CN110247795A - A kind of cloud net resource service chain method of combination and system based on intention - Google Patents
A kind of cloud net resource service chain method of combination and system based on intention Download PDFInfo
- Publication number
- CN110247795A CN110247795A CN201910461367.6A CN201910461367A CN110247795A CN 110247795 A CN110247795 A CN 110247795A CN 201910461367 A CN201910461367 A CN 201910461367A CN 110247795 A CN110247795 A CN 110247795A
- Authority
- CN
- China
- Prior art keywords
- service
- layout
- cost
- intention
- net resource
- 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
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
- H04L41/08—Configuration management of networks or 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
Abstract
The embodiment of the invention provides a kind of cloud net resource service chain method of combination and system based on intention, method include: to provide end-to-end service to cloud net resource based on preset northbound interface frame of reference；Based on the service chaining cradle of deeply study, online layout is carried out to the end-to-end service and dynamic adjusts, wherein in the online layout and dynamic adjustment, solves default multi-goal Optimization Model to minimize service chaining layout cost and delay.A kind of cloud net resource service chain method of combination and system based on intention provided in an embodiment of the present invention, the preset northbound interface frame of reference of offer and the SFC cradle based on DRL are provided, and a multi-goal Optimization Model is constructed, to reduce long term service chain layout cost to the maximum extent.
Description
Technical field
The present invention relates to field of communication technology more particularly to a kind of cloud net resource service chain method of combination based on intention
And system.
Background technique
The rapid growth of Internet of Things service with different QoS requirement brings quickly delivery and QoS to network operator
Guarantee the huge challenge of aspect.Network function virtualizes (Network Function Virtualization, NFV) and software
Define the pass that network (Software Defined Networking, SDN) has become flexible resource distribution and dynamic Service supply
Key technology.But both technologies still need application manual operation to define service model and Configuration network details, in turn
Need high professional qualification administrator and a large amount of time.It is unfavorable that these manual tasks have raising reliability and quick offer service
Effect.Therefore, it is proposed to which the network (the Intent Based Networking, IBN) based on intention is to simplify low-level configuration
And accelerate service offering.
A critical aspects for supporting the service provision based on intention are north orientations unrelated with supplier and unrelated with technology
Interface (northbound interface, NBI), for customer language to be converted to service chaining (Service Function
Chain, SFC) abstract definition.Another committed step is the online layout of the abstract definition based on SFC model, to realize
Requirement drive, the service offering mode of adjust automatically.
However the above method is there is still a need for obtaining complete network details in advance to obtain globally optimal solution, but these are accurate
Information be generally difficult to collect.Therefore a kind of cloud net resource service chain method of combination based on intention is needed now to solve
State problem.
Summary of the invention
To solve the above-mentioned problems, the embodiment of the present invention provides one kind and overcomes the above problem or at least be partially solved
A kind of cloud net resource service chain method of combination and system based on intention of the above problem.
The first aspect embodiment of the present invention provides a kind of cloud net resource service chain method of combination based on intention, comprising:
Based on preset northbound interface frame of reference, end-to-end service is provided to cloud net resource；
Based on the service chaining cradle of deeply study, online layout and dynamic are carried out to the end-to-end service
Adjustment, wherein in the online layout and dynamic adjustment, solve default multi-goal Optimization Model to minimize service
Chain layout cost and delay.
Wherein, the multi-goal Optimization Model indicates are as follows:
min{cost(server)+cost(link)}
Wherein, cost (server) is the cost of the relevant cost of server resource, the forwarding of cost (link) flow, C1,
C2,C3,C4,C5,C6,C7For resource constraint.
Wherein, described to solve default multi-goal Optimization Model to minimize service chaining layout cost and delay, it wraps
It includes:
The optimal solution of the multi-goal Optimization Model is obtained based on the preset double-deck depth Q network algorithm.
Wherein, described that the optimal of the multi-goal Optimization Model is obtained based on the preset double-deck depth Q network algorithm
Solution, comprising:
Operation flow is initialized；
Based on the preset double-deck depth Q network, arranging service is carried out to the operation flow after initialization.
It is wherein, described that operation flow is initialized, comprising:
The target protocol for concentrating random selection to meet the requirements from Cloud Server；
Based on shortest path selection algorithm, the target routing plan between VNF is determined；
Calculate the layout expense of all service chainings.
Wherein, described based on the preset double-deck depth Q network, arranging service, packet are carried out to the operation flow after initialization
It includes:
To the double-deck depth Q network inputs state after state space initialization；
It obtains the corresponding movement of input state and calculates target Q value；
Input state is updated based on gradient descent method until reaching preset termination condition.
The second aspect embodiment of the present invention also provides a kind of cloud net resource service chain arranging system based on intention, comprising:
Service module provides end-to-end service to cloud net resource for being based on preset northbound interface frame of reference；
Layout adjust module, the service chaining cradle for being learnt based on deeply, to the end-to-end service into
The online layout of row and dynamic adjust, wherein in the online layout and dynamic adjustment, solve default multi-objective optimization question
Model is to minimize service chaining layout cost and delay.
