
- 238000013079 data visualisation Methods 0.000 description 1
- 238000003066 decision tree Methods 0.000 description 1
- 230000001419 dependent effect Effects 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 230000002068 genetic effect Effects 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 230000003993 interaction Effects 0.000 description 1
- 238000005065 mining Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000003058 natural language processing Methods 0.000 description 1
- 239000002245 particle Substances 0.000 description 1
- 238000013439 planning Methods 0.000 description 1
- 238000002922 simulated annealing Methods 0.000 description 1
- 238000007619 statistical method Methods 0.000 description 1
- 238000013179 statistical model Methods 0.000 description 1
- 238000012549 training Methods 0.000 description 1
- 238000010200 validation analysis Methods 0.000 description 1
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L47/00—Traffic control in data switching networks
- H04L47/10—Flow control; Congestion control
- H04L47/12—Avoiding congestion; Recovering from congestion
- H04L47/125—Avoiding congestion; Recovering from congestion by balancing the load, e.g. traffic engineering
Abstract
The application discloses a control system and a control method of a converging flow divider, which automatically carry out integral time sequence association analysis on the residual computing resource quantity of global target equipment by utilizing a data processing and analysis algorithm so as to determine the distribution proportion of flow. Therefore, the occurrence of low efficiency, configuration errors and unreasonable conditions caused by manual intervention can be avoided, so that the whole resources of target equipment in the system can be fully utilized to realize load balancing, and the performance and the scalability of the system are improved.
Description
Technical Field
The application relates to the field of intelligent control, in particular to a control system and a control method of a converging shunt.
Background
A convergence splitter is a network device that is used to combine multiple network connections or data flows into one or more flows, or to distribute one flow to multiple targets, typically used in network load balancing, traffic management, and data center networks. The control system of the convergence diverter refers to a software system for managing and controlling the convergence diverter, and the system needs to monitor the computing resource condition of target equipment in real time and adjust the distribution proportion of the flow so as to realize load balancing and optimal resource utilization.
However, conventional aggregate shunt management systems typically employ a static configuration to determine the distribution ratio of the traffic, which means that the weight or distribution ratio of each target device needs to be manually specified at system start-up or at configuration change, and such a static configuration cannot adapt to dynamically changing environments and load conditions. Under the condition of high load or sudden load, the traditional management and control system may not be capable of timely adjusting the distribution proportion of the flow, so that the resource utilization is unbalanced, the performance is reduced, and the real load balance cannot be realized.
Moreover, the management and control system of the traditional convergence diverter needs manual intervention to perform configuration and adjustment, which not only increases the workload of management and maintenance, but also is easy to cause configuration errors or unreasonable conditions. At the same time, manual intervention also limits the automation and adaptation capabilities of the system. In addition, the conventional management and control system usually only focuses on the resource status of a single target device, and lacks a global view of the whole system, so that the resources of other target devices in the system cannot be fully utilized, and resource waste and performance bottleneck are caused.
Accordingly, an optimized converging splitter management system is desired.
Disclosure of Invention
The present application has been made to solve the above-mentioned technical problems. The embodiment of the application provides a control system and a control method of a converging splitter, which automatically carry out integral time sequence association analysis on the residual computing resource quantity of global target equipment by utilizing a data processing and analysis algorithm so as to determine the distribution proportion of flow. Therefore, the occurrence of low efficiency, configuration errors and unreasonable conditions caused by manual intervention can be avoided, so that the whole resources of target equipment in the system can be fully utilized to realize load balancing, and the performance and the scalability of the system are improved.
According to one aspect of the present application, there is provided a control system for a converging flow diverter, comprising:
the data acquisition module is used for acquiring the residual computing resource amounts of a plurality of target devices at a plurality of preset time points in a preset time period;
the data association analysis module is used for carrying out association analysis on the residual computing resource amounts of a plurality of target devices at a plurality of preset time points in a preset time period so as to obtain the time sequence association characteristics of the residual computing resources of the global target device;
and the traffic distribution module is used for determining traffic distribution proportion based on the time sequence related characteristics of the residual computing resources of the global target equipment.
According to another aspect of the present application, there is provided a method for controlling a converging splitter, including:
acquiring the residual computing resource amounts of a plurality of target devices at a plurality of preset time points in a preset time period;
performing association analysis on the residual computing resource amounts of a plurality of preset time points of the plurality of target devices in a preset time period to obtain a global target device residual computing resource time sequence association characteristic;
and determining a flow distribution proportion based on the time sequence related characteristics of the residual computing resources of the global target equipment.
Compared with the prior art, the control system and the control method for the converging current divider provided by the application automatically carry out integral time sequence association analysis on the residual computing resource quantity of the global target equipment by utilizing a data processing and analysis algorithm so as to determine the distribution proportion of the flow. Therefore, the occurrence of low efficiency, configuration errors and unreasonable conditions caused by manual intervention can be avoided, so that the whole resources of target equipment in the system can be fully utilized to realize load balancing, and the performance and the scalability of the system are improved.
Drawings
The above and other objects, features and advantages of the present application will become more apparent by describing embodiments of the present application in more detail with reference to the attached drawings. The accompanying drawings are included to provide a further understanding of embodiments of the application and are incorporated in and constitute a part of this specification, illustrate the application and together with the embodiments of the application, and not constitute a limitation to the application. In the drawings, like reference numerals generally refer to like parts or steps.
FIG. 1 is a block diagram of a system for managing a converging splitter according to an embodiment of the application.
Fig. 2 is a system architecture diagram of a control system for a converging splitter according to an embodiment of the application.
Fig. 3 is a block diagram of a data association analysis module in a management and control system of a convergence splitter according to an embodiment of the application.
