CN110855485A - Method and system for determining network flow of IP backbone network - Google Patents

Method and system for determining network flow of IP backbone network Download PDF

Info

Publication number
CN110855485A
CN110855485A CN201911086587.1A CN201911086587A CN110855485A CN 110855485 A CN110855485 A CN 110855485A CN 201911086587 A CN201911086587 A CN 201911086587A CN 110855485 A CN110855485 A CN 110855485A
Authority
CN
China
Prior art keywords
network
snapshot data
data
deep belief
network model
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
Application number
CN201911086587.1A
Other languages
Chinese (zh)
Inventor
聂来森
吴诒轩
尚润泽
王蕙质
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Qingdao Research Institute Of Northwest Polytechnic University
Original Assignee
Qingdao Research Institute Of Northwest Polytechnic University
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Qingdao Research Institute Of Northwest Polytechnic University filed Critical Qingdao Research Institute Of Northwest Polytechnic University
Priority to CN201911086587.1A priority Critical patent/CN110855485A/en
Publication of CN110855485A publication Critical patent/CN110855485A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L41/00—Arrangements for maintenance, administration or management of data switching networks, e.g. of packet switching networks
    • H04L41/14—Network analysis or design
    • H04L41/145—Network analysis or design involving simulating, designing, planning or modelling of a network
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L41/00—Arrangements for maintenance, administration or management of data switching networks, e.g. of packet switching networks
    • H04L41/14—Network analysis or design
    • H04L41/142—Network analysis or design using statistical or mathematical methods
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L41/00—Arrangements for maintenance, administration or management of data switching networks, e.g. of packet switching networks
    • H04L41/14—Network analysis or design
    • H04L41/147—Network analysis or design for predicting network behaviour

Landscapes

  • Engineering & Computer Science (AREA)
  • Computer Networks & Wireless Communication (AREA)
  • Signal Processing (AREA)
  • Physics & Mathematics (AREA)
  • Algebra (AREA)
  • General Physics & Mathematics (AREA)
  • Mathematical Analysis (AREA)
  • Mathematical Optimization (AREA)
  • Mathematical Physics (AREA)
  • Probability & Statistics with Applications (AREA)
  • Pure & Applied Mathematics (AREA)
  • Data Exchanges In Wide-Area Networks (AREA)

Abstract

The invention discloses a method and a system for determining network flow of an IP backbone network. The method comprises the following steps: constructing a deep belief network model; the deep belief network model is a neural network model which takes snapshot data of link load data as input and takes snapshot data of an end-to-end flow value as output; collecting link load data of an IP backbone network; carrying out normalization processing on the link load data to obtain a priori measurement value; extracting snapshot data in the prior measurement value; inputting snapshot data in the prior measurement value into the deep belief network model, and determining snapshot data of an end-to-end flow value of the IP backbone network; and determining the flow value of the IP backbone network according to the snapshot data of the end-to-end flow value. The method and the system for estimating the network flow of the IP backbone network can improve the accuracy of network flow estimation and have the characteristic of high estimation efficiency.