The embodiment of the invention provides a kind of electronic equipment for the third aspect, comprising:
Processor, memory, communication interface and bus；Wherein, the processor, memory, communication interface pass through described
Bus completes mutual communication；The memory is stored with the program instruction that can be executed by the processor, the processor
Described program instruction is called to be able to carry out the above-mentioned cloud net resource service chain method of combination based on intention.
The embodiment of the invention provides a kind of non-transient computer readable storage medium, the non-transient meters for fourth aspect
Calculation machine readable storage medium storing program for executing stores computer instruction, and it is above-mentioned based on intention that the computer instruction executes the computer
Cloud net resource service chain method of combination.
A kind of cloud net resource service chain method of combination and system based on intention provided in an embodiment of the present invention, by mentioning
For preset northbound interface frame of reference and the SFC cradle based on DRL, and construct a multi-objective optimization question mould
Type, to reduce long term service chain layout cost to the maximum extent.
Detailed description of the invention
In order to more clearly explain the embodiment of the invention or the technical proposal in the existing technology, below will to embodiment or
Attached drawing needed to be used in the description of the prior art is briefly described, it should be apparent that, the accompanying drawings in the following description is this
Some embodiments of invention without creative efforts, may be used also for those of ordinary skill in the art
To obtain other drawings based on these drawings.
Fig. 1 is a kind of cloud net resource service chain method of combination process signal based on intention provided in an embodiment of the present invention
Figure；
Fig. 2 is the train epochs schematic diagram under different learning rates provided in an embodiment of the present invention；
Fig. 3 is the train epochs schematic diagram under algorithms of different provided in an embodiment of the present invention；
Fig. 4 is the average delay schematic diagram of algorithms of different provided in an embodiment of the present invention；
Fig. 5 is the totle drilling cost schematic diagram of algorithms of different provided in an embodiment of the present invention；
Fig. 6 is a kind of cloud net resource service chain arranging system structural representation based on intention provided in an embodiment of the present invention
Figure；
Fig. 7 is the structural block diagram of electronic equipment provided in an embodiment of the present invention.
Specific embodiment
In order to make the object, technical scheme and advantages of the embodiment of the invention clearer, below in conjunction with the embodiment of the present invention
In attached drawing, technical solution in the embodiment of the present invention is explicitly described, it is clear that described embodiment is this hair
Bright a part of the embodiment, instead of all the embodiments.Based on the embodiments of the present invention, those of ordinary skill in the art exist
Every other embodiment obtained under the premise of creative work is not made, shall fall within the protection scope of the present invention.
Fig. 1 is a kind of cloud net resource service chain method of combination process signal based on intention provided in an embodiment of the present invention
Figure, as shown in Figure 1, comprising:
101, it is based on preset northbound interface frame of reference, provides end-to-end service to cloud net resource；
102, the service chaining cradle based on deeply study, to the end-to-end service carry out online layout and
Dynamic adjusts, wherein in the online layout and dynamic adjustment, solves default multi-goal Optimization Model to minimize
Service chaining layout cost and delay.
It should be noted that the application scenarios of the embodiment of the present invention are to be in Internet of Things how to realize to cloud net resource
Service chaining layout, in this scenario, the embodiment of the present invention proposes an IBN frame of reference and sets to manage Internet of Things basis
It applies and provides end-to-end service across multiple domains.Specifically, in a step 101, IBN frame of reference provided in an embodiment of the present invention
It include: VNF manager (VNF Manager, VNFM) and NFV composer (NFV Orchestrator, NFVO), management and control
Plane processed and data plane.VNF manager and NFV composer are for allowing client to state IR using human-readable language, so
Declaratively strategy it will be converted to high-level service by the NBI based on intention afterwards and be abstracted, such as VNF property, qos feature and threshold value.
Plane is managed and controlled including virtualization architecture manager (VIM) and controller, VIM is by high-level abstractions policy mappings
Low level services chain layout strategy coordinates SDN_C and cloud controller (Cloud_C) to automate SFC layout.Data plane is used for
It places and flows by southbound interface (Southbound Interface, SBI) receiving control message, and for the VNF in cloud domain
Amount routing provides physical resource, and the sensor and actuator of Internet of Things domain are responsible for data collection.So pass through aforementioned present invention
IBN frame of reference provided by embodiment can be realized management Internet of Things infrastructure and provide end-to-end service across multiple domains.
Further, in a step 102, the embodiment of the invention provides a kind of service chainings based on deeply study
Cradle, i.e. the SFC cradle based on DRL.The SFC cradle of the DRL can pass through NBI and net based on intention
Network learning model obtains SFC abstract model and environment detail.Be then based on DRL SFC cradle can tuning controller,
The corresponding actions of current network are realized by the specific SBI of technology, meanwhile, network provides the feedback in relation to rewarding or punishing,
To promote the SFC cradle of DRL to adjust its behavior, by effectively training so that the SFC cradle realization of DRL is optimal
Strategy, cloud net resource service chain layout scheme needed for the optimal policy, that is, embodiment of the present invention.Wherein, optimal plan is being obtained
During slightly, the embodiment of the present invention establishes a multi-goal Optimization Model to minimize SFC layout cost and prolong
Late, final required cloud net resource service chain layout scheme is obtained by the optimal solution to the multi-goal Optimization Model.
Wherein, in the multi-goal Optimization Model that the embodiment of the present invention is established, SFC is by triple s={ (vso,
vde)s,Fs,rsIndicate, wherein (vso,vde)sIndicate s source node to and destination node pair. vsoGenerating has data transmission speed
Rate is rsFlow.FsIndicate specific SFC information, the attribute including VNF, sequence and connectivity.Needed for Cloud Server
The f ∈ F of CPU, memory source and VNFsProcessing postpone by cpuf, memfAnd dfIt indicates.Virtual chain between the u and w of VNF
Connect byIt indicates.Define DsTo postpone threshold value.
So the physical network in cloud domain is indicated by weighted undirected graph G=(N, L), and wherein N and L respectively indicates node and has
Wired link.Node is divided into two classes: the first kind is the interchanger for converting flow, and the second class is for hosts virtual machine v ∈ V
Cloud Server.Server and the quantity of link are indicated with M and H.Cloud Server v has based on the CPU by placing VNF example
Calculation and memory source, these examples are respectively by Capcpu(v) and Capmem(v) it indicates.Physical link l between node i and jijTool
There is maximum data transfer rate bijWith transmission delay dij。 Indicate that the f of the VNF in s has been mapped to Cloud Server,
Otherwise it is Indicate the virtual link in sIt has been mapped to physical link, has been otherwise 0.
VNF example needs the memory source in CPU computing resource and Cloud Server.It needs to consider load balancing, because of dimension
Load balancing between shield server and link to avoid flow congestion and can further increase network cost efficiency.Therefore, originally
Inventive embodiments propose two load balancing factor ΦvAnd Θij, it is used to indicate the load condition of network, their value and money
Source utilization rate has positive correlation.ΦvIt calculates as follows:
Wherein α1,β1,χ1It is positive parameter, for Φ in Setup Cost calculating processvValue.ΦvIt is UvLinear function also
It is whether exponential function depends on UvRange.UvThe weighted sum for indicating CPU and memory usage, is calculated by following formula:
Wherein, epAnd emRepresent the weight of CPU and memory usage, ep+em=1.The relevant cost of server resource by
Following formula calculates:
The unit price of CPU and memory source uses c respectively1And c2It indicates.Next consider the forwarding cost in flow routing, bear
Carry balance factor ΘijIt calculates as follows:
Wherein α2,β2,χ2It is positive parameter and is used to adjust ΘijValue.UijIndicate link lijThe use of middle transmission rate
Rate, calculation are as follows:
The cost calculation mode of flow forwarding is as follows:
Wherein, c3Indicate the unit price of link transmission rate.(el·Θij+ed·dij/Ds) represent ΘijWeighted sum delay
dij, el+ed=1.Cost (link) is consisted of three parts: Θij, fixed delay dijAnd unit price.It can from above-mentioned calculating formula
Out, node or link with larger surplus resources have relatively low cost.
The totle drilling cost Cost_total of SFC layout process calculates as follows:
Cost_total=cost (server)+cost (link)；
Resource constraint is obtained by following formula:
Deferred constraint is obtained by following formula:
To establish the multi-goal Optimization Model for being intended to improve cost efficiency and guaranteed qos:
min{cost(server)+cost(link)}
A kind of cloud net resource service chain method of combination and system based on intention provided in an embodiment of the present invention, by mentioning
For preset northbound interface frame of reference and the SFC cradle based on DRL, and construct a multi-objective optimization question mould
Type, to reduce long term service chain layout cost to the maximum extent.
On the basis of the above embodiments, described to solve default multi-goal Optimization Model to minimize service chaining volume
Line up this and delay, comprising:
The optimal solution of the multi-goal Optimization Model is obtained based on preset bilayer depth Q network (DDQN) algorithm.
For the multi-goal Optimization Model proposed in above-described embodiment, the embodiment of the present invention devises a bilayer
Depth Q network algorithm finds out the optimal solution of multi-goal Optimization Model.
Specifically, optimization problem is expressed as Markovian decision process { ST, A, Rd, P } by the embodiment of the present invention,
Middle ST indicates that state space, A indicate motion space, and Rd is defined as reward function, and P is state transition probability.It is defined as follows:
State space are as follows: each agency is sometime having corresponding layout scheme.The state is defined as owning
The satisfaction degree of the qos requirement of SFC, and calculated by following formula:
ST={ st1,sts,...,stK}；
Wherein, sts={ 0,1 }, K are the quantity of SFC.sts=1 indicates that the delay of SFC is wanted under current layout scheme
Asking can be met.Otherwise, sts=0.Stateful quantity be 2K。
Motion space are as follows: between two states of SFC conversion indicate by take action to change VNF placement or
Routing.Behavior aggregate A is defined as follows shown in formula:
Wherein, X is designed to the actions available collection that VNF is placed in SFC operation flow.It, can be in addition, if providing X
The routing between VNF is obtained by shortest path first.
Y is designed to the actions available collection of the routing of the flow between VNF.Therefore, the operation that VNF is placed and flow routes
Space representation is A={ X, Y }.The movement number of s is 2M+H。
Reward are as follows: if agent takes certain action, state stsTransfer to new state st's.Acting on behalf of s can also be with
Rd is rewarded in acquisition immediatelys(as, st, st'), it is defined as stsIt is transferred to st'sWhen reduce cost.
