CN111526070A - Service function chain fault detection method based on prediction - Google Patents
Service function chain fault detection method based on prediction Download PDFInfo
- Publication number
- CN111526070A CN111526070A CN202010359298.0A CN202010359298A CN111526070A CN 111526070 A CN111526070 A CN 111526070A CN 202010359298 A CN202010359298 A CN 202010359298A CN 111526070 A CN111526070 A CN 111526070A
- Authority
- CN
- China
- Prior art keywords
- network
- service function
- function chain
- data
- 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.)
- Granted
Links
- 238000001514 detection method Methods 0.000 title claims abstract description 39
- 230000006870 function Effects 0.000 claims abstract description 76
- 238000000034 method Methods 0.000 claims abstract description 30
- 238000012544 monitoring process Methods 0.000 claims abstract description 19
- 230000005012 migration Effects 0.000 claims abstract description 12
- 238000013508 migration Methods 0.000 claims abstract description 12
- 230000036541 health Effects 0.000 claims abstract description 6
- 238000012549 training Methods 0.000 claims description 23
- 230000008569 process Effects 0.000 claims description 9
- 238000012545 processing Methods 0.000 claims description 8
- 230000002829 reductive effect Effects 0.000 claims description 7
- ORILYTVJVMAKLC-UHFFFAOYSA-N Adamantane Natural products C1C(C2)CC3CC1CC2C3 ORILYTVJVMAKLC-UHFFFAOYSA-N 0.000 claims description 6
- 230000009286 beneficial effect Effects 0.000 claims description 5
- 238000012937 correction Methods 0.000 claims description 4
- 230000000694 effects Effects 0.000 claims description 4
- 239000011159 matrix material Substances 0.000 claims description 4
- 230000009467 reduction Effects 0.000 claims description 4
- 238000004458 analytical method Methods 0.000 claims description 3
- 230000035876 healing Effects 0.000 claims description 3
- 230000002441 reversible effect Effects 0.000 claims description 3
- 238000009499 grossing Methods 0.000 claims description 2
- 230000036961 partial effect Effects 0.000 claims description 2
- 230000000306 recurrent effect Effects 0.000 claims description 2
- 238000012360 testing method Methods 0.000 claims description 2
- 238000004891 communication Methods 0.000 abstract description 5
- 230000008901 benefit Effects 0.000 description 5
- 230000003247 decreasing effect Effects 0.000 description 4
- 230000035772 mutation Effects 0.000 description 4
- 230000008859 change Effects 0.000 description 3
- 238000006243 chemical reaction Methods 0.000 description 3
- 238000010586 diagram Methods 0.000 description 3
- 238000005516 engineering process Methods 0.000 description 2
- 230000007246 mechanism Effects 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 238000010606 normalization Methods 0.000 description 2
- 238000007781 pre-processing Methods 0.000 description 2
- 238000013526 transfer learning Methods 0.000 description 2
- 230000007704 transition Effects 0.000 description 2
- 206010033799 Paralysis Diseases 0.000 description 1
- 230000002159 abnormal effect Effects 0.000 description 1
- 230000005856 abnormality Effects 0.000 description 1
- 230000002411 adverse Effects 0.000 description 1
- 238000013528 artificial neural network Methods 0.000 description 1
- 230000033228 biological regulation Effects 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 230000007423 decrease Effects 0.000 description 1
- 230000003862 health status Effects 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 230000001737 promoting effect Effects 0.000 description 1
- 238000011084 recovery Methods 0.000 description 1
- 230000007480 spreading Effects 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
Images
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L43/00—Arrangements for monitoring or testing data switching networks
- H04L43/08—Monitoring or testing based on specific metrics, e.g. QoS, energy consumption or environmental parameters
- H04L43/0805—Monitoring or testing based on specific metrics, e.g. QoS, energy consumption or environmental parameters by checking availability
- H04L43/0817—Monitoring or testing based on specific metrics, e.g. QoS, energy consumption or environmental parameters by checking availability by checking functioning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F9/00—Arrangements for program control, e.g. control units
- G06F9/06—Arrangements for program control, e.g. control units using stored programs, i.e. using an internal store of processing equipment to receive or retain programs
- G06F9/44—Arrangements for executing specific programs
- G06F9/455—Emulation; Interpretation; Software simulation, e.g. virtualisation or emulation of application or operating system execution engines
- G06F9/45533—Hypervisors; Virtual machine monitors
- G06F9/45558—Hypervisor-specific management and integration aspects
-
- 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
- H04L41/08—Configuration management of networks or network elements
- H04L41/0893—Assignment of logical groups to network elements
-
- 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
- H04L41/50—Network service management, e.g. ensuring proper service fulfilment according to agreements
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L43/00—Arrangements for monitoring or testing data switching networks
- H04L43/50—Testing arrangements
- H04L43/55—Testing of service level quality, e.g. simulating service usage
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F9/00—Arrangements for program control, e.g. control units
- G06F9/06—Arrangements for program control, e.g. control units using stored programs, i.e. using an internal store of processing equipment to receive or retain programs
- G06F9/44—Arrangements for executing specific programs
- G06F9/455—Emulation; Interpretation; Software simulation, e.g. virtualisation or emulation of application or operating system execution engines
- G06F9/45533—Hypervisors; Virtual machine monitors
- G06F9/45558—Hypervisor-specific management and integration aspects
- G06F2009/45591—Monitoring or debugging support
Landscapes
- Engineering & Computer Science (AREA)
- Computer Networks & Wireless Communication (AREA)
- Signal Processing (AREA)
- Software Systems (AREA)
- Theoretical Computer Science (AREA)
- Quality & Reliability (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Environmental & Geological Engineering (AREA)
- Data Exchanges In Wide-Area Networks (AREA)
Abstract
The invention relates to a service function chain fault detection method based on prediction, and belongs to the technical field of communication. According to the method, firstly, according to the performance correlation existing among the virtual network functions of a service function chain under a virtual network environment, data are collected in a mode of monitoring the performance data of each virtual network function on the whole service function chain, and the working state of the data is divided; secondly, considering the high-dimensional complexity and time-related characteristics of network monitoring data, combining the initiative requirement of fault detection, adopting a gated cycle unit network to detect faults, and predicting the health condition of the network by analyzing the historical performance data information of a service function chain; and migration learning is introduced to accelerate the convergence speed of the model by utilizing the similarity of the virtual network function nodes among different service function chains. The invention can effectively detect the occurrence of the fault while meeting the requirement of fault detection time.