Fig. 4 is a block diagram of a flow distribution module in a management and control system for a converging splitter according to an embodiment of the application.
Fig. 5 is a block diagram of a feature distribution optimizing unit in a management and control system of a converging splitter according to an embodiment of the present application.
Fig. 6 is a flow chart of a method of controlling a converging splitter according to an embodiment of the application.
Detailed Description
Hereinafter, exemplary embodiments according to the present application will be described in detail with reference to the accompanying drawings. It should be apparent that the described embodiments are only some embodiments of the present application and not all embodiments of the present application, and it should be understood that the present application is not limited by the example embodiments described herein.
As used in the specification and in the claims, the terms "a," "an," "the," and/or "the" are not specific to a singular, but may include a plurality, unless the context clearly dictates otherwise. In general, the terms "comprises" and "comprising" merely indicate that the steps and elements are explicitly identified, and they do not constitute an exclusive list, as other steps or elements may be included in a method or apparatus.
Although the present application makes various references to certain modules in a system according to embodiments of the present application, any number of different modules may be used and run on a user terminal and/or server. The modules are merely illustrative, and different aspects of the systems and methods may use different modules.
A flowchart is used in the present application to describe the operations performed by a system according to embodiments of the present application. It should be understood that the preceding or following operations are not necessarily performed in order precisely. Rather, the various steps may be processed in reverse order or simultaneously, as desired. Also, other operations may be added to or removed from these processes.
Hereinafter, exemplary embodiments according to the present application will be described in detail with reference to the accompanying drawings. It should be apparent that the described embodiments are only some embodiments of the present application and not all embodiments of the present application, and it should be understood that the present application is not limited by the example embodiments described herein.
Conventional converging splitter control systems typically employ a static configuration to determine the distribution ratio of the traffic, which means that the weight or distribution ratio of each target device needs to be manually specified at system start-up or at configuration change, and this static configuration cannot adapt to dynamically changing environments and load conditions. Under the condition of high load or sudden load, the traditional management and control system may not be capable of timely adjusting the distribution proportion of the flow, so that the resource utilization is unbalanced, the performance is reduced, and the real load balance cannot be realized. Moreover, the management and control system of the traditional convergence diverter needs manual intervention to perform configuration and adjustment, which not only increases the workload of management and maintenance, but also is easy to cause configuration errors or unreasonable conditions. At the same time, manual intervention also limits the automation and adaptation capabilities of the system. In addition, the conventional management and control system usually only focuses on the resource status of a single target device, and lacks a global view of the whole system, so that the resources of other target devices in the system cannot be fully utilized, and resource waste and performance bottleneck are caused. Accordingly, an optimized converging splitter management system is desired.
In the technical scheme of the application, a control system of a converging shunt is provided. FIG. 1 is a block diagram of a system for managing a converging splitter according to an embodiment of the application. Fig. 2 is a system architecture diagram of a control system for a converging splitter according to an embodiment of the application. As shown in fig. 1 and 2, a converging splitter management and control system 300 according to an embodiment of the present application includes: a data acquisition module 310, configured to acquire remaining amounts of computing resources of a plurality of target devices at a plurality of predetermined time points within a predetermined time period; a data association analysis module 320, configured to perform association analysis on the amounts of remaining computing resources of the plurality of target devices at a plurality of predetermined time points within a predetermined period of time, so as to obtain a global target device remaining computing resource timing association feature; and the traffic distribution module 330 is configured to determine a traffic distribution proportion based on the remaining computing resource time sequence correlation characteristics of the global target device.
In particular, the data acquisition module 310 is configured to acquire the amounts of remaining computing resources of a plurality of target devices at a plurality of predetermined time points within a predetermined time period. Notably, the amount of remaining computing resources can be understood as a measure of the computing power or computing resources currently available to the system or device. This generally refers to the amount of computing resources that may also be used to perform tasks or process workloads in a particular computing environment. The computing resources may include the number of cores of the processor, memory capacity, storage space, network bandwidth, and the like. The amount of remaining computing resources represents the amount or availability of computing resources that are not currently being used or allocated. This index may help evaluate the load situation of the system, the resource utilization, and whether there are sufficient resources to handle the new task or workload.
Accordingly, in one possible implementation, the remaining amount of computing resources for a plurality of target devices at a plurality of predetermined points in time within a predetermined period of time may be obtained by: a list of target devices to monitor is determined. This may be a server, virtual machine, container, or other computing resource; the period of time to be monitored is determined, for example, one day, one week or one month. Then determining a predetermined point in time to be monitored during the time period, such as a specific point in time every hour, day or week; each target device is connected over the network using a suitable management tool or protocol. This may involve the use of SSH, remote desktop or dedicated management tools, etc.; upon connection to the target device, computing resource information is obtained, such as processor core number, memory capacity, storage space, network bandwidth, and the like. Such information may be obtained through command line tools, API interfaces, or specific management tools provided by the operating system; at each predetermined point in time, the current computing resource usage is recorded. This may include the number of processor cores used, memory used, storage used, and network bandwidth usage; the remaining amount of computing resources is calculated by subtracting the amount of computing resources that have been used from the total amount of computing resources. This will provide the amount of remaining computing resources per predetermined point in time; repeating steps 5 and 6 at each predetermined point in time until the monitoring of the entire time period is completed; the amount of remaining computing resources at each predetermined point in time is recorded and analyzed. This can help you know the usage pattern of the computing resources, peak periods, and whether there is a bottleneck or insufficient situation in the resources.
