CN112422451A - Service sensing method and device of network edge system - Google Patents
Service sensing method and device of network edge system Download PDFInfo
- Publication number
- CN112422451A CN112422451A CN202011089568.7A CN202011089568A CN112422451A CN 112422451 A CN112422451 A CN 112422451A CN 202011089568 A CN202011089568 A CN 202011089568A CN 112422451 A CN112422451 A CN 112422451A
- Authority
- CN
- China
- Prior art keywords
- esn
- network
- service
- classifier
- service flow
- 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.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 65
- 238000012549 training Methods 0.000 claims abstract description 81
- 238000012545 processing Methods 0.000 claims description 19
- 238000004590 computer program Methods 0.000 claims description 18
- 239000011159 matrix material Substances 0.000 claims description 17
- 238000010606 normalization Methods 0.000 claims description 13
- 230000006870 function Effects 0.000 claims description 11
- 230000014509 gene expression Effects 0.000 claims description 10
- 238000003860 storage Methods 0.000 claims description 9
- 238000005457 optimization Methods 0.000 claims description 5
- 230000005284 excitation Effects 0.000 claims description 3
- 230000008447 perception Effects 0.000 abstract description 14
- 230000000694 effects Effects 0.000 abstract description 7
- 238000010586 diagram Methods 0.000 description 16
- 230000008569 process Effects 0.000 description 11
- 238000004891 communication Methods 0.000 description 9
- 238000013528 artificial neural network Methods 0.000 description 7
- 230000005540 biological transmission Effects 0.000 description 7
- 238000004422 calculation algorithm Methods 0.000 description 7
- 230000000903 blocking effect Effects 0.000 description 6
- 238000005516 engineering process Methods 0.000 description 6
- 238000004088 simulation Methods 0.000 description 6
- 239000010410 layer Substances 0.000 description 5
- 239000012792 core layer Substances 0.000 description 3
- 230000003190 augmentative effect Effects 0.000 description 2
- 238000007796 conventional method Methods 0.000 description 2
- 230000007246 mechanism Effects 0.000 description 2
- 210000002569 neuron Anatomy 0.000 description 2
- 230000004913 activation Effects 0.000 description 1
- 238000013459 approach Methods 0.000 description 1
- 238000013473 artificial intelligence Methods 0.000 description 1
- 230000006399 behavior Effects 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 230000015556 catabolic process Effects 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 230000007423 decrease Effects 0.000 description 1
- 238000006731 degradation reaction Methods 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000001514 detection method Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000005183 dynamical system Methods 0.000 description 1
- 239000000284 extract Substances 0.000 description 1
- 230000003993 interaction Effects 0.000 description 1
- 238000012417 linear regression Methods 0.000 description 1
- 230000007774 longterm Effects 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 238000003909 pattern recognition Methods 0.000 description 1
- 230000000306 recurrent effect Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 230000003595 spectral effect Effects 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 239000013589 supplement Substances 0.000 description 1
Images
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
- H04L47/24—Traffic characterised by specific attributes, e.g. priority or QoS
- H04L47/2483—Traffic characterised by specific attributes, e.g. priority or QoS involving identification of individual flows
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
-
- 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
- H04L47/24—Traffic characterised by specific attributes, e.g. priority or QoS
- H04L47/2425—Traffic characterised by specific attributes, e.g. priority or QoS for supporting services specification, e.g. SLA
- H04L47/2433—Allocation of priorities to traffic types
-
- 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
- H04L47/24—Traffic characterised by specific attributes, e.g. priority or QoS
- H04L47/2441—Traffic characterised by specific attributes, e.g. priority or QoS relying on flow classification, e.g. using integrated services [IntServ]
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Signal Processing (AREA)
- Computer Networks & Wireless Communication (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Data Exchanges In Wide-Area Networks (AREA)
Abstract
The invention provides a service perception method and a device of a network edge system, wherein the method comprises the following steps: acquiring a sample training set of services of a network edge system, wherein the network edge system comprises a plurality of network units; performing ESN training on the echo state network based on the sample training set to generate a trained ESN classifier; configuring the ESN classifier of each network unit based on the trained parameter information of the ESN classifier; extracting the characteristic parameters of each service flow in each network unit aiming at each network unit, inputting the characteristic parameters of each service flow into an ESN classifier of the network unit, and obtaining the service classification identification result of each service flow; and aiming at each network unit, carrying out optimized scheduling on the service flow according to the service classification identification result of each service flow in the network unit. The invention can sense the network edge service and has good effect.
Description
Technical Field
The present invention relates to the field of communications technologies, and in particular, to a method and an apparatus for service awareness in a network edge system.