Description
Technical Field
The invention belongs to the technical field of communication, and relates to a service function chain fault detection method based on prediction.
Background
The arrival of 5G will bring a revolution across the era, and the 5G network will provide differentiated business scenes and customized network services for the colorful service demands of people. The existing network architecture is mainly oriented to human-human communication, and is difficult to support emerging services such as Internet of things and Internet of vehicles and the like which need to realize large-scale machine communication. Therefore, 5G considers network slicing as an ideal network architecture as a key to solving the above problems. The 5G network slice architecture based on the SDN and the NFV can be flexibly arranged according to user requirements, and resources can be dynamically and efficiently allocated. The NFV technology decouples software and hardware of the traditional network, realizes network functions in a software mode, and is more beneficial to realizing network resource sharing.
However, the network slice architecture brings great flexibility to the 5G network, and also puts new requirements on the reliability of the network. Due to the sharing of the underlying resources, a failure of an underlying Network node easily causes a failure of multiple Virtual Network Functions (VNFs) sharing the node resource, resulting in functional paralysis of multiple Service Function Chains (SFCs). Therefore, it is very important to realize fast recovery of a network from a fault by using the SON technology, where fault detection is used as a main body of network performance analysis and is a primary premise for realizing a self-healing measure.
The inventor finds that the following disadvantages exist in the process of researching the prior art:
most of the existing fault detection methods are based on a reaction mechanism, but the mechanism is difficult to avoid the inherent delay of the reaction and is not beneficial to the timeliness requirement of the network; most methods do not consider the condition of network service quality reduction, and neglect the adverse effect of continuous performance reduction on the network; the failure in the service function chain scenario is not considered to have the characteristics of longitudinal propagation between an infrastructure layer and an application layer and transverse propagation between different service function chains, which will cause the performance abnormality of some normal VNF nodes, and it is difficult to find the VNF node position where the failure first occurs.
Disclosure of Invention
In view of the above, an object of the present invention is to provide a service function chain fault detection method based on prediction, which can effectively detect a node abnormal condition according to a change of performance data of a VNF node in a network virtualization environment, and meet a reliability requirement of a network.
In order to achieve the purpose, the invention provides the following technical scheme:
a service function chain fault detection method based on prediction is characterized by specifically comprising the following steps:
s1: collecting data by monitoring performance data of each VNF on the whole service function chain according to performance correlation existing among VNF nodes by combining fault propagation characteristics under a service function chain scene, and dividing the working state of the data into normal state, service quality reduction or fault state;
s2: aiming at the high-dimensional complexity and time-related characteristics of network monitoring data, combining the initiative requirement of fault detection, adopting a Gated Recurrent Unit (GRU) network to detect faults, and predicting the health condition of the network by analyzing the historical performance data information of a service function chain; aiming at the problem that modeling of a GRU network requires a long time and is not beneficial to the real-time requirement of fault detection, migration learning is introduced to accelerate the convergence speed of the model by utilizing the similarity of VNF nodes among different service function chains.
Further, in the step S1, in the service function chain scenario, the service function chain is generally formed by arranging a plurality of VNFs independent of each other according to a specific requirement of a user in a certain order, VNFs in different service function chains may partially overlap, and such a characteristic of the service function chain is very likely to cause a VNF fault to propagate to an associated VNF, thereby causing a slice network large-area fault.
Considering that the fault has the characteristic of spreading among different VNF nodes in the service function chain, the performance data is collected in a mode of monitoring the working state of each VNF node of the whole service function chain at the application layer of the slice network, the occurrence of the fault is detected through the continuous analysis of the performance data by the detection system, and the first VNF node showing the fault is taken as the starting point of the occurrence of the fault.