In particular, the data association analysis module 320 is configured to perform association analysis on the amounts of remaining computing resources of the plurality of target devices at a plurality of predetermined time points within a predetermined period of time to obtain a global target device remaining computing resource timing association characteristic. In particular, in one specific example of the present application, as shown in fig. 3, the data association analysis module 320 includes: a remaining computing resource timing distribution unit 321, configured to arrange remaining computing resource amounts of a plurality of predetermined time points of each target device in a predetermined time period as input vectors according to a time dimension, so as to obtain remaining computing resource timing input vectors of a plurality of target devices; a residual computing resource time sequence variation feature extraction unit 322, configured to perform feature extraction on the residual computing resource time sequence input vectors of the plurality of target devices by using a time sequence feature extractor based on a deep neural network model, so as to obtain residual computing resource time sequence feature vectors of the plurality of target devices; the target device global dimension computing resource association encoding unit 323 is configured to perform global association encoding on the plurality of target device remaining computing resource timing feature vectors to obtain a target device global dimension computing resource timing association feature vector as the global target device remaining computing resource timing association feature.
Specifically, the remaining computing resource timing distribution unit 321 is configured to arrange the remaining computing resource amounts of the target devices at a plurality of predetermined time points in a predetermined time period as input vectors according to a time dimension, so as to obtain remaining computing resource timing input vectors of the plurality of target devices. In order to effectively analyze and characterize the time sequence dynamic change condition and trend of the residual computing resource quantity of each target device so as to improve the accuracy of the flow distribution proportion, in the technical scheme of the application, the residual computing resource quantity of each target device at a plurality of preset time points in a preset time period is required to be further arranged as input vectors according to the time dimension so as to obtain the time sequence input vectors of the residual computing resource of the plurality of target devices, so that the time sequence distribution information of the residual computing resource quantity of each target device at the plurality of preset time points in the time dimension is respectively integrated.
Specifically, the residual computing resource time sequence variation feature extraction unit 322 is configured to perform feature extraction on the residual computing resource time sequence input vectors of the plurality of target devices by using a time sequence feature extractor based on a deep neural network model, so as to obtain residual computing resource time sequence feature vectors of the plurality of target devices. In other words, in the technical scheme of the present application, the time sequence input vectors of the residual computing resources of the plurality of target devices are respectively subjected to feature mining in a time sequence feature extractor based on a one-dimensional convolutional neural network model, so as to extract time sequence associated feature information of the residual computing resource amounts of a plurality of predetermined time points of each target device in a time dimension, namely, time sequence change feature information of the residual computing resource amounts of each target device, thereby obtaining time sequence feature vectors of the residual computing resources of the plurality of target devices.
According to an embodiment of the present application, passing the plurality of target device remaining computing resource timing input vectors through a one-dimensional convolutional neural network model-based timing feature extractor to obtain the plurality of target device remaining computing resource timing feature vectors, respectively, includes: each layer of the time sequence feature extractor based on the one-dimensional convolutional neural network model is used for respectively carrying out input data in forward transfer of the layer: carrying out convolution processing on input data to obtain a convolution characteristic diagram; pooling the convolution feature images based on a feature matrix to obtain pooled feature images; performing nonlinear activation on the pooled feature map to obtain an activated feature map; the output of the last layer of the time sequence feature extractor based on the one-dimensional convolutional neural network model is the time sequence feature vector of the residual computing resources of the plurality of target devices, and the input of the first layer of the time sequence feature extractor based on the one-dimensional convolutional neural network model is the time sequence input vector of the residual computing resources of the plurality of target devices.
Notably, the one-dimensional convolutional neural network (1D CNN) is a neural network model for processing one-dimensional sequence data. Unlike conventional fully-connected neural networks, 1D CNNs use convolution operations on input data to capture local patterns and features in the input data. Notably, the structure of the 1D CNN includes: input layer: receiving one-dimensional sequence data as input; convolution layer: the input data is feature extracted using a convolution operation. The convolution layer includes a plurality of convolution kernels (or filters), each of which may learn a different characteristic. A convolution operation generates a feature map by sliding a convolution kernel over an input sequence and performing element product sum-up; activation function: after the convolutional layer, an activation function is typically applied to introduce nonlinearities. Common activation functions include ReLU, sigmoid, tanh, etc.; pooling layer: the pooling layer is used to reduce the dimensionality of the feature map and extract the most important features. Common pooling operations include maximum pooling and average pooling; full tie layer: after passing through a plurality of convolution layers and pooling layers, a full connection layer can be added to carry out tasks such as classification or regression; output layer: the output layer may use different activation functions, such as Sigmoid, softmax, etc., depending on the task.
Specifically, the target device global dimension computing resource association encoding unit 323 is configured to perform global association encoding on the plurality of target device remaining computing resource timing feature vectors to obtain a target device global dimension computing resource timing association feature vector as the global target device remaining computing resource timing association feature. In an actual distributed system, a certain association relationship exists between resource utilization and load conditions of different target devices. That is, the remaining computing resource amount time sequence variation characteristics of each target device have a global association relationship. Therefore, in the technical scheme of the application, the time sequence feature vectors of the residual computing resources of the plurality of target devices are further encoded in a computing resource global semantic encoder based on a converter module, so that the time sequence change feature information based on the device global among the time sequence change features of the residual computing resources of each target device is extracted, and the time sequence association feature vectors of the computing resources of the target device global dimension are obtained. In particular, the target device global dimension computing resource timing association feature vector may reflect resource associations and global features between different target devices, such as a fluctuation pattern of resources, trends in resource utilization, and so on. Based on the global dimension calculation resource time sequence association feature vector of the target equipment, the system can more comprehensively analyze and understand the resource relation among the target equipment, so that the load balancing decision and the determination of the flow distribution proportion can be more accurately carried out.