Rds(as, st, st') and=cost (sts)-cost(st's)；
Wherein cost (sts) and cost (st's) represent state stsAnd st'sLayout expense and.By accumulating long-term prize
Encourage Rds(as, st, st') and highest cost efficiency may be implemented, according to current state, tactful π can obtain SFC and will take
Corresponding action.Optimizing behavior isQs(st a) is defined as state-function of movement, and indicates specified shape
The expected accumulation discount reward of state-movement.Qs(st a) is indicated are as follows:
Wherein, γ is discount factor, indicates importance of the following reward in study.It, can be with according to Bellman equation
It obtains as follows optimal
Ps(as, st, st') and it indicates from state st to the transition probability of state st'.Therefore, it can be obtained most based on above formula
Dominant strategyAnd it indicates are as follows:
In practice, it is generally difficult to obtain accurate transition probability.Therefore, Q study be designed to based on available information with
Iterative manner finds optimal solution, and it updates Q value function using following equation:
Wherein, δ is learning efficiency, influences Qs(st, turnover rate a).
It is understood that Q study completes iteration based on Q value table, so if state and motion space are very big, then
It is difficult to obtain optimal solution.In order to overcome this weakness, depth Q network (DQN) provided in an embodiment of the present invention passes through depth nerve
Network (DNN) rather than the approximate Q value function of Q value table.DNN can be considered as the depth map with multiple process layers.θ is indicated
The weight of these layers, and declined by gradient and updated.The approximation for the value function that DQN is used is calculated by following formula:
Qs(st,a,θ)≈Qs(st,a)。
In addition to this, DQN is reset with pinpoint target network using experience and eliminates data dependency.Define target network
To be based on weight θ-Calculate target Q value.θ and θ-Between difference be that θ updates in each iteration, but θ-In fixed number of times
It is updated in iteration.Target Q value function is given by:
The loss function of DQN is defined as mean square error, is calculated by following formula:
L (θ)=E [(Target_Qs-Qs(st,a,θ))2]；
In each iteration, need to update weight θ according to gradientMinimize loss function.Renewal function by
Following formula calculates:
It is emphasized that DQN and Q study all carries out calculating Target_Q using maximal functions, which results in over-evaluate
Problem.As an improvement, DDQN finds the corresponding actions with maximum Q value in current network first, rather than directly in target
The maximum Q value of everything in network:
amax(st', θ)=argmaxa'∈AQs(st',a',θ)；
Then selected movement a is usedmax(st', θ) rewrites Target_Qs.New Target_Q in DDQNsBy following formula meter
It calculates:
Similarly, L (θ) and θ ' is also required to update together in DDQN.
Specifically, bilayer depth Q network algorithm provided by the embodiment of the present invention may include two parts, first part
For operation flow initialization, operation flow initialization be may comprise steps of:
1, from σf,sRandomly choose f ∈ FsFeasible placement schemesσf,sIt indicates for f ∈ FsFeasible cloud service
Device collection.
2, the routing plan between VNF is obtained by shortest path first.
3, the layout expense of all SFC is calculated.
Second part is the arranging service using DDQN network, which may comprise steps of:
1, state space is initialized to ST={ st1,sts,...,stK}。
2, for SFCs, it is by stsQ value network Q is added to as inputs(sts,a,θ)。
3, it is acted by ε-greedy strategy acquisition.The strategy is respectively with probability ε and 1- ε selection random action and best
Movement.
4, the conversion of all SFC is stored in experience and resets in memory.
5, each agency samples (st, a, Rd, st') from ER, and according to whether calculates its target Q value for end-state.
6, they execute gradient decline step (Target_Q relative to the θ of Q value networks-Qs(st,a,θ))2.In every ufθ
After step, θ is replaced by θ.
If 7, { 1,1 ..., 1 } current state ST=, training will be terminated.
In summary the DDQN algorithm that process can be seen that design of the embodiment of the present invention can obtain multiple-objection optimization and ask
The optimal solution of model is inscribed, and the algorithm has better cost efficiency and convergence, and can require with guaranteed qos, makes flow
It is balanced.
In order to verify the performance of proposed method of the embodiment of the present invention, the embodiment of the present invention has carried out emulation experiment.Specifically
, the embodiment of the present invention utilizes the cloud net being made of 30 nodes (10 Cloud Servers and 20 interchangers) and 50 links
Network emulates mentioned algorithm.The maximum data transfer rate of link is fixed as 1Gbps.The transmission delay of link is 1-
3ms.The CPU and memory source of server are respectively set to 32 and 100-200GB.Each SFC needs 2-4 VNF, data
Transmission rate is 20-50Mbps.Each VNF needs 2-4 CPU and 5-10GB memory source, and processing delay is 2-5ms.
The structure of DNN includes the hidden layer of three neural networks being fully connected, and has 64,32,32 neurons.Mainly from algorithm
Constringency performance optimizes performance these two aspects to be emulated.
Assessing constringency performance under different learning rates first: Fig. 2 is different learning rates provided in an embodiment of the present invention
Under train epochs schematic diagram, as shown in Fig. 2, tool there are three types of learning rate DDQN algorithm when episode starts have it is huge
Big train epochs.Train epochs tend to decline with the increase of episode, and which reflects the good convergence performances of DDQN.
On the other hand, learning rate is the key factor of constringency performance.By taking EP=90 as an example, the DDQN of δ=0.001 needs 92 training
Step.As a comparison, δ=0.01 and the algorithm of δ=0.1 only need 40 and 26 steps to be obtained with optimal solution.
Then the constringency performance under more different nitrification enhancements, Fig. 3 are different calculations provided in an embodiment of the present invention again
Train epochs schematic diagram under method, as shown in Figure 3.As can be seen that Q study has lower convergence at different episode
Can, because measure needed for it eliminates data dependency is less.On the contrary, DQN and DDQN establish experience playback and pinpoint target
Network solves data dependence, therefore their training step learns less than Q.With Q-learning, the instruction of DQN, DDQN
For practicing step, respectively 51,32,26.Therefore, compared with DQN, solve the problems, such as to over-evaluate be also DDQN advantage.
Further, the embodiment of the present invention assesses average retardation as a comparison with following two algorithm, total reward and negative
Carry equilibrium state.QoS driving placement algorithm (QPA): it obtains the end-to-end path of SFC first, then expand to path it
On, to minimize cost and delay, while meeting resource requirement.Random fit Placement (RPA): the placement of VNF is with random
The form of fitting executes, to consider to meet the constrained all schemes of institute, and randomly chooses one of them, then also random
Select path therein.Fig. 4 is the average delay schematic diagram of algorithms of different provided in an embodiment of the present invention, as shown in figure 4, by
It is seldom in the quantity of SFC request, therefore the average retardation of four kinds of algorithms is lower when starting.With the increase of SFC, delay with
Different amplitudes and increase.When SFC quantity is 200, the average retardation of DDQN, DQN, QPA and RPA are respectively 38ms,
43ms, 48ms, 56ms.Due to the randomness of RPA, the delay performance of RPA is poor.Although DQN and QPA examine delay minimization
Including worry, but they have ignored the influence of load balance, this may cause network congestion.In contrast, it is asked in different SFC
It asks under quantity, DDQN has better delay performance.
Fig. 5 is the totle drilling cost schematic diagram of algorithms of different provided in an embodiment of the present invention, as shown in figure 5, in different number
Under SFC, the totle drilling cost of DDQN, DQN, QPA are always lower than RPA, because they all consider the cost minimization of objective function.Example
Such as, when SFC quantity is 300, cost the ratio QPA, DQN, DDQN high 16% of RPA, 20%, 27%.In these three algorithms,
DDQN has optimal cost efficiency, over-evaluates problem because it overcomes and lays particular emphasis on load balance, QPA and DQN have ignored this
A bit.Therefore, DDQN can obtain optimal delay and cost in SFC layout process.
For load balancing state, by taking SFCs=300 as an example, the variance ratio RPA, SPA of DDQN junctor usage and
DQN low 62%, 55%, 41%.Equally, the variation of DDQN server utilization rate is also lower than 81%, 65%, 48%.It is compiled in SFC
It arranges in Optimized model, ΦvAnd ΘijDesigned for maintenance network balance, so that DDQN is it is possible to prevente effectively from network congestion.
Fig. 6 is a kind of cloud net resource service chain arranging system structural representation based on intention provided in an embodiment of the present invention
Figure, as shown in Figure 6, comprising: service module 601 and layout adjust module 602, in which:
Service module 601 is used to be based on preset northbound interface frame of reference, provides end-to-end service to cloud net resource；
The service chaining cradle that layout adjustment module 602 is used to learn based on deeply, to the end-to-end service
It carries out online layout and dynamic adjusts, wherein in the online layout and dynamic adjustment, solve default multiple-objection optimization and ask
Model is inscribed to minimize service chaining layout cost and delay.
Specifically how can be used for executing by service module 601 and layout adjustment module 602 shown in FIG. 1 based on intention
Cloud net resource service chain method of combination embodiment technical solution, it is similar that the realization principle and technical effect are similar, no longer superfluous herein
It states.
A kind of cloud net resource service chain arranging system based on intention provided in an embodiment of the present invention, it is default by providing
Northbound interface frame of reference and SFC cradle based on DRL, and a multi-goal Optimization Model is constructed, with most
Reduce to limits long term service chain layout cost.
On the basis of the above embodiments, the multi-goal Optimization Model indicates are as follows:
min{cost(server)+cost(link)}
Wherein, cost (server) is the cost of the relevant cost of server resource, the forwarding of cost (link) flow, C1,
C2,C3,C4,C5,C6,C7For resource constraint.
On the basis of the above embodiments, the layout adjustment module includes:
DDQN unit, for obtaining the multi-goal Optimization Model based on the preset double-deck depth Q network algorithm
Optimal solution.
On the basis of the above embodiments, the DDQN unit includes:
Initialization section, for being initialized to operation flow；
Arranging service part, for carrying out industry to the operation flow after initialization based on the preset double-deck depth Q network
Business layout.
On the basis of the above embodiments, the initialization section is specifically used for:
The target protocol for concentrating random selection to meet the requirements from Cloud Server；
Based on shortest path selection algorithm, the target routing plan between VNF is determined；
Calculate the layout expense of all service chainings.
On the basis of the above embodiments, the arranging service part is specifically used for:
To the double-deck depth Q network inputs state after state space initialization；
It obtains the corresponding movement of input state and calculates target Q value；
Input state is updated based on gradient descent method until reaching preset termination condition.
The embodiment of the present invention provides a kind of electronic equipment, comprising: at least one processor；And it is logical with the processor
Believe at least one processor of connection, in which:
Fig. 7 is the structural block diagram of electronic equipment provided in an embodiment of the present invention, referring to Fig. 7, the electronic equipment, comprising:
Processor (processor) 701, communication interface (Communications Interface) 702, memory (memory) 703
With bus 704, wherein processor 701, communication interface 702, memory 703 complete mutual communication by bus 704.Place
Reason device 701 can call the logical order in memory 703, to execute following method: be referred to based on preset northbound interface
Framework provides end-to-end service to cloud net resource；Based on the service chaining cradle of deeply study, to described end-to-end
Service carries out online layout and dynamic adjusts, wherein in the online layout and dynamic adjustment, it is excellent to solve default multiple target
Change problem model to minimize service chaining layout cost and delay.
The embodiment of the present invention discloses a kind of computer program product, and the computer program product is non-temporary including being stored in
Computer program on state computer readable storage medium, the computer program include program instruction, when described program instructs
When being computer-executed, computer is able to carry out method provided by above-mentioned each method embodiment, for example, based on default
Northbound interface frame of reference, to cloud net resource provide end-to-end service；Service chaining layout frame based on deeply study
Frame carries out online layout to the end-to-end service and dynamic adjusts, wherein in the online layout and dynamic adjustment, asks
Default multi-goal Optimization Model is solved to minimize service chaining layout cost and delay.
The embodiment of the present invention provides a kind of non-transient computer readable storage medium, and the non-transient computer is readable to deposit
Storage media stores computer instruction, and the computer instruction executes the computer provided by above-mentioned each method embodiment
Method, for example, be based on preset northbound interface frame of reference, provide end-to-end service to cloud net resource；It is strong based on depth
The service chaining cradle that chemistry is practised carries out online layout to the end-to-end service and dynamic adjusts, wherein it is described
In line layout and dynamic adjustment, default multi-goal Optimization Model is solved to minimize service chaining layout cost and delay.
Through the above description of the embodiments, those skilled in the art can be understood that each embodiment
It can realize by means of software and necessary general hardware platform, naturally it is also possible to pass through hardware.Based on such reason
Solution, substantially the part that contributes to existing technology can embody above-mentioned technical proposal in the form of software products in other words
Out, which may be stored in a computer readable storage medium, such as ROM/RAM, magnetic disk, CD,
It uses including some instructions so that a computer equipment (can be personal computer, server or the network equipment etc.) is held
Method described in certain parts of each embodiment of row or embodiment.
Finally, it should be noted that the above embodiments are merely illustrative of the technical solutions of the present invention, rather than its limitations；To the greatest extent
Present invention has been described in detail with reference to the aforementioned embodiments for pipe, those skilled in the art should understand that: it is still
It is possible to modify the technical solutions described in the foregoing embodiments, or part of technical characteristic is equally replaced
It changes；And these are modified or replaceed, technical solution of various embodiments of the present invention that it does not separate the essence of the corresponding technical solution
Spirit and scope.
Claims (9)
1. a kind of cloud net resource service chain method of combination based on intention characterized by comprising
Based on preset northbound interface frame of reference, end-to-end service is provided to cloud net resource；
Based on the service chaining cradle of deeply study, online layout is carried out to the end-to-end service and dynamic adjusts,
Wherein, in the online layout and dynamic adjustment, default multi-goal Optimization Model is solved to minimize service chaining layout
Cost and delay.
2. the cloud net resource service chain method of combination according to claim 1 based on intention, which is characterized in that more mesh
Mark optimization problem model is expressed as:
min{cost(server)+cost(link)}
Wherein, cost (server) is the cost of the relevant cost of server resource, the forwarding of cost (link) flow, C1,C2,C3,
C4,C5,C6,C7For resource constraint.
3. the cloud net resource service chain method of combination according to claim 1 based on intention, which is characterized in that the solution
Multi-goal Optimization Model is preset to minimize service chaining layout cost and delay, comprising:
The optimal solution of the multi-goal Optimization Model is obtained based on the preset double-deck depth Q network algorithm.
4. the cloud net resource service chain method of combination according to claim 3 based on intention, which is characterized in that described to be based on
Preset bilayer depth Q network algorithm obtains the optimal solution of the multi-goal Optimization Model, comprising:
Operation flow is initialized；
Based on the preset double-deck depth Q network, arranging service is carried out to the operation flow after initialization.
5. the cloud net resource service chain method of combination according to claim 4 based on intention, which is characterized in that described to industry
Business process is initialized, comprising:
The target protocol for concentrating random selection to meet the requirements from Cloud Server；
Based on shortest path selection algorithm, the target routing plan between VNF is determined；
Calculate the layout expense of all service chainings.