Background
With the development of the internet of things, the appearance of more and more abundant services (such as vehicle-mounted communication, virtual reality, augmented reality), the interaction modes between people, between people and the environment, and between people and machines are also changing greatly. While more and more mobile devices and applications are beginning to use the undirected network. In order to improve user experience and ensure network service quality, the prior art proposes edge computing, which is also an important component of fifth generation networks. NFV and SDN are receiving increasing attention as key technologies for 5G networks. NFV uses virtualization technology to flexibly deploy network services. At the same time, SDN may provide all information about network topology and resources. The combination of NFV and SDN provides a new approach to the deployment of network services. But due to the performance limitations of the edge server, as the number of users and devices accessing the network increases, the quality of service of the network decreases if the edge server is unable to provide the corresponding network resources.
The self-similarity, asymmetry, dynamics, burstiness and diversity characteristics of the data service are different from the stable voice service characteristics. The characteristics of the traffic are different and the demands on the network edge are also different. In edge computing, users use various types of network services, such as video transmission, virtual reality, augmented reality, mailboxes, online gaming, online medical education, and the like. These network services have different demands on network resources. For example, video transmission requires high-speed, low-latency data transmission capabilities, while virtual reality requires the network to provide significant computing power. In addition, with the diversification of network services and applications, the amount of data involved is increasing exponentially, and the traffic flow shows characteristics of long-term dependence on terminal behavior, multi-fractal characteristics, and some nonlinear characteristics, which are very complicated. How to sense the change of service characteristics in real time at the edge of the network and adjust parameters in time to meet the requirements of different services, so as to provide differentiated services, has become a problem to be solved urgently.
Disclosure of Invention
The embodiment of the invention provides a service perception method of a network edge system, which is used for perceiving network edge services and has good effect, and the method comprises the following steps:
acquiring a sample training set of services of a network edge system, wherein the network edge system comprises a plurality of network units;
performing ESN training on the echo state network based on the sample training set to generate a trained ESN classifier;
configuring the ESN classifier of each network unit based on the trained parameter information of the ESN classifier;
extracting the characteristic parameters of each service flow in each network unit aiming at each network unit, inputting the characteristic parameters of each service flow into an ESN classifier of the network unit, and obtaining the service classification identification result of each service flow;
and aiming at each network unit, carrying out optimized scheduling on the service flow according to the service classification identification result of each service flow in the network unit.
The embodiment of the invention provides a service sensing device of a network edge system, which is used for sensing network edge services and has good effect, and the device comprises: the system comprises a training module, a main control module, a plurality of ESN agent modules and a plurality of scheduling modules, wherein each network unit of the network edge system corresponds to one ESN agent module and one scheduling module respectively;
the training module is used for acquiring a sample training set of the service of the network edge system;
the main control module is used for carrying out ESN training on the echo state network based on the sample training set to generate a trained ESN classifier;
the ESN agent module is used for configuring the ESN classifier of the corresponding network unit based on the parameter information of the ESN classifier; extracting the characteristic parameter of each service flow in the corresponding network unit, inputting the characteristic parameter of each service flow into an ESN classifier of the network unit, and obtaining a service classification identification result of each service flow;
and the scheduling module is used for carrying out service flow optimization scheduling according to the service classification identification result of each service flow in the corresponding network unit.
The embodiment of the present invention further provides a computer device, which includes a memory, a processor, and a computer program stored in the memory and executable on the processor, and when the processor executes the computer program, the service awareness method of the network edge system is implemented.
An embodiment of the present invention further provides a computer-readable storage medium, where a computer program for executing the service awareness method of the network edge system is stored in the computer-readable storage medium.
In the embodiment of the invention, a sample training set of the service of a network edge system is obtained, wherein the network edge system comprises a plurality of network units; performing ESN training on the echo state network based on the sample training set to generate a trained ESN classifier; configuring the ESN classifier of each network unit based on the trained parameter information of the ESN classifier; extracting the characteristic parameters of each service flow in each network unit aiming at each network unit, inputting the characteristic parameters of each service flow into an ESN classifier of the network unit, and obtaining the service classification identification result of each service flow; and aiming at each network unit, carrying out optimized scheduling on the service flow according to the service classification identification result of each service flow in the network unit. In the process, the echo state network ESN is adopted to finally obtain the service classification recognition result of each service flow, so that the optimized scheduling of the service flows can be performed, the service perception is realized, and the effect is good.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below, it is obvious that the drawings in the following description are only some embodiments of the present invention, and for those skilled in the art, other drawings can be obtained according to the drawings without creative efforts. In the drawings:
fig. 1 is a flowchart of a service awareness method of a network edge system according to an embodiment of the present invention;
fig. 2 is a schematic diagram of a service sensing apparatus of a network edge system according to an embodiment of the present invention;
FIG. 3 is a schematic diagram of an echo state network according to an embodiment of the present invention;
FIG. 4 is a diagram illustrating the relationship between the accuracy of service sensing and the training times of an echo state network according to an embodiment of the present invention;
FIGS. 5 and 6 are graphs comparing results of the method of the present invention and the conventional method in the example of the present invention, respectively;
fig. 7 is a relationship between a user arrival rate and a blocking rate corresponding to different methods in the embodiment of the present invention;
FIG. 8 is a diagram of a computer device in an embodiment of the invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the embodiments of the present invention more apparent, the embodiments of the present invention are further described in detail below with reference to the accompanying drawings. The exemplary embodiments and descriptions of the present invention are provided to explain the present invention, but not to limit the present invention.
In the description of the present specification, the terms "comprising," "including," "having," "containing," and the like are used in an open-ended fashion, i.e., to mean including, but not limited to. Reference to the description of the terms "one embodiment," "a particular embodiment," "some embodiments," "for example," etc., means that a particular feature, structure, or characteristic described in connection with the embodiment or example is included in at least one embodiment or example of the application. In this specification, the schematic representations of the terms used above do not necessarily refer to the same embodiment or example. Furthermore, the particular features, structures, or characteristics described may be combined in any suitable manner in any one or more embodiments or examples. The sequence of steps involved in the embodiments is for illustrative purposes to illustrate the implementation of the present application, and the sequence of steps is not limited and can be adjusted as needed.
Fig. 1 is a flowchart of a service awareness method of a network edge system in an embodiment of the present invention, as shown in fig. 1, the method includes:
102, performing ESN training on the echo state network based on the sample training set to generate a trained ESN classifier;
103, configuring the ESN classifier of each network unit based on the trained parameter information of the ESN classifier;
104, aiming at each network unit, extracting the characteristic parameter of each service flow in the network unit, inputting the characteristic parameter of each service flow into an ESN classifier of the network unit, and obtaining a service classification identification result of each service flow;
and 105, aiming at each network unit, performing optimized scheduling on the service flow according to the service classification identification result of each service flow in the network unit.
Similarly, fig. 2 is a schematic diagram of a service awareness apparatus of a network edge system in an embodiment of the present invention, as shown in fig. 2, the apparatus includes: the system comprises a training module 201, a main control module 202, a plurality of ESN agent modules 203 and a plurality of scheduling modules 204, wherein each network unit of the network edge system corresponds to one ESN agent module 203 and one scheduling module 204 respectively;
a training module 201, configured to obtain a sample training set of a service of a network edge system;
the main control module 202 is configured to perform echo state network ESN training based on the sample training set, and generate a trained ESN classifier;
the ESN agent module 203 is configured to configure the ESN classifier of the corresponding network element based on the trained parameter information of the ESN classifier; extracting the characteristic parameter of each service flow in the corresponding network unit, inputting the characteristic parameter of each service flow into an ESN classifier of the network unit, and obtaining a service classification identification result of each service flow;
and the scheduling module 204 is configured to perform service flow optimization scheduling according to the service classification identification result of each service flow in the corresponding network unit.
In the embodiment of the invention, the echo state network ESN is adopted to finally obtain the service classification recognition result of each service flow, so that the optimized scheduling of the service flows can be carried out, the service perception is realized, and the effect is good.
In specific implementation, fig. 3 is a schematic diagram of an echo state network structure in an embodiment of the present invention, and an Echo State Network (ESN) algorithm is a novel neural network algorithm that can be used for pattern recognition, and is a simplified form of a Recurrent Neural Network (RNN), and is a discrete neural network with a three-layer structure, and is also a neural network with a feedback structure. Compared with the traditional prediction algorithm, the method has stronger nonlinear prediction capability and superior simulation performance of the nonlinear dynamical system. The ESN algorithm is one of neural network prediction algorithms which are widely researched and applied at present, and is widely applied to the field of artificial intelligence. In the embodiment of the invention, based on a mobile service flow awareness mechanism of an Echo State Network (ESN), a discrete Echo State Network algorithm is adopted to identify and perceive the service flow, so that the high-efficiency matching capability of the service in the Network edge is realized, and the research of the service perception technology at the Network edge is realized.
In step 101, a sample training set (x) of traffic of a network edge system is obtained1(t),...,xK(t)),xK(t) represents the time series of the traffic flow, and K represents the total number of the time series of the traffic flow; referring to step 102 and the main control module 202, based on the sample training set, performing echo state network ESN training to generate a trained ESN classifier, and obtaining parameter information in the trained ESN classifier (i.e. an output weight connection matrix W of the trained ESN classifier)out). The ESN training method is the same as the existing ESN training method.
The ESN adopts a reserve pool composed of randomly sparsely connected neurons as a hidden layer for performing high-dimensional and nonlinear representation on input. The generation process of the reserve pool is independent of the training process of the echo state network, so that the training process of the network is simplified by only adopting a linear method to train the weight from the reserve pool to the output layer, the global optimality and good generalization capability of weight determination are ensured, and the problems of complex training algorithm, easy falling into local minimum and the like in the traditional neural network are avoided. The advantages enable the echo state network to have great application potential in traffic perception.
In addition, the network unit counts the record (including the characteristic value and the classification recognition result) of the traffic flow classification perception, and periodically feeds the record back to the training module, and continuously supplements a new sample training set. And the main control module periodically carries out ESN training again according to the sample training set so as to form a new ESN classifier.
The echogenic nature of the ESN is a prerequisite for successful ESN training. By echo state is meant that the internal state of the ESN is a finite function of the input history data, i.e. the "echo" of the input history data. To ensure that the ESN has an echo state characteristic, therefore, when initializing the intermediate weight connection matrix W that constructs the ESN, the spectral radius of W should be made smaller than 1. The ESN neural network has the greatest advantages that the training method is simple, and the dynamic pool structure adopts simple random and sparse connection. The echo state network is composed of K input units, N reserve pool processing units and N output units. In one embodiment, the expression of the echo state network ESN is as follows:
x(n+1)=f(Winu(n+1)+Wx(n)+Wbacky(n)) (1)
y(n+1)=fout(Woutu(n+1)+Wx(n+1)+Wbacky(n)) (2)
wherein u (n) is an input variable, and u (n) is (u)1(n),…,uK(n))T(ii) a x (n) is an internal state variable, x (n) ═ x1(n),...,xL(n))T(ii) a y (n) is an output variable y (n) ═ y1(n),...,yL(n))T;WinConnecting the matrix for the input weights; w is an intermediate weight connection matrix; woutConnecting a matrix for the output weights; wbackConnecting a matrix for feedback weights; f and foutIs an excitation function; n are different samples in the sample training set.
The expression (1) is a state update equation of an internal neuron node, and the expression (2) is an output prediction equation. f and foutExcitation functions of the reserve pool processing unit and the output unit respectively are generally sigmoid functions, and expressions of the excitation functions are shown as follows:
in ESN training, a sample training set is connected with moments through randomly generated weightsArray WinAnd WbackThe processing unit exciting the core layer can correct the ESN internal parameters after each round of training by adopting linear regression, and can reduce the mean square error to the minimum. Input variable is passed through WinIs connected with the processing units of the ESN, W is the connection weight between the processing units in the ESN, WbackRepresents the connection weight, W, of the output layer and the core layeroutThe connection weight of the core layer and the output layer. During the training process, the internal state weight matrix of the dynamic pool is kept unchanged, and only the output weight connection matrix W from the dynamic pool to the output is updatedoutAnd calculating the minimum Mean Square Error (MSE)min. Further, WinW and WbackUsually set as a constant, the output weight connection matrix W is obtained by a certain trainingout. Mean square error MSEminThe expression used is as follows:
when MSEminAt the minimum, W is calculated by using an offline generalized inverse matrixoutI.e. Wout=M-1T, where M ═ (x (n)), and T ═ d (n)). When W isoutAfter the calculation is finished, the trained ESN classifier is generated, prediction can be started, and the output weight at the moment is connected with the matrix WoutNamely the parameter information of the trained ESN classifier.
In the embodiment of the invention, the service perception device of the network edge system comprises an ESN agent module and a scheduling module corresponding to each network unit, and the main control module is used for acquiring the parameter information W of the trained ESN classifieroutAnd each ESN agent module configures the ESN classifier of the corresponding network element to ensure the consistency of the ESN classifiers of the ESN agent modules in service flow classification perception.
In an embodiment, the characteristic parameter of the traffic flow comprises at least one of a packet length, a packet arrival interval, a traffic duration and a load level of the network element. In order to fully consider the influence of the traffic load degree of the network unit ONU node on the traffic flow, the load degree is particularly introduced as one of characteristic parameters.