According to an embodiment of the present application, passing the plurality of target device remaining computing resource timing feature vectors through a computing resource global semantic encoder based on a converter module to obtain the target device global dimension computing resource timing association feature vector includes: one-dimensional arrangement is carried out on the residual computing resource time sequence feature vectors of the plurality of target devices so as to obtain global target device residual computing resource time sequence feature vectors; calculating the product between the global target equipment residual computing resource time sequence feature vector and the transpose vector of each target equipment residual computing resource time sequence feature vector in the target equipment residual computing resource time sequence feature vectors to obtain a plurality of self-attention correlation matrixes; respectively carrying out standardization processing on each self-attention correlation matrix in the plurality of self-attention correlation matrices to obtain a plurality of standardized self-attention correlation matrices; obtaining a plurality of probability values by using a Softmax classification function through each normalized self-attention correlation matrix in the normalized self-attention correlation matrices; weighting each target equipment residual computing resource time sequence feature vector in the plurality of target equipment residual computing resource time sequence feature vectors by taking each probability value in the plurality of probability values as a weight so as to obtain the plurality of context semantic target equipment residual computing resource time sequence feature vectors; and cascading the residual computing resource time sequence feature vectors of the context semantic target device to obtain a global dimension computing resource time sequence associated feature vector of the target device.
Notably, the computing resource global semantic encoder is a technique for converting context information into a fixed length vector representation. In natural language processing and machine learning, the computing resource global semantic encoder is often used to process sequence data, such as text, speech, or time sequence data. The goal of a computing resource global semantic encoder is to convert a variable-length input sequence into a fixed-length vector representation to capture the semantic and contextual information of the input sequence. This vector representation may be used as an abstract representation of the input sequence for subsequent tasks such as classification, generation or clustering.
It should be noted that, in other specific examples of the present application, the remaining computing resource amounts of the plurality of target devices at a plurality of predetermined time points within the predetermined time period may also be subjected to correlation analysis by other manners to obtain a global target device remaining computing resource timing correlation feature, for example: and in a preset time period, acquiring residual computing resource amount data of each target device at a plurality of preset time points according to the methods from step 1 to step 7. Ensuring that the recorded data includes device identification, time stamp and amount of remaining computing resources; preprocessing the collected data, including data cleaning, outlier removal and missing value processing. The consistency and the accuracy of the data format are ensured; the remaining computing resource amount data for each target device is integrated into one data set for correlation analysis. The merging of the data can be performed according to the time stamp and the equipment identification; the associated features to be analyzed are defined according to the requirements and the targets. For example, an average, maximum, minimum, standard deviation, etc. of the global remaining computing resource amounts may be calculated; the integrated dataset is analyzed in time series, exploring trends, seasonality and periodicity of the data. Analysis may be performed using statistical methods, time series models, or machine learning algorithms; the relevance of the amount of remaining computing resources between the different target devices is explored using suitable relevance analysis methods, such as correlation coefficients, covariance, correlation graphs, etc. A correlation matrix can be calculated or a correlation chart can be drawn to visualize the association relationship; and extracting time sequence correlation characteristics of residual computing resources of the global target equipment according to the result of the correlation analysis. This may include features such as synchronization, hysteresis, periodicity, or abrupt changes between the target devices; the extracted features are interpreted to understand the time sequence association relationship of the residual computing resources among different target devices. Based on these associated features, corresponding decisions may be made, such as resource scheduling, load balancing, or capacity planning.
In particular, the traffic distribution module 330 is configured to determine a traffic distribution ratio based on the remaining computing resource timing correlation characteristics of the global target device. In particular, in one specific example of the present application, as shown in fig. 4, the flow distribution module 330 includes: the feature distribution optimizing unit 331 is configured to perform feature distribution optimization on the target device global dimension computing resource time sequence associated feature vector to obtain an optimized target device global dimension computing resource time sequence associated feature vector; a target device remaining computing resource mapping association unit 332, configured to respectively calculate transfer matrices of the remaining computing resource timing feature vectors of each target device with respect to the global dimension computing resource timing association feature vector of the optimized target device to obtain a plurality of transfer matrices; a remaining resource allocation probability calculation unit 333 configured to pass the plurality of transition matrices through a classifier to obtain a plurality of probability values and perform maximum value-based normalization processing on the plurality of probability values; and a flow distribution ratio determining unit 334 configured to set the plurality of probability values as flow distribution ratios.
Specifically, the feature distribution optimizing unit 331 is configured to perform feature distribution optimization on the target device global dimension computing resource time sequence associated feature vector to obtain an optimized target device global dimension computing resource time sequence associated feature vector. In particular, in one specific example of the present application, as shown in fig. 5, the feature distribution optimizing unit 331 includes: a vector concatenation subunit 3311, configured to concatenate the residual computing resource timing feature vectors of the plurality of target devices to obtain a concatenated feature vector; and a tracking equalization fusion subunit 3312, configured to perform hilbert space heuristic sequence tracking equalization fusion on the cascade feature vector and the target device global dimension computing resource time sequence correlation feature vector to obtain the optimized target device global dimension computing resource time sequence correlation feature vector.
More specifically, the vector concatenation subunit 3311 is configured to concatenate the residual computing resource timing feature vectors of the plurality of target devices to obtain a concatenated feature vector. That is, specifically, the residual computing resource time sequence feature vectors of the plurality of target devices are cascaded in the following cascade formula to obtain cascade feature vectors; wherein, the formula is:wherein->Time sequence feature vector representing the remaining computing resources of the plurality of target devices,>representing a cascade function->Representing the concatenated feature vector.