Description

Method and system for determining network flow of IP backbone network
Technical Field
The invention relates to the technical field of network flow prediction, in particular to a method and a system for determining the network flow of an IP backbone network.
Background
The development of information and communication technology has greatly changed human life and production modes, and services based on Internet technology, such as smart power grids, office automation and the like, have entered people's lives. In addition, the proposal of advanced information concepts and technologies such as smart cities, big data and the like plays a positive promoting role in the development of the future human society. The development of the Internet enables the network scale to increase rapidly, the types of network bearing services are in diversified development, and particularly the rise of cloud computing and the Internet of things enables the network to become a complex heterogeneous network. According to the result of the 37 th statistical report of the development conditions of the Chinese Internet, the scale of Chinese netizens reaches 6.88 hundred million and the total number of newly added netizens is 3951 million in 2015 to 12 months. The Internet popularity reaches 50.3%, and is improved by 2.4% compared with the end of 2014. In order to provide convenient and fast information interaction service for users, network traffic is increased sharply, which makes network management problems increasingly prominent, so that the development of the Internet faces unprecedented challenges. In addition, in order to prevent network congestion and defend network attacks such as ddos (distributed denial of service), a guaranteed service quality is provided for users, and effective network management becomes a key link for maintaining normal operation of a network. When performing network management decisions, a network manager needs to know network operation states, such as time delay, packet loss rate, throughput, bandwidth, network traffic, and the like. The network traffic estimation technology provides a necessary solution and technical support for a network manager to acquire a network state, and can effectively improve the efficiency of network management measures such as network planning, load balancing, IGP (inter Gateway protocol) link weight setting and the like.
In order to improve the accuracy of network traffic estimation, scientists initially model network traffic using linear or non-linear models such as poisson distribution, markov models, gaussian models, AR models, ARIMA models, FARIMA models, MWM models, GARCH models, etc. as additional information to estimate end-to-end network traffic, however, these models only describe the short dependencies of network traffic. But as networks are commercialized, the number of network nodes grows exponentially. Meanwhile, due to the appearance of novel network application, network traffic has multiple statistical characteristics, such as heavy tail distribution, fractal characteristics, autocorrelation characteristics and the like. Therefore, for a complex heterogeneous network, the classical models cannot accurately estimate the network traffic.
Disclosure of Invention
The invention aims to provide a method and a system for determining the network flow of an IP backbone network, which can improve the accuracy of network flow estimation and have the characteristic of high estimation efficiency.
In order to achieve the purpose, the invention provides the following scheme:
a method for determining the network flow of an IP backbone network comprises the following steps:
constructing a deep belief network model; the deep belief network model is a neural network model which takes snapshot data of link load data as input and takes snapshot data of an end-to-end flow value as output;
collecting link load data of an IP backbone network;
carrying out normalization processing on the link load data to obtain a priori measurement value;
extracting snapshot data in the prior measurement value;
inputting snapshot data in the prior measurement value into the deep belief network model, and determining snapshot data of an end-to-end flow value of the IP backbone network;
and determining the flow value of the IP backbone network according to the snapshot data of the end-to-end flow value.
Optionally, before the constructing the deep belief network model, the method further includes:
collecting end-to-end flow value of IP backbone network
Figure BDA0002265595930000021
And link load data
Figure BDA0002265595930000022
The end-to-end flow value is obtained
Figure BDA0002265595930000023
And said link load data
Figure BDA0002265595930000024
By the formula
Figure BDA0002265595930000025
Carrying out normalization processing to obtain the link load dataA priori measurements of
Figure BDA0002265595930000027
And end-to-end flow valueA priori measurements of
Figure BDA0002265595930000029
Extracting the prior measurement
Figure BDA00022655959300000210
Snapshot data of
Figure BDA00022655959300000211
And the prior measurement value
Figure BDA00022655959300000212
Snapshot data of
Figure BDA00022655959300000213