Further, in order to eliminate the influence of the problems of different dimensions, different data value ranges, unobvious data change trend and the like in the original data sample on model training and improve the precision of the model and the network training speed, a linear maximum and minimum method is required to be adopted to carry out normalization preprocessing on the performance data collected from all VNF nodes, and the conversion function is as follows:
further, in step S1, the operating status of the service function chain is divided into three categories according to the cause of the fault:
(1) and (3) normal: the network state runs well;
(2) the service quality is reduced: the situation that network load is increased, flow is reduced, time delay is increased or packet loss occurs, but the VNF can still work; the specific reasons may be sudden change of the surrounding environment, software and hardware problems or insufficient resources, and the situation that the system is recovered to a normal working state in a short time or continuously deteriorated and becomes a fault may occur;
(3) and (4) failure: the network function can not be used at all, the time delay becomes infinite, and corresponding service can not be provided for the user any more; in order to ensure the normal operation of the SFC where the VNF node is located, operations such as node migration or software and hardware device restart need to be performed.
Further, the rule for executing the node migration is as follows: the failed VNF node may be generated by a mutation of a normal VNF node, or may be generated by a mutation of a VNF node in the process of decreasing the service quality, which is degraded to a certain extent or in the process of decreasing the performance. For VNF nodes that are in a reduced quality of service, if they can adapt to the optimal regulation of the system, they revert to normal operation or deteriorate into a failure, and the restoration of network functions needs to be achieved through necessary healing measures.
Further, in step S2, for the high-dimensional complexity and time-related characteristics of the network monitoring data, and in combination with the initiative requirement of fault detection, the method for detecting the fault by using the GRU network specifically includes the following steps:
s201: processing input time sequence sample data by using a three-layer GRU unit, and training a model in a small batch mode by using a historical monitoring data set;
s202: after passing through the three GRU networks, the characteristic information of the previous network is integrated through a full connection layer, so that the learning capability of the network is improved;
s203: the output of the full connection layer is used as the input of a softmax classifier, and reverse supervised fine tuning is carried out by combining tag data;
s204: the parameters are further optimized using real-time monitoring data.
Further, in step S2, predicting the health condition of the network by analyzing the historical performance data information of the service function chain specifically includes:
the historical performance data is assumed to be waiting time delay and processing time delay; in a training stage, firstly, feature acquisition is carried out on waiting time delay and processing time delay of all VNF nodes on a service function chain, a certain service function chain is set to be composed of m VNF nodes, monitoring data of all VNF nodes at each moment are recorded, the length of a sliding window is defined as d, and then in a time range from t-d to t, an input data set of a network model is represented as x ═ x { (x-x) } xt,xt-1,…,xt-d+1At time t, the data set of all VNF nodes is:
wherein,
and
respectively representing the waiting delay and the processing delay of the mth VNF;
the prediction method comprises the following steps: since the input of the GRU network model is time-series data, it is necessary to construct time-series samples by a sliding window method, and to use the length d as the size of a sliding window according to the length dThe time step h moves on the data set to obtain the sample X at the current momentt＝{xt,xt-1,…,xt-d+1And sample X at the next time instantt+h＝{xt+h,xt+h-1,…,xt+h-d+1And determining the label value as x according to the network actual state at the next momentt+1And xt+h+1；
Then, according to the dimension of GRU input, dividing a training set and a test set according to a certain proportion, training a model in a small batch mode, integrating the characteristic information of the previous network through a full connection layer after passing through a three-layer GRU network, improving the learning capacity of the network, and finally taking the output of the full connection layer as the input of a softmax classifier to obtain a final prediction result; in order to prevent overfitting when the network is trained, partial repeated information generated in the training process is discarded in a Dropout regularization mode;
training of a network model: the method is a process for continuously optimizing the model parameters, and in order to reduce the error between an output result and a real network, a back propagation algorithm is used for carrying out iterative updating on the network parameters; optimizing the network weight layer by layer in a gradient descending mode to enable the value of the target loss function to be minimum; compared with other parameter optimization algorithms, the Adam algorithm has the advantages of calculation efficiency, convergence speed and the like, so that the Adam algorithm with the self-adaptive learning rate is adopted to accelerate the convergence speed of the algorithm.
Further, the Adam algorithm dynamically adjusts the learning rate of each parameter by using the distance estimation of the gradient, is suitable for a large data set and a high-dimensional space, and is more modern as follows:
wherein theta is an iteration parameter, is a learning rate,
and
the bias correction for the first order estimate and the bias correction for the second order estimate of the gradient, respectively, are a smoothing term.
Further, in step S2, the introducing of transfer learning to accelerate the convergence rate of the model specifically includes: in a network slicing scene, the size of a monitoring data set is limited by the length of slicing operation time, and for a service function chain in the early stage of slicing operation, the condition that fault detection accuracy is not high due to insufficient monitoring data may occur, and a parameter-based transfer learning method needs to be introduced to accelerate the convergence speed of a model, so that a fault detection model based on GRU neural network prediction can keep higher detection accuracy under the condition of a smaller data sample.