More specifically, the tracking equalization fusion subunit 3312 is configured to perform hilbert space heuristic sequence tracking equalization fusion on the cascade feature vector and the target device global dimension computing resource timing correlation feature vector to obtain the optimized target device global dimension computing resource timing correlation feature vector. In particular, in the technical solution of the present application, when the plurality of target device remaining computing resource timing feature vectors pass through a computing resource global semantic encoder based on a converter module to obtain target device global dimension computing resource timing association feature vectors, the target device global dimension computing resource timing association feature vectors may express inter-device context association features of each target device remaining computing resource timing feature vector, but, considering a distribution difference of a remaining computing resource amount of each target device in a time domain, the target device global dimension computing resource timing association feature vectors obtained through context association encoding may have a distribution imbalance with respect to the plurality of target device remaining computing resource timing feature vectors, so as to affect a computing accuracy of a transition matrix of each target device remaining computing resource timing feature vector with respect to the target device global dimension computing resource timing association feature vectors, and then affect an accuracy of a plurality of probability values obtained by the classifier by the plurality of transition matrices. Thus, the target device global dimension computing resource timing correlation feature vector also conforms to the plurality of target device remaining computing resource timings, taking into account that the target device global dimension computing resource timing correlation feature vector is substantially obtained by concatenating the plurality of contextual target device remaining computing resource timing feature vectors obtained by the converter module-based computing resource global semantic encoder The applicant of the present application refers to a cascade feature vector of the residual computing resource time sequence feature vectors of the target devices, for example, asAnd said target device global dimension computing resource timing association feature vector, e.g. denoted +.>Performing Hilbert space heuristic sequence tracking equalization fusion to optimize the target device global dimension computing resource time sequence associated feature vector, for example, marked as +.>The method is specifically expressed as follows:
wherein, the liquid crystal display device comprises a liquid crystal display device,is the cascade feature vector,/->Is the target device global dimension computing resource time sequence associated feature vector,>representing feature vector +.>And->Is +.>Representing feature vector +.>And->Mean value of union set composed of all eigenvalues of (2), and eigenvector +.>And->Are all row vectors, +.>Representing multiplication by location +.>Representing vector addition, ++>Is the global dimension computing resource time sequence associated feature vector of the optimizing target equipment. Here, the complete inner product space characteristic of the hilbert space with inner product is utilized to pass the cascade feature vector +.>And said target device global dimension computing resource timing association feature vector +. >Is aggregated mean of sequence aggregation (collective average), exploring the cascade feature vector +.>And said target device global dimension computing resource timing association feature vector +.>Sequence-based spatial distribution heuristics (heuristics) within feature fusion space via context-dependent encoding to globally dimension the target deviceDegree computing resource timing associative feature vector>The local feature distribution of the sequence is converted into a sequence tracking instance (tracking instance) in a fusion space so as to realize tracking small-segment cognitive (tracking let-aware) distribution equalization of the feature space distribution of the sequence, and thus, the distribution imbalance of the target equipment global dimension computing resource time sequence correlation feature vector relative to the residual computing resource time sequence feature vectors of the plurality of target equipment is improved. Therefore, the distribution proportion of the actual flow can be determined based on the change condition of the residual computing resources of the global target equipment, so that the whole resources of the target equipment in the system are fully utilized to realize load balancing, and the performance and the scalability of the system are improved.
It should be noted that, in other specific examples of the present application, the feature distribution optimization may be performed on the target device global dimension computing resource time sequence associated feature vector in other manners to obtain an optimized target device global dimension computing resource time sequence associated feature vector, for example: and collecting the original data of the target equipment global dimension computing resource time sequence association characteristic vector. The data may include information about computing resource usage, task load, network traffic, etc. of the target device; the collected data is subjected to feature analysis, including statistic calculation, data visualization, correlation analysis and the like. This can help to understand the relationship, importance and distribution between features; based on the results of the feature analysis, features related to target device global dimension computing resource time sequence associated feature vector optimization are selected. Features related to computing resource utilization, load balancing, latency, etc. may be considered; the selected features are preprocessed, including data cleaning, missing value processing, outlier processing, and the like. The accuracy and the integrity of data are ensured; the selected features are normalized and converted to values having the same scale and range. Normalization can eliminate dimensional differences between different features so that they are comparable in the optimization process; and optimizing the target equipment global dimension computing resource time sequence association feature vector by using an optimization algorithm. Common optimization algorithms include genetic algorithms, particle swarm algorithms, simulated annealing algorithms, and the like. The goals of optimization may be to maximize computing resource utilization, minimize load imbalance, etc.; and evaluating the performance of the optimized target equipment global dimension computing resource time sequence association feature vector, and adjusting according to the requirement. Cross-validation, index evaluation, etc. may be used to evaluate the optimization effect of the feature vector; and applying the optimized target equipment global dimension computing resource time sequence association feature vector to actual flow distribution or resource scheduling, and monitoring the effect of the feature vector. According to the actual situation, iterative optimization can be performed, and the performance and adaptability of the feature vector are further improved.
Specifically, the target device remaining computing resource mapping association unit 332 is configured to calculate transfer matrices of the target device remaining computing resource timing feature vectors relative to the optimization target device global dimension computing resource timing association feature vector, so as to obtain a plurality of transfer matrices. That is, the transfer matrices of the time sequence feature vectors of the residual computing resources of each target device relative to the time sequence associated feature vectors of the computing resources of the global dimension of the target device are calculated respectively, so that the time sequence change features of the residual computing resources of each target device are mapped into the high-dimensional space of the time dimension and the global associated features of the computing resources of each target device in the sample dimension, and a plurality of transfer matrices are obtained to represent the change trend and the fluctuation features of the residual computing resources of each target device on the basis of the overall system.