In an embodiment, before performing the echo state network ESN training based on the sample training set, the method further includes:
carrying out normalization processing on the sample training set;
based on the sample training set, carrying out ESN training of the echo state network, comprising:
and performing ESN training on the echo state network based on the sample training set after the normalization processing.
In the above embodiment, the sample training set is a large amount of traffic flow data, and during training, feature parameters are also extracted, that is, the sample training set is normalized, that is, the feature parameters of the traffic flow data are normalized, so as to avoid an over-fitting phenomenon, and a feature set u (i) ═ x (x) describing the traffic flow is obtainedi,1,xi,2,xi,3,xi,4)。
The normalization process uses the following expression:
wherein: pSIZE(i) Is data packet length, PSIZE_MAXFor statistical maximum packet length, PINTERVAL(i) For packet inter-arrival, PINTERVAL_MAXIs the maximum arrival interval, PDUR(i) For the duration of a service, PDUR_MAXFor maximum service duration, PLOAD(i) Is the load level, P, of the network elementLOAD_MAXIs the maximum load rate of the network element.
Of course, it is understood that other normalization methods of the traffic stream data may be adopted besides the normalization method, and all the related variations should fall within the scope of the present invention.
An ESN agent module for configuring the ESN classifier of the corresponding network unit based on the trained parameter information of the ESN classifier, mainly referring to the corresponding networkW of the unitoutAnd (5) carrying out configuration.
In an embodiment, after configuring the ESN classifier of each network element based on the parameter information of the trained ESN classifier, the method further includes:
the ESN classifier for each network element is implemented and consolidated in hardware.
In the above embodiment, after the ESN agent module completes the above curing, the ESN agent modules in the network elements respectively and independently operate to perform service flow sensing under the condition that the ESN classifier of each network element is consistent.
In one embodiment, before inputting the characteristic parameter of each traffic flow into the ESN classifier of the network element, the method further includes:
carrying out normalization processing on the characteristic parameters of each service flow;
inputting the characteristic parameters of each traffic flow into the ESN classifier of the network element, comprising: and inputting the characteristic parameters of each service flow after normalization processing into an ESN classifier of the network unit to obtain a service classification identification result.
In the above embodiment, the normalization process may still be performed using expression (5).
In one embodiment, the expression for the ESN classifier for each network element is as follows:
where n represents different samples. In the classification process, the input samples must be kept unchanged all the time until the value of the reservoir state variable tends to be stable, so that the difference between the results of the two previous iterations and the results of the two subsequent iterations is minimized. Therefore, a service classification identification result is obtained, and in the embodiment of the invention, the service classification identification result is the priority of the service flow. In the embodiment of the invention, the state variable is processed to be stable only by the activation function of the reserve pool processing unit, and the characteristics of simple and direct training process of the echo state network are still kept, so that the overall complexity is reduced, and the operation result is ensured to have the globally optimal performance.
The scheduling module can perform service flow optimization scheduling according to the service classification identification result of each service flow in the corresponding network unit, realize the perception of the service at the network edge, and improve the regional branch support capability of the scheduling module on diversified service QoS while exerting the capacity and bandwidth advantages of the technology.
An embodiment is given below to illustrate a specific application of the method and apparatus of the present invention.
The embodiment of the invention constructs a simulation system for service perception at the network edge on a Mininet-based simulation software platform, wherein the simulation system is formed by mounting 32 network units on a terminal, an SDN network is designed, and edge nodes are added, so that the method and the device provided by the embodiment of the invention can be verified to obtain satisfactory effects. The network edge with the traffic aware mechanism and the network edge without traffic awareness are compared by emulation. The service types are divided into two types in simulation: the internal priority of each type of service is divided into two grades: i.e., high priority traffic and low priority traffic.
The main control module inputs a large number of sample training sets from the training module to perform uniform ESN training by adopting expressions (1) - (4) until a complete trained ESN classifier is formed, and the adopted ESN training method is the same as the existing ESN training method.
And then, the main control module distributes the parameter information of the trained ESN classifier to each network unit, an ESN agent module of each network unit configures the ESN classifier of each network unit, extracts the characteristic parameters of each service flow in the corresponding network unit, performs normalization processing, inputs the processed characteristic parameters of each service flow into the ESN classifier, and obtains a service classification identification result of each service flow.
And finally, the scheduling module performs service flow optimization scheduling according to the service classification identification result of each service flow in the corresponding network unit, realizes the perception of the service at the network edge, and improves the regional support capability of the scheduling module on diversified service QoS while exerting the capacity and bandwidth advantages of the technology.
Fig. 4 is a diagram illustrating the relationship between the service sensing accuracy and the training times of the echo state network in the embodiment of the present invention. Therefore, the accuracy of service perception is improved along with the increase of the training times, and the accuracy of service classification and classification can be ensured through the fully-trained echo state network.
At the network edge, the method proposed by the embodiment of the present invention and the conventional non-service-aware method are respectively adopted for simulation, and fig. 5 and fig. 6 are respectively comparison graphs of the results of the method of the present invention and the conventional method in the embodiment of the present invention. As can be seen from the simulation results shown in fig. 5 and fig. 6, as the network traffic load increases, the packet loss rate and the real-time performance of the method of the present invention both show a tendency of degradation. Under the condition of high network service load, the network edge system adopting the method of the invention is superior to the traditional method in two important indexes of packet loss rate and transmission delay of high-priority service. On the other hand, because the requirements of low-priority services on transmission delay and packet loss rate are low, the method of the invention reduces the performance cost of the services to a certain extent, and replaces the overall service quality, especially ensures the QoS requirement of high-priority services.
The comparison between the packet loss rate and the transmission delay in fig. 5 and fig. 6 illustrates that the method proposed herein can ensure that services of different types and different priorities at the network edge can obtain the service quality matched therewith and ensure the overall service quality of the services under the condition of low computational complexity.
Fig. 7 is a relationship between user arrival rates and blocking rates corresponding to different methods in the embodiment of the present invention, and as can be seen from fig. 7, at the beginning, the user arrival rate is small, network resources are sufficient, and the user requirements can be met, so that the blocking rate is almost 0. However, as the user arrival rate is continuously increased, the blocking rate gradually increases, and it can be seen that under the condition that the user arrival rate is the same, the blocking rate of the method of the present invention is lower than that of the AHP-SAW and TOPSIS, because the method of the present invention classifies users according to the service types when the users request the switch, and designs different strategies aiming at the users of different service types, thereby avoiding the network congestion caused by the simultaneous switch of a large number of users to the same network, and effectively reducing the blocking rate of the network.
In summary, in the method and apparatus provided in the embodiments of the present invention, a sample training set of a service of a network edge system is obtained, where the network edge system includes a plurality of network units; performing ESN training on the echo state network based on the sample training set to generate a trained ESN classifier; configuring the ESN classifier of each network unit based on the trained parameter information of the ESN classifier; extracting the characteristic parameters of each service flow in each network unit aiming at each network unit, inputting the characteristic parameters of each service flow into an ESN classifier of the network unit, and obtaining the service classification identification result of each service flow; and aiming at each network unit, carrying out optimized scheduling on the service flow according to the service classification identification result of each service flow in the network unit. In the process, the echo state network ESN is adopted to finally obtain the service classification recognition result of each service flow, so that the optimized scheduling of the service flows can be performed, the service perception is realized, and the effect is good.
An embodiment of the present application further provides a computer device, and fig. 8 is a schematic diagram of a computer device in an embodiment of the present invention, where the computer device is capable of implementing all steps in the service awareness method of the network edge system in the foregoing embodiment, and the computer device specifically includes the following contents:
a processor (processor)801, a memory (memory)802, a communication Interface (Communications Interface)803, and a communication bus 804;
the processor 801, the memory 802 and the communication interface 803 complete mutual communication through the communication bus 804; the communication interface 803 is used for realizing information transmission among related devices such as server-side devices, detection devices, client-side devices and the like;
the processor 801 is configured to call a computer program in the memory 802, and when the processor executes the computer program, the processor implements all the steps in the service awareness method of the network edge system in the above embodiments.
An embodiment of the present application further provides a computer-readable storage medium, which can implement all the steps in the service awareness method of the network edge system in the foregoing embodiment, where the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the computer program implements all the steps of the service awareness method of the network edge system in the foregoing embodiment.
As will be appreciated by one skilled in the art, embodiments of the present invention may be provided as a method, system, or computer program product. Accordingly, the present invention may take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, the present invention may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, and the like) having computer-usable program code embodied therein.
The present invention is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each flow and/or block of the flow diagrams and/or block diagrams, and combinations of flows and/or blocks in the flow diagrams and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means which implement the function specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
The above-mentioned embodiments are intended to illustrate the objects, technical solutions and advantages of the present invention in further detail, and it should be understood that the above-mentioned embodiments are only exemplary embodiments of the present invention, and are not intended to limit the scope of the present invention, and any modifications, equivalent substitutions, improvements and the like made within the spirit and principle of the present invention should be included in the scope of the present invention.
Claims (10)
1. A service awareness method for a network edge system, comprising:
acquiring a sample training set of services of a network edge system, wherein the network edge system comprises a plurality of network units;
performing ESN training on the echo state network based on the sample training set to generate a trained ESN classifier;
configuring the ESN classifier of each network unit based on the trained parameter information of the ESN classifier;
extracting the characteristic parameters of each service flow in each network unit aiming at each network unit, inputting the characteristic parameters of each service flow into an ESN classifier of the network unit, and obtaining the service classification identification result of each service flow;
and aiming at each network unit, carrying out optimized scheduling on the service flow according to the service classification identification result of each service flow in the network unit.
2. The traffic-aware method of a network edge system of claim 1, further comprising, prior to performing echo-state network ESN training based on the sample training set:
carrying out normalization processing on the sample training set;
based on the sample training set, carrying out ESN training of the echo state network, comprising:
and performing ESN training on the echo state network based on the sample training set after the normalization processing.
3. The method of service awareness for a network edge system according to claim 1, wherein after configuring the ESN classifier of each network element based on the trained parameter information of the ESN classifier, the method further comprises:
the ESN classifier for each network element is implemented and consolidated in hardware.
4. The traffic-aware method of a network edge system of claim 1, wherein the characteristic parameters of the traffic flow comprise at least one of packet length, packet inter-arrival interval, traffic duration, and load level of the network element.
5. The traffic-aware method of a network edge system of claim 1, wherein before inputting the characteristic parameters of each traffic flow into the ESN classifier of the network element, further comprising:
carrying out normalization processing on the characteristic parameters of each service flow;
inputting the characteristic parameters of each traffic flow into the ESN classifier of the network element, comprising: and inputting the characteristic parameters of each service flow after normalization processing into an ESN classifier of the network unit to obtain a service classification identification result.
6. The service aware method of a network edge system as claimed in claim 1, wherein the expression of the ESN is as follows:
x(n+1)=f(Winu(n+1)+Wx(n)+Wbacky(n))
y(n+1)=fout(Woutu(n+1)+Wx(n+1)+Wbacky(n))
wherein u (n) is an input variable, and u (n) is (u)1(n),…,uK(n))T(ii) a x (n) is an internal state variable, x (n) ═ x1(n),...,xL(n))T(ii) a y (n) is an output variable y (n) ═ y1(n),...,yL(n))T;WinConnecting the matrix for the input weights; w is an intermediate weight connection matrix; woutConnecting a matrix for the output weights; wbackConnecting a matrix for feedback weights; f and foutIs an excitation function; n are different samples in the sample training set.
7. The traffic-aware method of a network edge system of claim 6, wherein the ESN classifier for each network element is expressed as follows:
wherein u (n) is an input variable, and u (n) is (u)1(n),…,uK(n))T(ii) a x (n) is an internal state variable, x (n) ═ x1(n),...,xL(n))T;WinConnecting the matrix for the input weights; w is an intermediate weight connection matrix; n are different samples in the sample training set.
8. A traffic aware apparatus for a network edge system, comprising: the system comprises a training module, a main control module, a plurality of ESN agent modules and a plurality of scheduling modules, wherein each network unit of the network edge system corresponds to one ESN agent module and one scheduling module respectively;
the training module is used for acquiring a sample training set of the service of the network edge system;
the main control module is used for carrying out ESN training on the echo state network based on the sample training set to generate a trained ESN classifier;
the ESN agent module is used for configuring the ESN classifier of the corresponding network unit based on the trained parameter information of the ESN classifier; extracting the characteristic parameter of each service flow in the corresponding network unit, inputting the characteristic parameter of each service flow into an ESN classifier of the network unit, and obtaining a service classification identification result of each service flow;
and the scheduling module is used for carrying out service flow optimization scheduling according to the service classification identification result of each service flow in the corresponding network unit.
9. A computer device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, characterized in that the processor implements the method of any of claims 1 to 7 when executing the computer program.
10. A computer-readable storage medium, characterized in that the computer-readable storage medium stores a computer program for executing the method of any one of claims 1 to 7.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202011089568.7A CN112422451A (en) | 2020-10-13 | 2020-10-13 | Service sensing method and device of network edge system |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202011089568.7A CN112422451A (en) | 2020-10-13 | 2020-10-13 | Service sensing method and device of network edge system |
Publications (1)
Publication Number | Publication Date |
---|---|
CN112422451A true CN112422451A (en) | 2021-02-26 |
Family
ID=74854701
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202011089568.7A Pending CN112422451A (en) | 2020-10-13 | 2020-10-13 | Service sensing method and device of network edge system |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN112422451A (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN113128955A (en) * | 2021-04-02 | 2021-07-16 | 中国科学院计算技术研究所 | Electronic government affair processing method and system based on intelligent measurement |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104346517A (en) * | 2013-08-02 | 2015-02-11 | 杨凤琴 | Echo state network based prediction method and prediction device |
CN106375136A (en) * | 2016-11-17 | 2017-02-01 | 北京智芯微电子科技有限公司 | Optical access network service flow sensing method and optical access network service flow sensing device |
CN108901036A (en) * | 2018-07-04 | 2018-11-27 | 广东海格怡创科技有限公司 | Method of adjustment, device, computer equipment and the storage medium of subzone network parameter |
CN111542073A (en) * | 2020-04-16 | 2020-08-14 | 全球能源互联网研究院有限公司 | Heterogeneous network selection method and system for power service and network adaptation |
-
2020
- 2020-10-13 CN CN202011089568.7A patent/CN112422451A/en active Pending
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104346517A (en) * | 2013-08-02 | 2015-02-11 | 杨凤琴 | Echo state network based prediction method and prediction device |
CN106375136A (en) * | 2016-11-17 | 2017-02-01 | 北京智芯微电子科技有限公司 | Optical access network service flow sensing method and optical access network service flow sensing device |
CN108901036A (en) * | 2018-07-04 | 2018-11-27 | 广东海格怡创科技有限公司 | Method of adjustment, device, computer equipment and the storage medium of subzone network parameter |
CN111542073A (en) * | 2020-04-16 | 2020-08-14 | 全球能源互联网研究院有限公司 | Heterogeneous network selection method and system for power service and network adaptation |
Non-Patent Citations (2)
Title |
---|
李钟: "基于回声状态网络的电力EPON业务感知技术", 《电力系统保护与控制》 * |
王雨竹等: "基于层次回声状态网络的电力EPON业务流感知", 《计算机测量与控制》 * |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN113128955A (en) * | 2021-04-02 | 2021-07-16 | 中国科学院计算技术研究所 | Electronic government affair processing method and system based on intelligent measurement |
CN113128955B (en) * | 2021-04-02 | 2023-09-12 | 中国科学院计算技术研究所 | Electronic government affair processing method and system based on intelligent measurement |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
US20220391771A1 (en) | Method, apparatus, and computer device and storage medium for distributed training of machine learning model | |
Tang et al. | Joint multiuser DNN partitioning and computational resource allocation for collaborative edge intelligence | |
US20210256403A1 (en) | Recommendation method and apparatus | |
JP7112919B2 (en) | Smart device task processing method and device | |
CN107688493B (en) | Method, device and system for training deep neural network | |
JP2022137182A (en) | Federated learning method, device, equipment and storage medium | |
CN109768940A (en) | The flow allocation method and device of multi-service SDN network | |
CN111176820B (en) | Deep neural network-based edge computing task allocation method and device | |
CN104639466B (en) | A kind of application network Bandwidth Dynamic priority support method based on Storm real-time streams Computational frames | |
Djigal et al. | Machine and deep learning for resource allocation in multi-access edge computing: A survey | |
CN113193984A (en) | Air-space-ground integrated network resource mapping method and system | |
CN112667400B (en) | Edge cloud resource scheduling method, device and system managed and controlled by edge autonomous center | |
Liu et al. | When wireless video streaming meets AI: A deep learning approach | |
Huang et al. | Toward decentralized and collaborative deep learning inference for intelligent iot devices | |
Guo et al. | A delay-sensitive resource allocation algorithm for container cluster in edge computing environment | |
CN112422451A (en) | Service sensing method and device of network edge system | |
Lorido-Botran et al. | ImpalaE: Towards an optimal policy for efficient resource management at the edge | |
AlOrbani et al. | Load balancing and resource allocation in smart cities using reinforcement learning | |
CN113240100B (en) | Parallel computing method and system based on discrete Hopfield neural network | |
Wang et al. | On Jointly Optimizing Partial Offloading and SFC Mapping: A Cooperative Dual-Agent Deep Reinforcement Learning Approach | |
CN114896061B (en) | Training method of computing resource control model, computing resource control method and device | |
CN116367190A (en) | Digital twin function virtualization method for 6G mobile network | |
Xia et al. | Learn to optimize: Adaptive VNF provisioning in mobile edge clouds | |
CN116976461A (en) | Federal learning method, apparatus, device and medium | |
CN110971451A (en) | NFV resource allocation method |
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 | ||
RJ01 | Rejection of invention patent application after publication | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20210226 |