The selected source domain model parameters are trained from service function chain data with similar performance index requirements to the target domain; then, migrating the fault detection model parameters in other service function chains similar to the current service function chain structure to the current service function chain to help the current service function chain fault detection model to obtain better training effect; the method comprises the following specific steps: utilizing a historical data sample set of a service function chain SFC b to carry out fault detection model training of a GRU network of the SFCb, obtaining an optimal parameter matrix of model convergence, and taking W as the referencei bFor example, let the migration ratio phi (t) ∈ (0, 1) represent the degree of migration of a parameter from the SFC b model to the SFC a model, where phi (t) is 1/t and decreases as time t increases, let it be known that the parameter matrix of SFC a is Wi a＝φ(t)Wi b+(1-φ(t))Wi aParameter W at initial timei a＝Wi bTraining and fine tuning the model by using the data sample set of the SFC a to obtain the optimal GRU network model parameter W of the SFC ai a'。
The invention has the beneficial effects that: aiming at the problem of fault detection in a 5G end-to-end network slicing scene, the invention can effectively extract massive and high-dimensional data characteristics in a complex network on the basis of meeting the requirement of the system on detection accuracy, simultaneously ensures the timeliness of fault detection and has high application value in a wireless communication system.
Additional advantages, objects, and features of the invention will be set forth in part in the description which follows and in part will become apparent to those having ordinary skill in the art upon examination of the following or may be learned from practice of the invention. The objectives and other advantages of the invention may be realized and attained by the means of the instrumentalities and combinations particularly pointed out hereinafter.
Drawings
For the purposes of promoting a better understanding of the objects, aspects and advantages of the invention, reference will now be made to the following detailed description taken in conjunction with the accompanying drawings in which:
FIG. 1 is a schematic diagram of a scenario in which the present invention may be applied;
FIG. 2 is a node state transition diagram according to an embodiment of the present invention;
FIG. 3 is a fault detection model based on a GRU network in an embodiment of the present invention;
fig. 4 is a flowchart illustrating a service function chain fault detection method according to an embodiment of the present invention.
Detailed Description
The embodiments of the present invention are described below with reference to specific embodiments, and other advantages and effects of the present invention will be easily understood by those skilled in the art from the disclosure of the present specification. The invention is capable of other and different embodiments and of being practiced or of being carried out in various ways, and its several details are capable of modification in various respects, all without departing from the spirit and scope of the present invention. It should be noted that the drawings provided in the following embodiments are only for illustrating the basic idea of the present invention in a schematic way, and the features in the following embodiments and examples may be combined with each other without conflict.
Referring to fig. 1 to 4, fig. 1 is a schematic view of a scenario in which the present invention can be implemented. Referring to fig. 1, the application layer is mainly responsible for providing an ordered set of VNFs for each sliced request to process traffic, and the infrastructure layer provides physical nodes and links of many types of resources, such as computational resources, bandwidth resources, storage resources, etc., that support various sliced network functional requirements. The virtualization layer realizes functions of slice life cycle management, network performance data monitoring and the like through the NFV MANO and the SDN controller. The system generates a specific service function chain according to different service requests, thereby meeting the service requirements of users. The traffic, bandwidth, latency and other requirements of different VNFs in each service function chain are also different, but there is a certain correlation between the resources required by two adjacent VNFs and the virtual link connecting them. In order to ensure the stability and the service quality of the network, the node states of different VNFs of the whole service function chain need to be monitored, and the occurrence of a fault is detected in time.
FIG. 2 is a node state transition diagram in the present embodiment. The failed VNF node may be generated by a mutation of a normal VNF node, or may be generated by a mutation of a VNF node in the process of decreasing the service quality, which is degraded to a certain extent or in the process of decreasing the performance. For VNF nodes that are in a reduced quality of service, if the optimal adjustment of the system can be adapted, it may be restored to a normal operating state, otherwise, it deteriorates to a failure, and the restoration of the network function needs to be achieved through necessary healing measures.
Fig. 3 is a fault detection model based on a GRU network in an embodiment of the present invention. Referring to fig. 3, the fault detection model based on GRU network prediction of the present invention is composed of three layers of GRU units, a full connectivity layer and a softmax classifier. Defining each input sample at time t as the health status feature information x of all VNF nodes on the same service function chaintAnd the hidden layer state h of the model at the moment can be obtained by training the GRU network modeltAnd then derive the predicted state y of each VNFt+1. Since the state of the VNF at the next time is affected by the observed data at that time and the state of the hidden layer at the previous time, the GRU network can use the historical observed data for the health state prediction of future VNF nodes.
Fig. 4 is a flowchart illustrating a service function chain fault detection method according to an embodiment of the present invention. The method comprises the following specific steps:
step 401: initializing system parameters, wherein the GRU network model parameters including SFC b and SFC a, the learning rate and the iteration times k are 0;
step 402: inputting historical performance data of all VNF nodes of the SFC b and real-time performance data of all VNF nodes of the SFC a before t moment;
step 403: carrying out normalization preprocessing on input data;
step 404: constructing time sequence input data according to the length of the sliding window and the time step;
step 405: training a GRU network model by using historical performance data of all VNF nodes of the SFC b, carrying out reverse fine adjustment on the model by using an Adam algorithm of a self-adaptive learning rate, and updating corresponding parameters;
step 406: and judging whether the convergence condition is met. If the convergence condition is not satisfied, let k be k +1, continue to execute step 405, otherwise, extract the latest model parameter and execute step 407;
step 407: transferring model parameters extracted from the GRU network model of the SFC b to the GRU network model of the SFC a, and further training the model by utilizing real-time performance data before t moment in the SFC a;
step 408: and judging whether the maximum iteration number K is reached. If not, let k be k +1, go to step 407, otherwise execute step 409;
step 409: and outputting the working state of each VNF node in the SFC a at the moment of t + 1.
Finally, the above embodiments are only intended to illustrate the technical solutions of the present invention and not to limit the present invention, and although the present invention has been described in detail with reference to the preferred embodiments, it will be understood by those skilled in the art that modifications or equivalent substitutions may be made on the technical solutions of the present invention without departing from the spirit and scope of the technical solutions, and all of them should be covered by the claims of the present invention.
Claims (9)
1. A service function chain fault detection method based on prediction is characterized by specifically comprising the following steps:
s1: collecting data by monitoring performance data of each Virtual Network Function (VNF) on the whole service function chain according to performance correlation existing between VNF nodes by combining fault propagation characteristics under a service function chain scene, and dividing the working state of the VNF into normal state, service quality reduction or fault state;
s2: aiming at the high-dimensional complexity and time-related characteristics of network monitoring data, combining the initiative requirement of fault detection, adopting a Gated Recurrent Unit (GRU) network to detect faults, and predicting the health condition of the network by analyzing the historical performance data information of a service function chain; aiming at the problem that modeling of a GRU network requires a long time and is not beneficial to the real-time requirement of fault detection, migration learning is introduced to accelerate the convergence speed of the model by utilizing the similarity of VNF nodes among different service function chains.
2. The method according to claim 1, wherein in the step S1, in the service function chain scenario, the service function chain is formed by arranging a plurality of VNFs independent of functions according to a specific requirement of a user in a certain order; the method comprises the steps of collecting performance data by monitoring the working state of each VNF node of the whole service function chain on the application layer of the slice network, detecting the occurrence of faults through continuous analysis of the performance data by a detection system, and taking the first VNF node showing the faults as the starting point of the occurrence of the faults.
3. The prediction-based service function chain fault detection method of claim 2, wherein the performance data collected from all VNF nodes is normalized and preprocessed by a linear maximum-minimum method.
4. The method according to claim 1, wherein in step S1, the operation status of the service function chain is divided into three categories according to the cause of the failure:
(1) and (3) normal: the network state runs well;
(2) the service quality is reduced: network load is increased, flow is reduced, time delay is increased or packet loss occurs, but the VNF can still work;
(3) and (4) failure: the network function can not be used at all, the time delay becomes infinite, and corresponding service can not be provided for the user any more; and carrying out node migration or software and hardware equipment restart.
5. The prediction-based service function chain fault detection method according to claim 4, wherein the rule for performing node migration is: for VNF nodes in a reduced quality of service, if the optimal adjustment of the system can be adapted, the VNF nodes are restored to a normal operating state or deteriorated to a failure, and the restoration of the network function needs to be achieved through a healing measure.
6. The method according to claim 1, wherein in the step S2, for the high-dimensional complexity and time-related characteristics of the network monitoring data, the method for detecting the fault by using the GRU network in combination with the initiative requirement of fault detection specifically includes the following steps:
s201: processing input time sequence sample data by using a three-layer GRU unit, and training a model in a small batch mode by using a historical monitoring data set;
s202: after passing through the three GRU networks, the characteristic information of the previous network is integrated through a full connection layer, so that the learning capability of the network is improved;
s203: the output of the full connection layer is used as the input of a softmax classifier, and reverse supervised fine tuning is carried out by combining tag data;
s204: the parameters are further optimized using real-time monitoring data.