Specifically, the calculated remaining resource allocation probability calculation unit 333 is configured to pass the plurality of transition matrices through a classifier to obtain a plurality of probability values and perform maximum value-based normalization processing on the plurality of probability values. That is, classification processing is performed based on the overall time-series variation characteristic information using the remaining computing resources of each of the target devices, so that the traffic distribution ratio for each of the target devices is determined.
According to an embodiment of the present application, passing the plurality of transition matrices through a classifier to obtain a plurality of probability values includes: expanding the transfer matrix into classification feature vectors based on row vectors or column vectors; performing full-connection coding on the classification feature vectors by using a plurality of full-connection layers of the classifier to obtain coded classification feature vectors; and passing the coding classification feature vector through a Softmax classification function of the classifier to obtain the classification result.
A Classifier (Classifier) refers to a machine learning model or algorithm that is used to classify input data into different categories or labels. The classifier is part of supervised learning, which performs classification tasks by learning mappings from input data to output categories.
The fully connected layer (Fully Connected Layer) is one type of layer commonly found in neural networks. In the fully connected layer, each neuron is connected to all neurons of the upper layer, and each connection has a weight. This means that each neuron in the fully connected layer receives inputs from all neurons in the upper layer, and weights these inputs together, and then passes the result to the next layer.
The Softmax classification function is a commonly used activation function for multi-classification problems. It converts each element of the input vector into a probability value between 0 and 1, and the sum of these probability values equals 1. The Softmax function is commonly used at the output layer of a neural network, and is particularly suited for multi-classification problems, because it can map the network output into probability distributions for individual classes. During the training process, the output of the Softmax function may be used to calculate the loss function and update the network parameters through a back propagation algorithm. Notably, the output of the Softmax function does not change the relative magnitude relationship between elements, but rather normalizes them. Thus, the Softmax function does not change the characteristics of the input vector, but simply converts it into a probability distribution form.
Specifically, the flow distribution ratio determining unit 334 is configured to set the plurality of probability values as the flow distribution ratio. That is, in order to facilitate the generation of the flow distribution ratio after the plurality of probability values are obtained, in the technical solution of the present application, the plurality of probability values are further subjected to normalization processing based on a maximum value, and the plurality of probability values after normalization processing are set as the flow distribution ratio. Thus, the occurrence of low efficiency, configuration errors and unreasonable conditions caused by manual intervention can be avoided, and the whole resource of target equipment in the system can be fully utilized to realize load balancing.
It should be noted that, in other specific examples of the present application, the traffic distribution proportion may also be determined by other manners based on the remaining computing resource timing related features of the global target device, for example: acquiring time sequence associated characteristic data of residual computing resources of the global target equipment, wherein the time sequence associated characteristic data comprise residual computing resource quantity, associated characteristic values (such as average value, maximum value, minimum value and the like) and other relevant information of each time point; preprocessing the collected data, including data cleaning, outlier removal and missing value processing. The consistency and the accuracy of the data format are ensured; and selecting the associated characteristics related to the flow distribution proportion according to the task requirements and the targets. These features may be features directly related to the amount of computing resources, or may be features related to computing resource utilization, load balancing, etc.; the selected associated features are data normalized and converted to values having the same scale and range. This may ensure that the different characteristics have a comparability in the impact on the traffic distribution ratio; based on the selected correlation characteristics and the target variables (flow distribution ratio), an appropriate machine learning or statistical model is established. Modeling can be performed by using regression models, decision trees, neural networks and other methods; the built model is trained using the labeled dataset and performance of the model is assessed using the validation set. Techniques such as cross-validation can be used to evaluate the generalization ability of the model; predicting the time sequence associated characteristics of the residual computing resources of the new global target equipment by using the trained model to obtain corresponding flow distribution proportion; the predicted traffic distribution ratio is interpreted to understand the impact of different correlation characteristics on the distribution ratio. According to the prediction result, corresponding flow adjustment or resource allocation decision can be made to realize better load balancing and resource utilization.
As described above, the convergence splitter management and control system 300 according to the embodiment of the present application may be implemented in various wireless terminals, for example, a server or the like having a convergence splitter management and control algorithm. In one possible implementation, the aggregation splitter management system 300 according to an embodiment of the present application may be integrated into a wireless terminal as a software module and/or a hardware module. For example, the management and control system 300 of the convergence splitter may be a software module in the operating system of the wireless terminal or may be an application developed for the wireless terminal; of course, the centralized control system 300 of the aggregation splitter may also be one of a plurality of hardware modules of the wireless terminal.
Alternatively, in another example, the centralized control system 300 of the centralized shunt and the wireless terminal may be separate devices, and the centralized control system 300 of the centralized shunt may be connected to the wireless terminal through a wired and/or wireless network and transmit the interaction information according to the agreed data format.
Further, a control method of the converging flow divider is also provided.
Fig. 6 is a flow chart of a method of controlling a converging splitter according to an embodiment of the application. As shown in fig. 6, a method for controlling a converging splitter according to an embodiment of the present application includes the steps of: s1, acquiring residual computing resource amounts of a plurality of target devices at a plurality of preset time points in a preset time period; s2, performing association analysis on the residual computing resource amounts of a plurality of target devices at a plurality of preset time points in a preset time period to obtain a global target device residual computing resource time sequence association characteristic; s3, determining a flow distribution proportion based on the time sequence related characteristics of the residual computing resources of the global target equipment.
In summary, a method for controlling a converging splitter according to an embodiment of the present application is explained, which automatically performs overall time-series association analysis on the remaining amount of computing resources of a global target device by using a data processing and analysis algorithm to determine a distribution proportion of traffic. Therefore, the occurrence of low efficiency, configuration errors and unreasonable conditions caused by manual intervention can be avoided, so that the whole resources of target equipment in the system can be fully utilized to realize load balancing, and the performance and the scalability of the system are improved.