The snapshot data is processed
Figure BDA00022655959300000214
As input, the snapshot data
Figure BDA00022655959300000215
As output, carrying out optimization training on the deep belief network model;
wherein t represents the number of snapshots, L represents the link number, L represents the number of links in the network, N represents the end-to-end traffic number, and N represents the number of end-to-end traffic in the network.
Optionally, the deep belief network model is a limited boltzmann model based on an actual measurement unit.
Optionally, the conditional probability P (v) of the visual layer in the deep belief network modeli| h) is:
Figure BDA0002265595930000031
conditional probability P (h) of hidden layer in the deep belief network model j1| v) is:
Figure BDA0002265595930000032
wherein the content of the first and second substances,
Figure BDA0002265595930000033
is expressed as
Figure BDA0002265595930000034
And the variance is 1, sigmm (·) denotes sigmoid function, v ═ v1,v2,...,vI) Denotes a visible layer unit, h ═ h1,h2,...,hJ) Representing hidden layer elements, I representing the number of visual layer elements, J representing the number of hidden layer elements, biIndicates a visible layer deviation, ajIndicating the hidden layer bias.
An IP backbone network traffic determination system comprising:
the network model building module is used for building a deep belief network model; the deep belief network model is a neural network model which takes snapshot data of link load data as input and takes snapshot data of an end-to-end flow value as output;
the first data acquisition module is used for acquiring link load data of the IP backbone network;
the first prior measurement value acquisition module is used for carrying out normalization processing on the link load data to obtain a prior measurement value;
the first snapshot data extraction module is used for extracting snapshot data in the prior measurement value;
an end-to-end flow determining module, configured to input snapshot data in the first prior measurement value into the deep belief network model, and determine snapshot data of an end-to-end flow value of the IP backbone network;
and the flow value determining module is used for determining the flow value of the IP backbone network according to the snapshot data of the end-to-end flow value.
Optionally, the system further includes:
a second data acquisition module for acquiring end-to-end flow value of the IP backbone network
Figure BDA0002265595930000041
And link load data
A second prior measurement value obtaining module for obtaining the end-to-end flow value
Figure BDA0002265595930000043
And said link load data
Figure BDA0002265595930000044
By the formula
Figure BDA0002265595930000045
Carrying out normalization processing to obtain the link load dataA priori measurements of
Figure BDA0002265595930000047
And end-to-end flow value
Figure BDA0002265595930000048
A priori measurements of
Figure BDA0002265595930000049
A second snapshot data extraction module for extracting the prior measurement value
Figure BDA00022655959300000410
Snapshot data of
Figure BDA00022655959300000411
And the prior measurement value
Figure BDA00022655959300000412
Snapshot data of
Figure BDA00022655959300000413
An optimization training module for matching the snapshot data
Figure BDA00022655959300000414
As input, the snapshot data
Figure BDA00022655959300000415
As output, carrying out optimization training on the deep belief network model;
wherein t represents the number of snapshots, L represents the link number, L represents the number of links in the network, N represents the end-to-end traffic number, and N represents the number of end-to-end traffic in the network.
Optionally, the deep belief network model is a limited boltzmann model based on an actual measurement unit.
Optionally, the conditional probability P (v) of the visual layer in the deep belief network modeli| h) is:
conditional probability P (h) of hidden layer in the deep belief network model j1| v) is:
Figure BDA00022655959300000417
wherein the content of the first and second substances,
Figure BDA00022655959300000418
is expressed as
Figure BDA00022655959300000419
And the variance is 1, sigmm (·) denotes sigmoid function, v ═ v1,v2,...,vI) Denotes a visible layer unit, h ═ h1,h2,...,hJ) Representing hidden layer elements, I representing the number of visual layer elements, J representing the number of hidden layer elements, biIndicates a visible layer deviation, ajIndicating the hidden layer bias.
According to the specific embodiment provided by the invention, the invention discloses the following technical effects: according to the method and the system for estimating the network flow of the IP backbone network, the constructed deep belief network model is adopted, and the snapshot data of the link load data are deeply trained to obtain the end-to-end network flow value of the current network, so that the estimation accuracy of the network flow can be ensured, and the estimation efficiency is improved.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings needed in the embodiments will be briefly described below, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and it is obvious for those skilled in the art to obtain other drawings without creative efforts.
Fig. 1 is a flowchart of a method for estimating network traffic of an IP backbone network according to an embodiment of the present invention;
FIG. 2 is a diagram illustrating a deep belief network architecture according to an embodiment of the present invention;
FIG. 3 is a diagram illustrating the result of performing bias estimation on an Abilene network by using two methods, namely DBN and PCA, according to an embodiment of the present invention;
FIG. 4 is a diagram illustrating the results of estimating the standard deviation and the sampling standard deviation of the Abilene network by using two methods, namely DBN and PCA, according to the embodiment of the present invention;
fig. 5a is a diagram of relative root mean square error results when estimating an Abilene network by using two methods, namely a DBN method and a PCA method, according to an embodiment of the present invention;
fig. 5b is a diagram of cumulative distribution function of relative root mean square error when estimating the Abilene network by using two methods, namely DBN and PCA according to the embodiment of the present invention;
fig. 6 is a schematic structural diagram of an IP backbone network traffic estimation system according to an embodiment of the present invention.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
The invention aims to provide a method and a system for determining the network flow of an IP backbone network, which can improve the accuracy of network flow estimation and have the characteristic of high estimation efficiency.
In order to make the aforementioned objects, features and advantages of the present invention comprehensible, embodiments accompanied with figures are described in further detail below.
The simulation experiment carried out by the invention adopts the Abilene backbone network in the United states and the Abilene backbone network in Europe
Figure BDA0002265595930000061
The backbone network, Abilene backbone network, is mainly used for scientific research and education, and has 12 nodes, 30 internal links, 24 external links, 144 end-to-end flows, and 2016 total moments in simulation data with 5min time intervals. In this embodiment, 2000 time data are collected as a training set, and the traffic matrix may be a matrix of 144 × 2000.
Fig. 1 is a flowchart of a method for estimating network traffic of an IP backbone network according to an embodiment of the present invention, and as shown in fig. 1, a method for determining network traffic of an IP backbone network includes:
s100, constructing a deep belief network model; the deep belief network model is a neural network model which takes snapshot data of link load data as input and takes snapshot data of an end-to-end flow value as output;
s101, collecting link load data of an IP backbone network;
s102, carrying out normalization processing on the link load data to obtain a prior measured value;
s103, extracting snapshot data in the prior measurement value;
s104, inputting snapshot data in the prior measurement value into the deep belief network model, and determining snapshot data of an end-to-end flow value of the IP backbone network;
and S105, determining the flow value of the IP backbone network according to the snapshot data of the end-to-end flow value.
The method for acquiring the end-to-end network flow and the link load data in the S100 specifically comprises the following steps:
step A: acquiring network end-to-end network flow through a NetFlow;
and B: measuring link load data through a simple network management protocol;
the link load data can be obtained by Simple Network Management Protocol (SNMP), and the Abilene backbone network has 54 links, and the link load is a 56 × 2000 matrix.
The normalization processing of the end-to-end network traffic and the link load data described in S101 is specifically as follows:
end-to-end network flow and link load data are normalized according to the following formula:
Figure BDA0002265595930000071
wherein the content of the first and second substances,
Figure BDA0002265595930000072
representing link load data
Figure BDA0002265595930000073
Maximum value of (2).
Figure BDA0002265595930000074
And
Figure BDA0002265595930000075
respectively representing normalized end-to-end network traffic and link load data.
The optimization training process of the adopted deep network belief model comprises the following steps:
will normalize the prior measurements
Figure BDA0002265595930000076
And
Figure BDA0002265595930000077
each snapshot of (expressed as a vector)
Figure BDA0002265595930000078
And
Figure BDA0002265595930000079
) The deep architecture is trained by using the data as input and output respectively. In the invention, a deep belief network architecture facing traffic matrix estimation is shown in fig. 2 and comprises an input layer, an output layer and 12 hidden layers, wherein the input layer comprises 54 input units, the hidden layers and the output layer respectively comprise 100 units and 144 units, and the hidden layers are formed by the deep belief network architecture. In order to ensure the accuracy of the flow matrix estimation, the section adopts an RBM (restricted Boltzmann machine) model based on a real-value unit instead of the traditional binary RBM model.
For training data sets
Figure BDA00022655959300000710
Update of RBM parameters (including weights and offsets)
Figure BDA00022655959300000711
Is performed, and therefore it is described here how to calculate
Figure BDA00022655959300000712
Partial derivatives of, i.e.
Figure BDA00022655959300000713
Where θ represents a parameter of the depth architecture. In the formula (2)
Figure BDA00022655959300000714
Is shown as
From the formula (3), it can be obtained
Figure BDA00022655959300000716
According to the formula (4), there are
Figure BDA0002265595930000081
From equation (5):
Figure BDA0002265595930000082
wherein the content of the first and second substances,from equation (6):
Figure BDA0002265595930000084
for the left term of equation (7), the following derivation applies:
Figure BDA0002265595930000085
at this time, the product can be obtained
Figure BDA0002265595930000086
Similarly, the right term in equation (7) is collated with the following conclusions:
Figure BDA0002265595930000091
at this time, the product can be obtained
Obtained from the formulae (9) and (11)
Therefore, we can calculate by the above formula
Figure BDA0002265595930000094
Partial derivatives of (a). In equation (12), it is very easy to calculate the first term on the right side of the equal sign, however, it needs to traverse when calculating the second term
Figure BDA0002265595930000095
And
Figure BDA0002265595930000096
all possible value results. Thus, the term may be calculated according to the CD algorithm (contextual divide). The CD algorithm samples the visual layer, calculates the hidden layer unit and the visual layer unit through repeated iteration, finally obtains the steady distribution, and calculates the second item according to the hidden layer unit and the visual layer unit. According to the RBM training method and the greedy training method, the depth architecture is trained layer by layer。
After the training is finished, the link load Y at 16 moments is addedtAs input, after forward conduction, the flow matrix estimated value at the corresponding moment can be obtained at the output end
Figure BDA0002265595930000098
Through the steps, an end-to-end network traffic estimation value can be obtained according to the link load data.
Firstly, analyzing the estimation deviation and the sampling standard deviation of a flow matrix estimation method (DBN) and a PCA method based on a deep belief network, wherein the estimation deviation and the sampling standard deviation are respectively defined as follows:
Figure BDA0002265595930000097
Figure BDA0002265595930000101
fig. 3 and 4 show the estimated deviation and the sampling standard deviation of the Abilene network data by the two methods, respectively. Fig. 3 shows the estimated deviation of Abilene network data for two methods, and the x-axis represents the ID of the OD flow, and will be sorted in descending order according to the flow mean. It can be seen that as the OD flow decreases, the estimation deviation gradually decreases, and the DBN method has a smaller estimation deviation. The PCA method exhibits more negative estimation phenomena when estimating small OD flows. Fig. 4 shows the estimated deviation and the sampling standard deviation of the Abilene network data by the two methods, and it can be seen that the DBN method not only has a lower estimated deviation, but also has a smaller sampling standard deviation. From fig. 3-4, it can be seen that the DBN algorithm is both suitable for characterizing long time scale variations of the flow and good at capturing short time scale jitter of the flow, compared to the PCA algorithm.
The ability of the DBN method and the PCA method to capture the time-dependent nature of the end-to-end network traffic is next analyzed. In the section, relative root mean square error is used as error measurement, and two methods are analyzed and compared. The relative root mean square error is defined as:
Figure BDA0002265595930000102
wherein X (n, t) and
Figure BDA0002265595930000103
the true and estimated values of the flow matrix are represented separately. Fig. 5a shows the relative root mean square error of Abilene network data, and in fig. 5(a), the x-axis represents time and the y-axis represents the relative root mean square error. It can be seen that the relative root mean square error of the DBN method is significantly lower than that of the PCA method for other time instants, except that at time 2014, the relative root mean square errors of the DBN method and the PCA method are relatively similar. The relative root mean square error averages for the DBN method and the PCA method were 20% and 27%, respectively. Fig. 5(b) shows the cumulative distribution function of relative root mean square error, and from the simulation results, it is apparent that the DBN method has a lower relative root mean square error.
Therefore, the method and the system for estimating the network flow of the IP backbone network provided by the invention can be obtained, the constructed deep belief network model is adopted, and the end-to-end network flow value of the current network can be obtained by deep training of snapshot data of link load data, so that the estimation accuracy of the network flow can be ensured, and the estimation efficiency is improved.
In addition, the invention also provides a system for determining the network flow of the IP backbone network. As shown in fig. 6, the system includes: the flow rate measurement system comprises a network model construction module 1, a first data acquisition module 2, a first prior measurement value acquisition module 3, a first snapshot data extraction module 4, an end-to-end flow rate determination module 5 and a flow rate value determination module 6.
The network model building module 1 builds a deep belief network model; the deep belief network model is a neural network model which takes snapshot data of link load data as input and takes snapshot data of an end-to-end flow value as output; the first data acquisition module 2 acquires link load data of an IP backbone network; the first prior measurement value acquisition module 3 normalizes the link load data to obtain a prior measurement value; the first snapshot data extraction module 4 extracts snapshot data in the prior measurement value; the end-to-end flow determining module 5 inputs snapshot data in the first prior measurement value into the deep belief network model, and determines snapshot data of an end-to-end flow value of the IP backbone network; and the flow value determining module 6 determines the flow value of the IP backbone network according to the snapshot data of the end-to-end flow value.
In order to optimally train the deep belief network, the system further comprises: the system comprises a second data acquisition module, a second prior measured value acquisition module, a second snapshot data extraction module and an optimization training module.
The second data acquisition module acquires the end-to-end flow value of the IP backbone network
Figure BDA0002265595930000111
And link load data
Figure BDA0002265595930000112
A second prior measurement value acquisition module acquires the end-to-end flow value
Figure BDA0002265595930000113
And said link load data
Figure BDA0002265595930000114
By the formulaCarrying out normalization processing to obtain the link load data
Figure BDA0002265595930000116
A priori measurements of
Figure BDA0002265595930000117
And end-to-end flow valueA priori measurements of
Figure BDA0002265595930000119
A second snapshot data extraction module extracts the prior measurement
Figure BDA00022655959300001110
Snapshot data of
Figure BDA00022655959300001111
And the prior measurement value
Figure BDA00022655959300001112
Snapshot data ofOptimizing the snapshot data by the training module
Figure BDA00022655959300001114
As input, the snapshot data
Figure BDA00022655959300001115
As output, carrying out optimization training on the deep belief network model;
wherein t represents the number of snapshots, L represents the link number, L represents the number of links in the network, N represents the end-to-end traffic number, and N represents the number of end-to-end traffic in the network.
In order to ensure the accuracy of the flow matrix estimation, the deep belief network model is a limited Boltzmann machine model based on a measured unit.
Conditional probability P (v) of visual layer in the deep belief network modeli| h) is:
Figure BDA0002265595930000121
conditional probability P (h) of hidden layer in the deep belief network model j1| v) is:
Figure BDA0002265595930000122
wherein the content of the first and second substances,
Figure BDA0002265595930000123
is expressed as
Figure BDA0002265595930000124
And the variance is 1, sigmm (·) denotes sigmoid function, v ═ v1,v2,...,vI) Denotes a visible layer unit, h ═ h1,h2,...,hJ) Representing hidden layer elements, I representing the number of visual layer elements, J representing the number of hidden layer elements, biIndicates a visible layer deviation, ajIndicating the hidden layer bias.
Since the technical problems solved by the above system and the advantages achieved by the above system are the same as those of the above method, further description is omitted here.
The embodiments in the present description are described in a progressive manner, each embodiment focuses on differences from other embodiments, and the same and similar parts among the embodiments are referred to each other. For the system disclosed by the embodiment, the description is relatively simple because the system corresponds to the method disclosed by the embodiment, and the relevant points can be referred to the method part for description.
The principles and embodiments of the present invention have been described herein using specific examples, which are provided only to help understand the method and the core concept of the present invention; meanwhile, for a person skilled in the art, according to the idea of the present invention, the specific embodiments and the application range may be changed. In view of the above, the present disclosure should not be construed as limiting the invention.

Claims (8)

1. A method for determining the network flow of an IP backbone network is characterized by comprising the following steps:
constructing a deep belief network model; the deep belief network model is a neural network model which takes snapshot data of link load data as input and takes snapshot data of an end-to-end flow value as output;
collecting link load data of an IP backbone network;
carrying out normalization processing on the link load data to obtain a priori measurement value;
extracting snapshot data in the prior measurement value;
inputting snapshot data in the prior measurement value into the deep belief network model, and determining snapshot data of an end-to-end flow value of the IP backbone network;
and determining the flow value of the IP backbone network according to the snapshot data of the end-to-end flow value.
2. The method for determining the network traffic of the IP backbone network according to claim 1, further comprising, before the constructing the deep belief network model:
collecting end-to-end flow value of IP backbone network
Figure FDA0002265595920000011
And link load data
Figure FDA0002265595920000012
The end-to-end flow value is obtained
Figure FDA0002265595920000013
And said link load data
Figure FDA0002265595920000014
By the formula
Figure FDA0002265595920000015
Carrying out normalization processing to obtain the link load dataA priori measurements of
Figure FDA0002265595920000017
And end-to-end flow valueA priori measurements of
Figure FDA0002265595920000019
Extracting the prior measurement
Figure FDA00022655959200000110
Snapshot data of
Figure FDA00022655959200000111
And the prior measurement value
Figure FDA00022655959200000112
Snapshot data of
Figure FDA00022655959200000113
The snapshot data is processed
Figure FDA00022655959200000114
As input, the snapshot data
Figure FDA00022655959200000115
As output, carrying out optimization training on the deep belief network model;
wherein t represents the number of snapshots, L represents the link number, L represents the number of links in the network, N represents the end-to-end traffic number, and N represents the number of end-to-end traffic in the network.
3. The method for determining the network traffic of the IP backbone network according to claim 1, wherein the deep belief network model is a limited boltzmann model based on a measured cell.
4. The method for determining the network traffic of the IP backbone network according to claim 3, wherein the conditional probability P (v) of the visual layer in the deep belief network modeli| h) is:
Figure FDA0002265595920000021
conditional probability P (h) of hidden layer in the deep belief network modelj1| v) is:
Figure FDA0002265595920000022
wherein the content of the first and second substances,
Figure FDA0002265595920000023
is expressed as
Figure FDA0002265595920000024
And the variance is 1, sigmm (·) denotes sigmoid function, v ═ v1,v2,...,vI) Denotes a visible layer unit, h ═ h1,h2,...,hJ) Representing hidden layer elements, I representing the number of visual layer elements, J representing the number of hidden layer elements, biIndicates a visible layer deviation, ajIndicating the hidden layer bias.
5. An IP backbone network traffic determination system, comprising:
the network model building module is used for building a deep belief network model; the deep belief network model is a neural network model which takes snapshot data of link load data as input and takes snapshot data of an end-to-end flow value as output;
the first data acquisition module is used for acquiring link load data of the IP backbone network;
the first prior measurement value acquisition module is used for carrying out normalization processing on the link load data to obtain a prior measurement value;
the first snapshot data extraction module is used for extracting snapshot data in the prior measurement value;
an end-to-end flow determining module, configured to input snapshot data in the first prior measurement value into the deep belief network model, and determine snapshot data of an end-to-end flow value of the IP backbone network;
and the flow value determining module is used for determining the flow value of the IP backbone network according to the snapshot data of the end-to-end flow value.
6. The IP backbone network traffic determination system of claim 5, further comprising:
a second data acquisition module for acquiring end-to-end flow value of the IP backbone network
Figure FDA0002265595920000025
And link load data
Figure FDA0002265595920000031
A second prior measurement value obtaining module for obtaining the end-to-end flow value
Figure FDA0002265595920000032
And said link load data
Figure FDA0002265595920000033
By the formula
Figure FDA0002265595920000034
Carrying out normalization processing to obtain the link load data
Figure FDA0002265595920000035
A priori measurements of
Figure FDA0002265595920000036
And end-to-end trafficValue of
Figure FDA0002265595920000037
A priori measurements of
Figure FDA0002265595920000038
A second snapshot data extraction module for extracting the prior measurement value
Figure FDA0002265595920000039
Snapshot data of
Figure FDA00022655959200000310
And the prior measurement value
Figure FDA00022655959200000311
Snapshot data of
Figure FDA00022655959200000312
An optimization training module for matching the snapshot dataAs input, the snapshot data
Figure FDA00022655959200000314
As output, carrying out optimization training on the deep belief network model;
wherein t represents the number of snapshots, L represents the link number, L represents the number of links in the network, N represents the end-to-end traffic number, and N represents the number of end-to-end traffic in the network.
7. The IP backbone network traffic determination system of claim 5, wherein the deep belief network model is a limited Boltzmann model based on a measured cell.
8. The system according to claim 7, wherein the conditional probability P (v) of the visual layer in the deep belief network model is P (v)i| h) is:
Figure FDA00022655959200000315
conditional probability P (h) of hidden layer in the deep belief network modelj1| v) is:
Figure FDA00022655959200000316
wherein the content of the first and second substances,is expressed as
Figure FDA00022655959200000318
And the variance is 1, sigmm (·) denotes sigmoid function, v ═ v1,v2,...,vI) Denotes a visible layer unit, h ═ h1,h2,...,hJ) Representing hidden layer elements, I representing the number of visual layer elements, J representing the number of hidden layer elements, biIndicates a visible layer deviation, ajIndicating the hidden layer bias.
CN201911086587.1A 2019-11-08 2019-11-08 Method and system for determining network flow of IP backbone network Pending CN110855485A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201911086587.1A CN110855485A (en) 2019-11-08 2019-11-08 Method and system for determining network flow of IP backbone network

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201911086587.1A CN110855485A (en) 2019-11-08 2019-11-08 Method and system for determining network flow of IP backbone network

Publications (1)

Publication Number Publication Date
CN110855485A true CN110855485A (en) 2020-02-28

Family

ID=69599841

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201911086587.1A Pending CN110855485A (en) 2019-11-08 2019-11-08 Method and system for determining network flow of IP backbone network

Country Status (1)

Country Link
CN (1) CN110855485A (en)

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102724078A (en) * 2012-06-29 2012-10-10 蒋定德 End-to-end network flow reconstruction method based on compression sensing in dynamic network
CN105071963A (en) * 2015-08-06 2015-11-18 东北大学 Multi-service traffic estimation method facing big data Internet
US20160180214A1 (en) * 2014-12-19 2016-06-23 Google Inc. Sharp discrepancy learning
CN109816095A (en) * 2019-01-14 2019-05-28 湖南大学 Based on the network flow prediction method for improving gating cycle neural network

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102724078A (en) * 2012-06-29 2012-10-10 蒋定德 End-to-end network flow reconstruction method based on compression sensing in dynamic network
US20160180214A1 (en) * 2014-12-19 2016-06-23 Google Inc. Sharp discrepancy learning
CN105071963A (en) * 2015-08-06 2015-11-18 东北大学 Multi-service traffic estimation method facing big data Internet
CN109816095A (en) * 2019-01-14 2019-05-28 湖南大学 Based on the network flow prediction method for improving gating cycle neural network

Non-Patent Citations (3)

* Cited by examiner, † Cited by third party
Title
LAISEN NIE, XIAOJIE WANG, LIANGTIAN WAN, SHUI YU, HOUBING SONG: "Network Traffic Prediction Based on Deep Belief Network and Spatiotemporal Compressive Sensing in Wireless Mesh Backbone Networks", 《WIRELESS COMMUNICATIONS AND MOBILE COMPUTING》 *
LAISEN NIE; DINGDE JIANG; LEI GUO; SHUI YU; HOUBING SONG: "Traffic Matrix Prediction and Estimation Based on Deep Learning for Data Center Networks", 《 2016 IEEE GLOBECOM WORKSHOPS (GC WKSHPS)》 *
聂来森: "基于压缩感知的端到端网络流量重构策略研究", 《中国博士学位论文全文数据库》 *

Similar Documents

Publication Publication Date Title
CN107483251B (en) Network service abnormity detection method based on distributed probe monitoring
Ribeiro et al. Estimating and sampling graphs with multidimensional random walks
CN111159425A (en) Temporal knowledge graph representation method based on historical relationship and double-graph convolution network
CN107705212B (en) Role identification method based on particle swarm random walk
Zhang et al. LNTP: An end-to-end online prediction model for network traffic
CN111431819A (en) Network traffic classification method and device based on serialized protocol flow characteristics
CN106126607B (en) User relationship analysis method facing social network
Yin et al. Network traffic classification via HMM under the guidance of syntactic structure
Toral-Cruz et al. Accurate modeling of VoIP traffic QoS parameters in current and future networks with multifractal and Markov models
Wang et al. A multitask learning-based network traffic prediction approach for SDN-enabled industrial internet of things
Tian et al. A network traffic prediction method based on IFS algorithm optimised LSSVM
CN112134873B (en) IoT network abnormal flow real-time detection method and system
CN108270608B (en) Link prediction model establishment and link prediction method
Zhang et al. Intelligent and application-aware network traffic prediction in smart access gateways
CN110855485A (en) Method and system for determining network flow of IP backbone network
Memon et al. Design and implementation of a robust convolutional neural network-based traffic matrix estimator for cloud networks
CN110839253A (en) Method for determining wireless grid network flow
CN113379092B (en) Backbone network multi-service traffic estimation method and system facing big data
Nie et al. A reconstructing approach to end‐to‐end network traffic based on multifractal wavelet model
Lei et al. Optimizing traffic classification using hybrid feature selection
Jiang et al. Network traffic anomaly detection based on sliding window
Mirtchev et al. New constant service time Polya/D/n traffic model with peaked input stream
Deri et al. Graph sampling: Estimation of degree distributions
Yang et al. An accurate approach for traffic matrix estimation in large-scale backbone networks
CN114979017B (en) Deep learning protocol identification method and system based on original flow of industrial control system

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

Application publication date: 20200228

RJ01 Rejection of invention patent application after publication