6. the cloud net resource service chain method of combination according to claim 4 based on intention, which is characterized in that described to be based on
Preset bilayer depth Q network, carries out arranging service to the operation flow after initialization, comprising:
To the double-deck depth Q network inputs state after state space initialization；
It obtains the corresponding movement of input state and calculates target Q value；
Input state is updated based on gradient descent method until reaching preset termination condition.
7. a kind of cloud net resource service chain arranging system based on intention characterized by comprising
Service module provides end-to-end service to cloud net resource for being based on preset northbound interface frame of reference；
Layout adjusts module, and the service chaining cradle for being learnt based on deeply carries out the end-to-end service
Line layout and dynamic adjust, wherein in the online layout and dynamic adjustment, solve default multi-goal Optimization Model with
Minimize service chaining layout cost and delay.
8. a kind of electronic equipment including memory, processor and stores the calculating that can be run on a memory and on a processor
Machine program, which is characterized in that the processor is realized as described in any one of claim 1 to 6 when executing described program based on meaning
The step of cloud net resource service chain method of combination of figure.
9. a kind of non-transient computer readable storage medium, is stored thereon with computer program, which is characterized in that the computer journey
The cloud net resource service chain method of combination as described in any one of claim 1 to 6 based on intention is realized when sequence is executed by processor
The step of.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910461367.6A CN110247795B (en) | 2019-05-30 | 2019-05-30 | Intent-based cloud network resource service chain arranging method and system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910461367.6A CN110247795B (en) | 2019-05-30 | 2019-05-30 | Intent-based cloud network resource service chain arranging method and system |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110247795A true CN110247795A (en) | 2019-09-17 |
| CN110247795B CN110247795B (en) | 2020-09-25 |
Family
ID=67885322
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910461367.6A Active CN110247795B (en) | 2019-05-30 | 2019-05-30 | Intent-based cloud network resource service chain arranging method and system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110247795B (en) |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111865686A (en) * | 2020-07-20 | 2020-10-30 | 北京百度网讯科技有限公司 | Cloud product capacity expansion method, device, equipment and storage medium |
| CN112040011A (en) * | 2020-06-30 | 2020-12-04 | 北京来也网络科技有限公司 | Simulation training method and device based on RPA and AI, computing equipment and medium |
| CN112583629A (en) * | 2019-09-30 | 2021-03-30 | 华为技术有限公司 | Information processing method, related equipment and computer storage medium |
| CN112637064A (en) * | 2020-12-31 | 2021-04-09 | 广东电网有限责任公司电力调度控制中心 | Service chain arranging method based on improved depth-first search algorithm |
| CN114827284A (en) * | 2022-04-21 | 2022-07-29 | 中国电子技术标准化研究院 | Service function chain arrangement method and device in industrial Internet of things and federal learning system |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107438983A (en) * | 2014-09-22 | 2017-12-05 | 沃尔廷控股公司 | Determine software defined network, storage and the physics and the compiler and method of virtual resource that calculate |
| US20180123974A1 (en) * | 2015-09-28 | 2018-05-03 | Centurylink Intellectual Property Llc | Intent-Based Services Orchestration |
| CN108536144A (en) * | 2018-04-10 | 2018-09-14 | 上海理工大学 | A kind of paths planning method of fusion dense convolutional network and competition framework |
| CN108600101A (en) * | 2018-03-21 | 2018-09-28 | 北京交通大学 | A kind of network for the optimization of end-to-end time delay performance services cross-domain method of combination |
| CN108900419A (en) * | 2018-08-17 | 2018-11-27 | 北京邮电大学 | Route decision method and device based on deeply study under SDN framework |
| CN109245916A (en) * | 2018-08-15 | 2019-01-18 | 西安电子科技大学 | A kind of the cloud access net system and method for intention driving |
| CN109358971A (en) * | 2018-10-30 | 2019-02-19 | 电子科技大学 | Quick and load balancing service function chain dispositions method in dynamic network environment |
-
2019
- 2019-05-30 CN CN201910461367.6A patent/CN110247795B/en active Active
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107438983A (en) * | 2014-09-22 | 2017-12-05 | 沃尔廷控股公司 | Determine software defined network, storage and the physics and the compiler and method of virtual resource that calculate |
| US20180123974A1 (en) * | 2015-09-28 | 2018-05-03 | Centurylink Intellectual Property Llc | Intent-Based Services Orchestration |
| CN108600101A (en) * | 2018-03-21 | 2018-09-28 | 北京交通大学 | A kind of network for the optimization of end-to-end time delay performance services cross-domain method of combination |
| CN108536144A (en) * | 2018-04-10 | 2018-09-14 | 上海理工大学 | A kind of paths planning method of fusion dense convolutional network and competition framework |
| CN109245916A (en) * | 2018-08-15 | 2019-01-18 | 西安电子科技大学 | A kind of the cloud access net system and method for intention driving |
| CN108900419A (en) * | 2018-08-17 | 2018-11-27 | 北京邮电大学 | Route decision method and device based on deeply study under SDN framework |
| CN109358971A (en) * | 2018-10-30 | 2019-02-19 | 电子科技大学 | Quick and load balancing service function chain dispositions method in dynamic network environment |
Non-Patent Citations (8)
| Title |
|---|
| FRANCO CALLEGATI: "Performance of Intent-based Virtualized Network Infrastructure Management", 《IEEE ICC 2017 SAC SYMPOSIUM SDN & NFV TRACK》 * |
| HAMED RAHMANI KHEZRI: "Deep Q-Learning for Dynamic Reliability Aware NFV-Based Service Provisioning", 《谷歌学术》 * |
| NGUYEN CONG LUONG: "Applications of Deep Reinforcement Learning in communication and networking: A Survey", 《IEEE》 * |
| WENCHEN HE: "Markov Approximation Method for Optimal Service Orchestration in IoT Network", 《IEEE ACCESS》 * |
| 左国玉: "基于动作注意策略的树形DDQN目标候选区域提取方法", 《电子与信息学报》 * |
| 肖逸凯: "基于软件定义网络的虚拟网络功能服务链在线调度技术", 《万方学位论文》 * |
| 马骋乾: "强化学习研究综述", 《指挥控制与仿真》 * |
| 魏亮: "基于强化学习的服务链映射算法，", 《通信学报》 * |
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112583629A (en) * | 2019-09-30 | 2021-03-30 | 华为技术有限公司 | Information processing method, related equipment and computer storage medium |
| CN112583629B (en) * | 2019-09-30 | 2022-06-10 | 华为技术有限公司 | Information processing method, related equipment and computer storage medium |
| CN112040011A (en) * | 2020-06-30 | 2020-12-04 | 北京来也网络科技有限公司 | Simulation training method and device based on RPA and AI, computing equipment and medium |
| CN112040011B (en) * | 2020-06-30 | 2023-06-09 | 北京来也网络科技有限公司 | Simulation training method, device, computing equipment and medium based on RPA and AI |
| CN111865686A (en) * | 2020-07-20 | 2020-10-30 | 北京百度网讯科技有限公司 | Cloud product capacity expansion method, device, equipment and storage medium |
| CN112637064A (en) * | 2020-12-31 | 2021-04-09 | 广东电网有限责任公司电力调度控制中心 | Service chain arranging method based on improved depth-first search algorithm |
| CN114827284A (en) * | 2022-04-21 | 2022-07-29 | 中国电子技术标准化研究院 | Service function chain arrangement method and device in industrial Internet of things and federal learning system |
| CN114827284B (en) * | 2022-04-21 | 2023-10-03 | 中国电子技术标准化研究院 | Service function chain arrangement method and device in industrial Internet of things and federal learning system |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110247795B (en) | 2020-09-25 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110247795A (en) | A kind of cloud net resource service chain method of combination and system based on intention | |
| Liu et al. | Dynamic service function chain orchestration for NFV/MEC-enabled IoT networks: A deep reinforcement learning approach | |
| CN110365514B (en) | SDN multistage virtual network mapping method and device based on reinforcement learning | |
| CN109768940A (en) | The flow allocation method and device of multi-service SDN network | |
| Rjoub et al. | Trust-driven reinforcement selection strategy for federated learning on IoT devices | |
| Yuan et al. | A Q-learning-based approach for virtual network embedding in data center | |
| CN110022230A (en) | The parallel dispositions method of service chaining and device based on deeply study | |
| CN108111335A (en) | A kind of method and system dispatched and link virtual network function | |
| Wang et al. | Optimal deployment of cloudlets based on cost and latency in Internet of Things networks | |
| CN110198280A (en) | A kind of SDN link allocation method based on BP neural network | |
| WO2023216489A1 (en) | Computing power network node evaluation and operation method and apparatus | |
| Wang et al. | A reinforcement learning approach for online service tree placement in edge computing | |
| Dalgkitsis et al. | Dynamic resource aware VNF placement with deep reinforcement learning for 5G networks | |
| Li et al. | MEC-based dynamic controller placement in SD-IoV: A deep reinforcement learning approach | |
| Liu et al. | Multi-objective optimization service function chain placement algorithm based on reinforcement learning | |
| Yamansavascilar et al. | Deepedge: A deep reinforcement learning based task orchestrator for edge computing | |
| Dalgkitsis et al. | SCHE2MA: Scalable, energy-aware, multidomain orchestration for beyond-5G URLLC services | |
| CN113543160B (en) | 5G slice resource allocation method, device, computing equipment and computer storage medium | |
| CN110233763B (en) | Virtual network embedding algorithm based on time sequence difference learning | |
| Faraji-Mehmandar et al. | A self-learning approach for proactive resource and service provisioning in fog environment | |
| CN115225512B (en) | Multi-domain service chain active reconfiguration mechanism based on node load prediction | |
| Rădulescu et al. | Analysing congestion problems in multi-agent reinforcement learning | |
| CN113596138B (en) | Heterogeneous information center network cache allocation method based on deep reinforcement learning | |
| Jin et al. | Resource management of cloud-enabled systems using model-free reinforcement learning | |
| Xia et al. | Learn to optimize: Adaptive VNF provisioning in mobile edge clouds |
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 |