The foregoing description of the embodiments of the present disclosure has been presented for purposes of illustration and description, and is not intended to be exhaustive or limited to the embodiments disclosed. Many modifications and variations will be apparent to those of ordinary skill in the art without departing from the scope and spirit of the various embodiments described. The terminology used herein was chosen in order to best explain the principles of the embodiments, the practical application, or the improvement of technology in the marketplace, or to enable others of ordinary skill in the art to understand the embodiments disclosed herein.
Claims (5)
1. A control system for a converging splitter, comprising:
The data acquisition module is used for acquiring the residual computing resource amounts of a plurality of target devices at a plurality of preset time points in a preset time period;
the data association analysis module is used for carrying out association analysis on the residual computing resource amounts of a plurality of target devices at a plurality of preset time points in a preset time period so as to obtain the time sequence association characteristics of the residual computing resources of the global target device;
the traffic distribution module is used for determining traffic distribution proportion based on the time sequence associated characteristics of the residual computing resources of the global target equipment;
wherein, the flow distribution module includes:
the feature distribution optimization unit is used for performing feature distribution optimization on the target equipment global dimension computing resource time sequence associated feature vector to obtain an optimized target equipment global dimension computing resource time sequence associated feature vector;
the target equipment residual computing resource mapping association unit is used for respectively calculating transfer matrixes of the residual computing resource time sequence feature vectors of each target equipment relative to the optimizing target equipment global dimension computing resource time sequence association feature vectors so as to obtain a plurality of transfer matrixes;
a calculation unit for calculating remaining resource allocation probability, which is used for passing the plurality of transition matrixes through a classifier to obtain a plurality of probability values and carrying out maximum value-based normalization processing on the plurality of probability values; and
A flow distribution ratio determining unit configured to set the plurality of probability values as flow distribution ratios;
wherein the feature distribution optimizing unit includes:
the vector cascading subunit is used for cascading the residual computing resource time sequence feature vectors of the plurality of target devices to obtain cascading feature vectors;
the tracking equalization fusion subunit is used for carrying out Hilbert space heuristic sequence tracking equalization fusion on the cascade feature vector and the target equipment global dimension computing resource time sequence correlation feature vector so as to obtain the optimized target equipment global dimension computing resource time sequence correlation feature vector;
the tracking equalization fusion subunit is configured to: carrying out Hilbert space heuristic sequence tracking equalization fusion on the cascade feature vector and the target equipment global dimension computing resource time sequence association feature vector by using the following optimization fusion formula to obtain the optimization target equipment global dimension computing resource time sequence association feature vector;
wherein, the optimized fusion formula is:
wherein V is 1 Is the cascade feature vector, V 2 Is the target device global dimension computing resource time sequence association feature vector, | (V) 1 ；V 2 )‖ 2 Representing feature vector V 1 And V 2 Is used to determine the two norms of the cascade of vectors,representing feature vector V 1 And V 2 Mean value of union set composed of all eigenvalues of (c), and eigenvector V 1 And V 2 Are all row vectors, +. 2 ' is the optimization target device global dimension computing resource timing association feature vector.
2. The system for controlling a converging splitter according to claim 1, wherein the data correlation analysis module comprises:
the residual computing resource time sequence distribution unit is used for respectively arranging residual computing resource amounts of a plurality of preset time points of each target device in a preset time period into input vectors according to a time dimension so as to obtain residual computing resource time sequence input vectors of a plurality of target devices;
the residual computing resource time sequence change feature extraction unit is used for respectively carrying out feature extraction on the residual computing resource time sequence input vectors of the plurality of target devices through a time sequence feature extractor based on a deep neural network model so as to obtain residual computing resource time sequence feature vectors of the plurality of target devices;
and the target equipment global dimension computing resource association coding unit is used for carrying out global association coding on the plurality of target equipment residual computing resource time sequence feature vectors to obtain target equipment global dimension computing resource time sequence association feature vectors serving as the global target equipment residual computing resource time sequence association features.
3. The system of claim 2, wherein the deep neural network model is a one-dimensional convolutional neural network model.
4. The system for controlling a convergence splitter as claimed in claim 3, wherein the target device global dimension computing resource association encoding unit is configured to: and passing the residual computing resource time sequence feature vectors of the plurality of target devices through a computing resource global semantic encoder based on a converter module to obtain the target device global dimension computing resource time sequence association feature vector.