7. The method according to claim 6, wherein the step S2 of predicting the health condition of the network by analyzing the historical performance data information of the service function chain specifically comprises:
the historical performance data is assumed to be waiting time delay and processing time delay; in trainingIn the training stage, firstly, feature acquisition is carried out on the waiting time delay and the processing time delay of all VNF nodes on a service function chain, a certain service function chain is set to be composed of m VNF nodes, monitoring data of all VNF nodes at each moment are recorded, the length of a sliding window is defined to be d, and then in the time range from t-d to t, the input data set of the network model is represented as x ═ x { (x-x)t,xt-1,…,xt-d+1At time t, the data set of all VNF nodes is:
wherein,
and
respectively representing the waiting delay and the processing delay of the mth VNF;
the prediction method comprises the following steps: constructing a time sequence sample by a sliding window method, taking the length d as the size of a sliding window, moving on a data set according to a time step h, and obtaining a sample X at the current momentt＝{xt,xt-1,…,xt-d+1And sample X at the next time instantt+h＝{xt+h,xt+h-1,…,xt+h-d+1And determining the label value as x according to the network actual state at the next momentt+1And xt+h+1；
Then, according to the dimension of GRU input, dividing a training set and a test set according to a certain proportion, training a model in a small batch mode, integrating the characteristic information of the previous network through a full connection layer after passing through a three-layer GRU network, improving the learning capacity of the network, and finally taking the output of the full connection layer as the input of a softmax classifier to obtain a final prediction result; in order to prevent overfitting when the network is trained, partial repeated information generated in the training process is discarded in a Dropout regularization mode;
training of a network model: iteratively updating the network parameters by using a back propagation algorithm; optimizing the network weight layer by layer in a gradient descending mode to enable the value of the target loss function to be minimum; and the Adam algorithm of the self-adaptive learning rate is adopted to accelerate the convergence speed of the algorithm.
8. The prediction-based service function chain fault detection method according to claim 7, wherein the Adam algorithm dynamically adjusts the learning rate of each parameter by using the distance estimation of the gradient, and the update formula is as follows:
wherein theta is an iteration parameter, is a learning rate,
and
the bias correction for the first order estimate and the bias correction for the second order estimate of the gradient, respectively, are a smoothing term.
9. The method according to claim 1, wherein in the step S2, migration learning is introduced to accelerate the convergence speed of the model, and specifically includes: the selected source domain model parameters are trained from service function chain data with similar performance index requirements to the target domain; then, migrating the fault detection model parameters in other service function chains similar to the current service function chain structure to the current service function chain to help the current service function chain fault detection model to obtain better training effect; the method comprises the following specific steps: utilizing a historical data sample set of a Service Function Chain (SFC) b to carry out fault detection model training of a GRU network of the SFC b, and obtaining an optimal parameter matrix W of model convergencei bLet the migration ratio φ (t) ∈ (0, 1) denote the degree of migration of a parameter from the SFC b model to the SFCa model,wherein phi (t) is 1/t; SFCa has a parameter matrix of Wi a＝φ(t)Wi b+(1-φ(t))Wi aParameter W at initial timei a＝Wi bTraining and fine tuning the model by using the data sample set of SFCa to obtain the optimal GRU network model parameter W of SFCai a'。
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010359298.0A CN111526070B (en) | 2020-04-29 | 2020-04-29 | Service function chain fault detection method based on prediction |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010359298.0A CN111526070B (en) | 2020-04-29 | 2020-04-29 | Service function chain fault detection method based on prediction |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN111526070A true CN111526070A (en) | 2020-08-11 |
| CN111526070B CN111526070B (en) | 2022-06-03 |
Family
ID=71906161
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010359298.0A Active CN111526070B (en) | 2020-04-29 | 2020-04-29 | Service function chain fault detection method based on prediction |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111526070B (en) |
Cited By (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111932376A (en) * | 2020-08-14 | 2020-11-13 | 腾讯科技（深圳）有限公司 | Virtual resource data simulation method and device |
| CN112291804A (en) * | 2020-10-22 | 2021-01-29 | 深圳供电局有限公司 | Service fault diagnosis method for noise network under 5G network slice |
| CN112887145A (en) * | 2021-01-27 | 2021-06-01 | 重庆邮电大学 | Distributed network slice fault detection method |
| CN113490254A (en) * | 2021-08-11 | 2021-10-08 | 重庆邮电大学 | VNF migration method based on bidirectional GRU resource demand prediction in federal learning |
| CN113608830A (en) * | 2021-07-13 | 2021-11-05 | 全球能源互联网研究院有限公司 | VNF migration method and device based on fault prediction |
| CN113805561A (en) * | 2021-08-16 | 2021-12-17 | 天津大学 | Spacecraft intelligent fault diagnosis method based on deep migration learning |
| CN114423035A (en) * | 2022-01-12 | 2022-04-29 | 重庆邮电大学 | Service function chain abnormity detection method under network slice scene |
| CN114785672A (en) * | 2022-03-31 | 2022-07-22 | 北京航空航天大学 | Cloud network service fault prediction method based on fault marker |
| WO2023011618A1 (en) * | 2021-08-06 | 2023-02-09 | International Business Machines Corporation | Predicting root cause of alert using recurrent neural network |
Citations (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20170126792A1 (en) * | 2015-11-02 | 2017-05-04 | Telefonaktiebolaget L M Ericsson (Publ) | System and methods for intelligent service function placement and autoscale based on machine learning |
| CN107332913A (en) * | 2017-07-04 | 2017-11-07 | 电子科技大学 | A kind of Optimization deployment method of service function chain in 5G mobile networks |
| CN107819662A (en) * | 2017-11-23 | 2018-03-20 | 电子科技大学 | A kind of service function chain real-time migration method applied to cloud and mist calculating network |
| CN108900358A (en) * | 2018-08-01 | 2018-11-27 | 重庆邮电大学 | Virtual network function dynamic migration method based on deepness belief network resource requirement prediction |
| US20180349202A1 (en) * | 2017-05-30 | 2018-12-06 | Hewlett Packard Enterprise Development Lp | Virtual Network Function Resource Allocation |
| CN108965024A (en) * | 2018-08-01 | 2018-12-07 | 重庆邮电大学 | A kind of virtual network function dispatching method of the 5G network slice based on prediction |
| TWI650979B (en) * | 2017-07-25 | 2019-02-11 | 中華電信股份有限公司 | Load balance adjustment system and method thereof |
| CN109711620A (en) * | 2018-12-26 | 2019-05-03 | 浙江大学 | A kind of Short-Term Load Forecasting Method based on GRU neural network and transfer learning |
| CN110335168A (en) * | 2019-04-22 | 2019-10-15 | 山东大学 | Method and system based on GRU optimization power information acquisition terminal fault prediction model |
| CN110737732A (en) * | 2019-10-25 | 2020-01-31 | 广西交通科学研究院有限公司 | electromechanical equipment fault early warning method |
| CN110766212A (en) * | 2019-10-15 | 2020-02-07 | 哈尔滨工程大学 | Ultra-short-term photovoltaic power prediction method for historical data missing electric field |
-
2020
- 2020-04-29 CN CN202010359298.0A patent/CN111526070B/en active Active
Patent Citations (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20170126792A1 (en) * | 2015-11-02 | 2017-05-04 | Telefonaktiebolaget L M Ericsson (Publ) | System and methods for intelligent service function placement and autoscale based on machine learning |
| US20180349202A1 (en) * | 2017-05-30 | 2018-12-06 | Hewlett Packard Enterprise Development Lp | Virtual Network Function Resource Allocation |
| CN107332913A (en) * | 2017-07-04 | 2017-11-07 | 电子科技大学 | A kind of Optimization deployment method of service function chain in 5G mobile networks |
| TWI650979B (en) * | 2017-07-25 | 2019-02-11 | 中華電信股份有限公司 | Load balance adjustment system and method thereof |
| CN107819662A (en) * | 2017-11-23 | 2018-03-20 | 电子科技大学 | A kind of service function chain real-time migration method applied to cloud and mist calculating network |
| CN108900358A (en) * | 2018-08-01 | 2018-11-27 | 重庆邮电大学 | Virtual network function dynamic migration method based on deepness belief network resource requirement prediction |
| CN108965024A (en) * | 2018-08-01 | 2018-12-07 | 重庆邮电大学 | A kind of virtual network function dispatching method of the 5G network slice based on prediction |
| CN109711620A (en) * | 2018-12-26 | 2019-05-03 | 浙江大学 | A kind of Short-Term Load Forecasting Method based on GRU neural network and transfer learning |
| CN110335168A (en) * | 2019-04-22 | 2019-10-15 | 山东大学 | Method and system based on GRU optimization power information acquisition terminal fault prediction model |
| CN110766212A (en) * | 2019-10-15 | 2020-02-07 | 哈尔滨工程大学 | Ultra-short-term photovoltaic power prediction method for historical data missing electric field |
| CN110737732A (en) * | 2019-10-25 | 2020-01-31 | 广西交通科学研究院有限公司 | electromechanical equipment fault early warning method |
Non-Patent Citations (4)
| Title |
|---|
| LYUNSIK YANG,ET AL.,: ""Alarm-Based Monitoring for High Availability in Service Function Chain"", 《2016 INTERNATIONAL CONFERENCE ON CLOUD COMPUTING RESEARCH AND INNOVATIONS (ICCCRI)》 * |
| 唐伦等，: ""基于强化学习的5G网络切片虚拟网络功能迁移算法"", 《电子与信息学报》 * |
| 孙士清，: ""NFV环境下服务功能链资源优化部署方法研究"", 《中国优秀硕士学位论文全文数据库 (信息科技辑)》 * |
| 高鹏，: ""5G-C-RAN中最大化效用服务功能链部署算法"", 《计算机工程与应用》 * |
Cited By (14)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111932376A (en) * | 2020-08-14 | 2020-11-13 | 腾讯科技（深圳）有限公司 | Virtual resource data simulation method and device |
| CN111932376B (en) * | 2020-08-14 | 2024-01-23 | 腾讯科技（深圳）有限公司 | Simulation method and device for virtual resource data |
| CN112291804A (en) * | 2020-10-22 | 2021-01-29 | 深圳供电局有限公司 | Service fault diagnosis method for noise network under 5G network slice |
| CN112887145A (en) * | 2021-01-27 | 2021-06-01 | 重庆邮电大学 | Distributed network slice fault detection method |
| CN113608830A (en) * | 2021-07-13 | 2021-11-05 | 全球能源互联网研究院有限公司 | VNF migration method and device based on fault prediction |
| WO2023011618A1 (en) * | 2021-08-06 | 2023-02-09 | International Business Machines Corporation | Predicting root cause of alert using recurrent neural network |
| US11928009B2 (en) | 2021-08-06 | 2024-03-12 | International Business Machines Corporation | Predicting a root cause of an alert using a recurrent neural network |
| CN113490254A (en) * | 2021-08-11 | 2021-10-08 | 重庆邮电大学 | VNF migration method based on bidirectional GRU resource demand prediction in federal learning |
| CN113490254B (en) * | 2021-08-11 | 2022-03-29 | 重庆邮电大学 | VNF migration method based on bidirectional GRU resource demand prediction in federal learning |
| CN113805561A (en) * | 2021-08-16 | 2021-12-17 | 天津大学 | Spacecraft intelligent fault diagnosis method based on deep migration learning |
| CN114423035A (en) * | 2022-01-12 | 2022-04-29 | 重庆邮电大学 | Service function chain abnormity detection method under network slice scene |
| CN114423035B (en) * | 2022-01-12 | 2023-09-19 | 北京宇卫科技有限公司 | Service function chain abnormality detection method in network slice scene |
| CN114785672A (en) * | 2022-03-31 | 2022-07-22 | 北京航空航天大学 | Cloud network service fault prediction method based on fault marker |
| CN114785672B (en) * | 2022-03-31 | 2023-09-26 | 北京航空航天大学 | Cloud network service fault prediction method based on fault markers |
Also Published As
| Publication number | Publication date |
|---|---|
| CN111526070B (en) | 2022-06-03 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111526070B (en) | Service function chain fault detection method based on prediction | |
| CN111368888B (en) | Service function chain fault diagnosis method based on deep dynamic Bayesian network | |
| CN112887145B (en) | Distributed network slice fault detection method | |
| Tuli et al. | Pregan: Preemptive migration prediction network for proactive fault-tolerant edge computing | |
| Wang et al. | A multitask learning-based network traffic prediction approach for SDN-enabled industrial Internet of Things | |
| Canzian et al. | Real-time stream mining: online knowledge extraction using classifier networks | |
| CN113766544A (en) | Multi-edge cooperation-based power Internet of things slice optimization method | |
| Tuli et al. | DRAGON: Decentralized fault tolerance in edge federations | |
| Tian et al. | JMSNAS: Joint model split and neural architecture search for learning over mobile edge networks | |
| Tham et al. | Active learning for IoT data prioritization in edge nodes over wireless networks | |
| Shayesteh et al. | Automated concept drift handling for fault prediction in edge clouds using reinforcement learning | |
| Chen et al. | Traffic prediction-assisted federated deep reinforcement learning for service migration in digital twins-enabled MEC networks | |
| Tuli et al. | Deepft: Fault-tolerant edge computing using a self-supervised deep surrogate model | |
| Ponmalar et al. | Machine Learning Based Network Traffic Predictive Analysis | |
| Jeon et al. | Intelligent resource scaling for container based digital twin simulation of consumer electronics | |
| Aziz et al. | Content-aware network traffic prediction framework for quality of service-aware dynamic network resource management | |
| Agate et al. | Anomaly Detection for Reoccurring Concept Drift in Smart Environments | |
| Ramya et al. | QoS in multimedia application for IoT devices through edge intelligence | |
| Halkidi et al. | Online clustering of distributed streaming data using belief propagation techniques | |
| Al-Saedi et al. | Group-personalized federated learning for human activity recognition through cluster eccentricity analysis | |
| Takele et al. | Resource aware long short-term memory model (RALSTMM) based on-device incremental learning for industrial Internet of Things | |
| CN116400963A (en) | Model automatic parallel method, device and storage medium based on load balancing | |
| Leyva-Pupo et al. | An intelligent scheduling for 5G user plane function placement and chaining reconfiguration | |
| Tang et al. | Enabling combined relay selection in stochastic wireless networks by recurrent neural computing | |
| Mingshi et al. | Failure prediction based vnf migration mechanism for multimedia services in power grid substation monitoring |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| TR01 | Transfer of patent right |
Effective date of registration: 20230710 Address after: Room 2004, 20/F, Hangzhou Bay Smart Valley Building, 233 Yonghui Road, Ningwei Street, Xiaoshan District, Hangzhou City, Zhejiang Province 310000 Patentee after: Hangzhou Mingshi Technology Co.,Ltd. Address before: 1003, Building A, Zhiyun Industrial Park, No. 13 Huaxing Road, Henglang Community, Dalang Street, Longhua District, Shenzhen City, Guangdong Province, 518000 Patentee before: Shenzhen Wanzhida Technology Transfer Center Co.,Ltd. Effective date of registration: 20230710 Address after: 1003, Building A, Zhiyun Industrial Park, No. 13 Huaxing Road, Henglang Community, Dalang Street, Longhua District, Shenzhen City, Guangdong Province, 518000 Patentee after: Shenzhen Wanzhida Technology Transfer Center Co.,Ltd. Address before: 400065 Chongqing Nan'an District huangjuezhen pass Chongwen Road No. 2 Patentee before: CHONGQING University OF POSTS AND TELECOMMUNICATIONS |
|
| TR01 | Transfer of patent right |