5. A method for controlling a converging splitter, comprising:
acquiring the residual computing resource amounts of a plurality of target devices at a plurality of preset time points in a preset time period;
performing association analysis on the residual computing resource amounts of a plurality of preset time points of the plurality of target devices in a preset time period to obtain a global target device residual computing resource time sequence association characteristic;
determining a flow distribution proportion based on the residual computing resource time sequence correlation characteristics of the global target equipment;
wherein determining a traffic distribution ratio based on the global target device remaining computing resource timing correlation characteristics comprises:
Performing feature distribution optimization on the target equipment global dimension computing resource time sequence associated feature vector to obtain an optimized target equipment global dimension computing resource time sequence associated feature vector;
calculating transfer matrixes of residual computing resource time sequence feature vectors of each target device relative to the global dimension computing resource time sequence associated feature vectors of the optimized target device respectively to obtain a plurality of transfer matrixes;
passing the plurality of transition matrixes through a classifier to obtain a plurality of probability values and carrying out maximum value-based normalization processing on the plurality of probability values; and
setting the probability values as flow distribution ratios;
performing feature distribution optimization on the target equipment global dimension computing resource time sequence associated feature vector to obtain an optimized target equipment global dimension computing resource time sequence associated feature vector, including:
cascading the residual computing resource time sequence feature vectors of the plurality of target devices to obtain cascading feature vectors;
performing Hilbert space heuristic sequence tracking equalization fusion on the cascade feature vector and the target equipment global dimension computing resource time sequence association feature vector to obtain the optimized target equipment global dimension computing resource time sequence association feature vector;
The performing hilbert space heuristic sequence tracking equalization fusion on the cascade feature vector and the target equipment global dimension computing resource time sequence association feature vector to obtain the optimized target equipment global dimension computing resource time sequence association feature vector comprises the following steps: carrying out Hilbert space heuristic sequence tracking equalization fusion on the cascade feature vector and the target equipment global dimension computing resource time sequence association feature vector by using the following optimization fusion formula to obtain the optimization target equipment global dimension computing resource time sequence association feature vector;
wherein, the optimized fusion formula is:
wherein V is 1 Is the cascade feature vector, V 2 Is the target device global dimension computing resource time sequence association feature vector, | (V) 1 ；V 2 )‖ 2 Representing feature vector V 1 And V 2 Is used to determine the two norms of the cascade of vectors,representing feature vector V 1 And V 2 Mean value of union set composed of all eigenvalues of (c), and eigenvector V 1 And V 2 Are all row vectors, +. 2 ' is the optimization target device global dimension computing resource timing association feature vector.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311051172.7A CN116760772B (en) | 2023-08-21 | 2023-08-21 | Control system and method for converging flow divider |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311051172.7A CN116760772B (en) | 2023-08-21 | 2023-08-21 | Control system and method for converging flow divider |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN116760772A CN116760772A (en) | 2023-09-15 |
| CN116760772B true CN116760772B (en) | 2023-10-27 |
Family
ID=87951920
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202311051172.7A Active CN116760772B (en) | 2023-08-21 | 2023-08-21 | Control system and method for converging flow divider |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116760772B (en) |
Families Citing this family (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116953859A (en) * | 2023-09-18 | 2023-10-27 | 广东连捷精密技术有限公司 | Multi-channel high-speed optical port connector and optical connection method |
| CN117254960A (en) * | 2023-09-25 | 2023-12-19 | 深圳市云钜天成信息技术有限公司 | Detection method for detecting API interface verification risk from flow data |
| CN117061322A (en) * | 2023-09-27 | 2023-11-14 | 广东云百科技有限公司 | Internet of things flow pool management method and system |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114860462A (en) * | 2022-07-11 | 2022-08-05 | 南京普道电子技术有限公司 | Intelligent computing resource distribution system and method for two-way rack-mounted server |
| CN116385365A (en) * | 2023-03-06 | 2023-07-04 | 华能太仓发电有限责任公司 | Coal flow monitoring system and method based on scanning image |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20180284758A1 (en) * | 2016-05-09 | 2018-10-04 | StrongForce IoT Portfolio 2016, LLC | Methods and systems for industrial internet of things data collection for equipment analysis in an upstream oil and gas environment |
-
2023
- 2023-08-21 CN CN202311051172.7A patent/CN116760772B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114860462A (en) * | 2022-07-11 | 2022-08-05 | 南京普道电子技术有限公司 | Intelligent computing resource distribution system and method for two-way rack-mounted server |
| CN116385365A (en) * | 2023-03-06 | 2023-07-04 | 华能太仓发电有限责任公司 | Coal flow monitoring system and method based on scanning image |
Non-Patent Citations (1)
| Title |
|---|
| 支持随机服务请求的云虚拟机按需物理资源分配方法;曹洁;曾国荪;匡桂娟;张建伟;马海英;胡克坤;钮俊;;软件学报(02);全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN116760772A (en) | 2023-09-15 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN116760772B (en) | Control system and method for converging flow divider | |
| CN104951425B (en) | A kind of cloud service performance self-adapting type of action system of selection based on deep learning | |
| US11650968B2 (en) | Systems and methods for predictive early stopping in neural network training | |
| CN111435463A (en) | Data processing method and related equipment and system | |
| CN110390425A (en) | Prediction technique and device | |
| CN110571792A (en) | Analysis and evaluation method and system for operation state of power grid regulation and control system | |
| CN109445935A (en) | A kind of high-performance big data analysis system self-adaption configuration method under cloud computing environment | |
| CN113506009B (en) | Equipment management method and system based on smart cloud service | |
| WO2019214455A1 (en) | Data sequence prediction method and computing apparatus | |
| CN110147389A (en) | Account number treating method and apparatus, storage medium and electronic device | |
| CN114066073A (en) | Power grid load prediction method | |
| CN117041017B (en) | Intelligent operation and maintenance management method and system for data center | |
| CN113379059B (en) | Model training method for quantum data classification and quantum data classification method | |
| CN114625477A (en) | Service node capacity adjusting method, equipment and computer readable storage medium | |
| CN116909712A (en) | Intelligent task scheduling system and method based on machine learning | |
| CN116861924A (en) | Project risk early warning method and system based on artificial intelligence | |
| CN116976461A (en) | Federal learning method, apparatus, device and medium | |
| CN115983497A (en) | Time sequence data prediction method and device, computer equipment and storage medium | |
| CN113010687B (en) | Exercise label prediction method and device, storage medium and computer equipment | |
| CN114970357A (en) | Energy-saving effect evaluation method, system, device and storage medium | |
| CN115392102A (en) | Method and device for establishing energy consumption prediction model and method and system for predicting energy consumption | |
| CN113554010A (en) | Power grid line fault recognition model training method | |
| Du et al. | OctopusKing: A TCT-aware task scheduling on spark platform | |
| CN116957304B (en) | Unmanned aerial vehicle group collaborative task allocation method and system | |
| CN116466672B (en) | Data center machine room parameter regulation and control method based on artificial intelligence and related device |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |