CN113535399B

CN113535399B - NFV resource scheduling method, device and system

Info

Publication number: CN113535399B
Application number: CN202110800472.5A
Authority: CN
Inventors: 吴立军; 李志圆
Original assignee: University of Electronic Science and Technology of China
Current assignee: University of Electronic Science and Technology of China
Priority date: 2021-07-15
Filing date: 2021-07-15
Publication date: 2023-04-25
Anticipated expiration: 2041-07-15
Also published as: CN113535399A

Abstract

The application discloses a NFV resource scheduling method, device and system. The NFV resource scheduling method acquires original request data and dynamic network topology information; acquiring service request characteristics according to the original request data; obtaining topological characteristics according to the dynamic network topological information; fusing the service request features and the topology features to form fused features; acquiring a trained resource allocation scheduling model; and inputting the fusion characteristics into a resource allocation scheduling model so as to acquire scheduling processing action information. The NFV resource scheduling method has the following advantages: in the NFV resource allocation problem, three sub-problems are solved in a coordinated scheme for the first time, and a new thought is provided for NFV-RA problem research. In the NFV-RA problem, a graph convolution neural network is used for the first time, and is improved, feature extraction is carried out on node and link states at the same time, and topology change is effectively perceived.

Description

NFV resource scheduling method, device and system

Technical Field

The application relates to the technical field of network function virtualization, in particular to an NFV resource scheduling method, an NFV resource scheduling device and an NFV resource scheduling system.

Background

Network Function Virtualization (NFV) software the network functions by virtualization technology, separate from proprietary hardware devices, and can run on standard server virtualization software. Thanks to the flexibility of NFV, network operators can save a lot of costs while improving the network resource utilization. However, to dig deeper into the potential of NFV, an important challenge is the network resource allocation problem in NFV (NFV-RA).

NFV-RA can be divided into VNFs chain components (VNFs-CC); VNF forwarding graph embedding (VNF-FGE); VNFs scheduling (VNFs-SCH) three phases. Designing a resource allocation algorithm that effectively and cooperatively solves the three phases is a key to solving the NFV-RA problem. However, most of the current solutions only solve a single stage, or solve the problem of more than one stage in a non-coordinated way, and do not consider the online arrival mode of the service.

It is therefore desirable to have a solution that overcomes or at least alleviates at least one of the above-mentioned drawbacks of the prior art.

Disclosure of Invention

It is an object of the present invention to provide a NFV resource scheduling method that overcomes or at least alleviates at least one of the above-mentioned drawbacks of the prior art.

In one aspect of the present invention, there is provided an NFV resource scheduling method, including:

acquiring original request data and dynamic network topology information;

acquiring service request characteristics according to the original request data;

obtaining topological characteristics according to the dynamic network topological information;

fusing the service request features and the topology features to form fused features;

acquiring a trained resource allocation scheduling model;

and inputting the fusion characteristics into the resource allocation scheduling model so as to acquire scheduling processing action information.

Optionally, the acquiring service request features according to the original request data includes:

service request features in the original request data are acquired through a multi-head self-attention mechanism.

Optionally, the obtaining the topology feature according to the dynamic network topology information includes:

and acquiring topological characteristics in the dynamic network topological information through the double-component graph convolutional neural network.

Optionally, after the inputting the fusion feature into the resource allocation scheduling model to obtain scheduling processing action information, the NFV resource scheduling method further includes:

and controlling the NFV resource to schedule the NFV environment according to the scheduling processing action information, and acquiring evaluation information returned by the NFV environment after scheduling by the scheduling processing action information.

Optionally, the evaluation information includes an evaluation of the scheduling processing action information and index information.

Optionally, after acquiring the evaluation information returned by the NFV environment after the scheduling by the scheduling processing action information, the NFV resource scheduling method further includes:

and adjusting parameters in the resource allocation scheduling model according to the acquired evaluation information.

The application also provides an NFV resource scheduling device, which includes:

the system comprises an original request data acquisition module, a data processing module and a data processing module, wherein the original request data acquisition module is used for acquiring original request data;

the dynamic network topology information acquisition module is used for acquiring dynamic network topology information;

the service request feature acquisition module is used for acquiring service request features according to the original request data;

the topological feature acquisition module is used for acquiring topological features according to the dynamic network topological information;

the fusion module is used for fusing the service request characteristics and the topology characteristics to form fusion characteristics;

the model acquisition module is used for acquiring a trained resource allocation scheduling model;

and the model calculation module is used for inputting the fusion characteristics into the resource allocation scheduling model so as to acquire scheduling processing action information.

The application also provides an NFV resource scheduling system, which comprises a plurality of NFV resource scheduling devices.

The application also provides an electronic device comprising a memory, a processor and a computer program stored in the memory and capable of running on the processor, wherein the processor implements the NFV resource scheduling method as described above when executing the computer program.

The present application also provides a computer readable storage medium storing a computer program which, when executed by a processor, is capable of implementing the NFV resource scheduling method as described above.

The beneficial effects are that:

the NFV resource scheduling method has the following advantages:

1. in the NFV resource allocation problem, three sub-problems are solved in a coordinated scheme for the first time, and a new thought is provided for NFV-RA problem research.

2. In the NFV-RA problem, a graph convolution neural network is used for the first time, and is improved, feature extraction is carried out on node and link states at the same time, and topology change is effectively perceived.

3. For the NFV-RA problem and the inadaptability of traditional deep reinforcement learning to the problem, a new reinforcement learning training algorithm is proposed.

4. In the NFV-RA problem, multiple NFV resource scheduling devices are used for the first time for online processing.

5. In the NFV-RA problem, a self-attention mechanism is first used, position coding is added, and feature extraction is performed on a service request.

6. The NFV resource scheduling device adopts end-to-end training, and is convenient to deploy and debug.

Drawings

Fig. 1 is a flow chart of an NFV resource scheduling method according to an embodiment of the present application.

Fig. 2 is an electronic device for implementing the NFV resource scheduling method shown in fig. 1.

Detailed Description

In order to make the purposes, technical solutions and advantages of the implementation of the present application more clear, the technical solutions in the embodiments of the present application will be described in more detail below with reference to the accompanying drawings in the embodiments of the present application. In the drawings, the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The described embodiments are some, but not all, of the embodiments of the present application. The embodiments described below by referring to the drawings are exemplary and intended for the purpose of explaining the present application and are not to be construed as limiting the present application. All other embodiments, which can be made by one of ordinary skill in the art without undue burden from the present disclosure, are within the scope of the present disclosure. Embodiments of the present application are described in detail below with reference to the accompanying drawings.

It should be noted that in the description of the present invention, the terms "first," "second," and the like are used for descriptive purposes only and are not to be construed as indicating or implying relative importance.

The NFV resource scheduling method shown in fig. 1 includes:

step 1, acquiring original request data and dynamic network topology information;

step 2, obtaining service request characteristics according to the original request data;

step 3, obtaining topological characteristics according to the dynamic network topological information;

step 4, fusing the service request characteristics and the topology characteristics to form fusion characteristics;

step 5, acquiring a trained resource allocation scheduling model;

and 6, inputting the fusion characteristics into a resource allocation scheduling model so as to acquire scheduling processing action information.

The NFV resource scheduling method has the following advantages:

1 in the NFV resource allocation problem, three sub-problems are solved in a coordinated scheme for the first time, and a new thought is provided for NFV-RA problem research. Specifically, the coordinated method attempts to execute multiple phases as a whole in one step. Previous techniques have addressed one or both sub-problems in a non-coordinated manner. The uncoordinated approach is to solve the problems sequentially, but does not consider the dependency between each problem, that is, the output of one stage is the input of the next stage, and if the sequential solution is adopted, there may be a solution as a whole, but since the solution generated by solving the first problem is taken as the input of the next problem, there is no solution. NFV resource allocation problems can be seen as a mapping: mapping the directed graph representing the service request into the undirected graph representing the physical network, and performing scheduling execution. Therefore, after the obtained characteristics are fused, the obtained action information is a sub-graph of a physical network, wherein the sub-graph comprises node information and link information. Thus we solve three sub-problems in a single step, i.e. coordinated fashion.

In this embodiment, the obtaining the service request feature according to the original request data includes: service request features in the original request data are acquired through a multi-head self-attention mechanism.

Specifically, the self-attention mechanism is utilized to extract the characteristics of the original request data, and the service request characteristics of the original request data are fully mined. In NFV, service requests usually come in the form of Service Function Chains (SFCs) where VNFs have dependencies, in order to relate VNFs of different locations to calculate a representation of the service request, the present algorithm introduces location coding in the self-attention mechanism. Meanwhile, in order to extract information of different representing subspaces, the method uses multiple heads of self-attentions, uses different linear mappings to project the original request to different dimensions, performs self-attentions calculation respectively, and finally performs multiple heads of serial connection, thereby obtaining the final service request characteristics.

In the present embodiment, in the natural language processing NLP, each word of the input character string is changed into a vector using the ebedding algorithm. However, the present application is directed to NFV, where service requests have non-negligible features, such as the number of modules involved, type, qoS requirements, time of arrival, requested data rate, etc., and the present application incorporates the features of the number of modules, type, qoS requirements, time of arrival, requested data rate, etc. in addition to encoding the service request as input to the self-attention mechanism; in addition, in the present embodiment, position coding is introduced in the self-attention mechanism. In NLP, position coding represents word order in the input sequence, however, service requests are not only simple order relations, but there are parallel and segmentation modules, so we code the positions of parallel and segmentation modules specifically based on position coding.

In this embodiment, obtaining the topology feature according to the dynamic network topology information includes: and acquiring topological characteristics in the dynamic network topological information through the double-component graph convolutional neural network.

Specifically, feature extraction of dynamic topology is performed, a double-component graph convolution is introduced to explicitly model the correlation of nodes and links, and automatic feature extraction is performed on the network topology of non-Euclidean domain.

In this embodiment, after inputting the fusion feature to the resource allocation scheduling model to obtain scheduling processing action information, the NFV resource scheduling method further includes:

and controlling the NFV resources to schedule the NFV environment according to the scheduling processing action information, and acquiring evaluation information returned by the NFV environment after the scheduling processing action information is scheduled.

In the present embodiment, the evaluation information includes two kinds: 1. because the scheduling processing action information is composed of the generated countermeasure network, the action generated by the generator is not necessarily the real physical network, so a discriminator is required to judge whether the action is the real physical network, thereby giving the first evaluation information; 2. in order to judge the degree of the generated action, an evaluation information is required for guidance. This assessment information is referred to as rewards (Reward) in reinforcement learning, and is returned after an action is performed by the environment. The reward in the present invention is designed with the goals and constraints of the mathematical problem formalized by NFV resource allocation.

In the present embodiment, the evaluation information includes evaluation of scheduling processing operation information and index information.

In this embodiment, after obtaining the evaluation information returned by the NFV environment after the scheduling by the scheduling processing action information, the NFV resource scheduling method further includes:

In this embodiment, the resource allocation scheduling model is a generative countermeasure model composed of a generator and a arbiter, and the topology features and the service request features are used as inputs, and the true network sub-topology is used as a sample, and the sub-topology is generated by input. Specifically, the generator takes samples from the a priori distribution (i.e., the true network sub-topology) and generates a topology G with node and link information representing the topology. The nodes and edges of G are respectively associated with nodes representing VNFs to be embedded and instance information and link bearer information and traffic information in the nodes. The arbiter extracts two samples from the dataset and the generator and learns how to distinguish them. The generator and the arbiter are trained using the improved WGAN, causing the generator to learn the matching empirical distribution and ultimately output an effective topology.

In this embodiment, the multi-headed self-attention mechanism, the two-component graph convolutional neural network, and the resource allocation scheduling model described above constitute a single agent. The method adopts a time slot mode when the prior resource allocation method processes the request, and possibly causes long waiting time of the request, so that the method adopts multiple agents, the number of the specific agents is determined by probability distribution of service request arrival, and zero waiting processing after the service request arrives can be basically ensured. The use of multiple agents may allow for reduced latency of requests.

Because model early training is unstable and is not easy to converge, and the data obtained by solving the simplified model by using a solver can accelerate the early training, the model is adopted to assist and simulate learning, so that the model can converge to a better strategy more quickly.

Also, it is considered that the experience of an agent from the environment can be generally slow, and the agent must wait for the environment to perform an operation before a training step. Second, the states and actions in one track are highly correlated, thereby reducing the robustness and efficiency of the training process. We inspired by A3C to use parallel training to accelerate the training process while enhancing its robustness. We use multiple worker agents to collect experiences from their own environment and send them to a central agent that is responsible for training and updating network parameters. For example, if parallel training is used in a single agent as described above, this single agent acts as a central master agent, after which this agent is replicated several times as a plurality of worker agents; the same is true for the multi-agent, except that the central master agent is the multi-agent.

The present application is described in further detail below by way of examples, which are not to be construed as limiting the present application in any way.

According to the NFV resource scheduling method provided by the embodiment, a Dell Precision T7920 tower workstation is selected as a hardware platform, and programming is performed by using a Python language.

Firstly, the resource scheduling system initializes the relevant interfaces, and when receiving the service request arriving in real time, the resource scheduling system starts the processing flow of the request.

The service request analysis function request_player () extracts original request data from DDR3 mounted under a workstation;

obtaining each sub-module in the parsed request, and extracting the dependency relationship between the original feature and each sub-module by utilizing an initial feature extraction function init_feature ();

then taking the characteristic and the dependency relationship as the input of a self_attribute () function to obtain the extracted characteristic of the service request, namely the service request characteristic;

meanwhile, feature extraction is carried out on the dynamic network topology by utilizing the convolutional neural network of the double-component graph, so as to obtain topology features;

and fusing the obtained topological characteristic with the service request characteristic, and taking the fused topological characteristic and the service request characteristic as the input of a resource allocation scheduling model to obtain scheduling processing action information aiming at the request and combining with the consideration of the network state.

After the NFV environment processes the generated scheduling processing action information, the NFV environment returns an evaluation of the action and various indexes (such as success or failure, delay, etc.), and the resource allocation scheduling model adjusts the neural network parameters according to the returned values. Specifically, the parameters are adjusted according to the multi-agent reinforcement learning madppg algorithm.

The application also provides an NFV resource scheduling device, which comprises an original request data acquisition module, a dynamic network topology information acquisition module, a service request feature acquisition module, a topology feature acquisition module, a fusion module, a model acquisition module and a model calculation module, wherein the original request data acquisition module is used for acquiring original request data; the dynamic network topology information acquisition module is used for acquiring dynamic network topology information; the service request feature acquisition module is used for acquiring service request features according to the original request data; the topology characteristic acquisition module is used for acquiring topology characteristics according to the dynamic network topology information; the fusion module is used for fusing the service request characteristics and the topology characteristics to form fusion characteristics; the model acquisition module is used for acquiring a trained resource allocation scheduling model; the model calculation module is used for inputting the fusion characteristics into the resource allocation scheduling model so as to acquire scheduling processing action information.

The application also provides an NFV resource scheduling system, which comprises a plurality of NFV resource scheduling devices, wherein each NFV resource scheduling device is used for realizing the NFV resource scheduling method.

The above-mentioned NFV resource scheduling method is a processing flow of each NFV resource scheduling device, and similarly, when the NFV resource scheduling system of the present application is adopted, the NFV resource scheduling devices cooperate with each other in addition to processing the respective requests, and the receiving rate of the request processing is increased in consideration of the actions of each other.

The above description of the method is equally applicable to the description of the device and the system.

Fig. 2 is an exemplary block diagram of an electronic device capable of implementing the NFV resource scheduling method provided according to one embodiment of the present application.

As shown in fig. 2, the electronic device includes an input device 501, an input interface 502, a central processor 503, a memory 504, an output interface 505, and an output device 506. The input interface 502, the central processing unit 503, the memory 504, and the output interface 505 are connected to each other through a bus 507, and the input device 501 and the output device 506 are connected to the bus 507 through the input interface 502 and the output interface 505, respectively, and further connected to other components of the electronic device. Specifically, the input device 501 receives input information from the outside, and transmits the input information to the central processor 503 through the input interface 502; the central processor 503 processes the input information based on computer executable instructions stored in the memory 504 to generate output information, temporarily or permanently stores the output information in the memory 504, and then transmits the output information to the output device 506 through the output interface 505; the output device 506 outputs the output information to the outside of the electronic device for use by the user.

That is, the electronic device shown in fig. 2 may also be implemented to include: a memory storing computer-executable instructions; and one or more processors that, when executing the computer-executable instructions, can implement the NFV resource scheduling method described in connection with fig. 1.

In one embodiment, the electronic device shown in FIG. 2 may be implemented to include: a memory 504 configured to store executable program code; the one or more processors 503 are configured to execute the executable program code stored in the memory 504 to perform the NFV resource scheduling method in the above embodiment.

In one typical configuration, a computing device includes one or more processors (CPUs), input/output interfaces, network interfaces, and memory.

The memory may include volatile memory in a computer-readable medium, random Access Memory (RAM) and/or nonvolatile memory, such as Read Only Memory (ROM) or flash memory (flash RAM). Memory is an example of computer-readable media.

Computer-readable media include both permanent and non-permanent, removable and non-removable media, and the media may be implemented in any method or technology for storage of information. The information may be computer readable instructions, data structures, modules of a program, or other data. Examples of storage media for a computer include, but are not limited to, phase change memory (PRAM), static Random Access Memory (SRAM), dynamic Random Access Memory (DRAM), other types of Random Access Memory (RAM), read Only Memory (ROM), electrically Erasable Programmable Read Only Memory (EEPROM), flash memory or other memory technology, compact disc read only memory (CD-ROM), digital Versatile Discs (DVD) or other optical storage, magnetic cassettes, magnetic tape disk storage or other magnetic storage devices, or any other non-transmission medium, which can be used to store information that can be accessed by a computing device.

It will be appreciated by those skilled in the art that embodiments of the present application may be provided as a method, system, or computer program product. Accordingly, the present application may take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, the present application may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, and the like) having computer-usable program code embodied therein.

Furthermore, it is evident that the word "comprising" does not exclude other elements or steps. A plurality of units, modules or means recited in the apparatus claims can also be implemented by means of software or hardware by means of one unit or total means. The terms first, second, etc. are used to identify names, and not any particular order.

The flowcharts and block diagrams in the figures illustrate the architecture, functionality, and operation of possible implementations of systems, methods and computer program products according to various embodiments of the present application. In this regard, each block in the flowchart or block diagrams may represent a module, segment, or portion of code, which comprises one or more executable instructions for implementing the specified logical function(s). It should also be noted that, in some alternative implementations, the functions noted in the block may occur out of the order noted in the figures. For example, two blocks shown in succession may, in fact, be executed substantially concurrently, or the blocks may sometimes be executed in the reverse order, depending upon the functionality involved. It will also be noted that each block of the block diagrams and/or flowchart illustration, and combinations of blocks in the block diagrams and/or flowchart illustration, can be implemented by special purpose hardware-based systems which perform the specified functions or acts, or combinations of special purpose hardware and computer instructions.

The processor referred to in this embodiment may be a central processing unit (Central Processing Unit, CPU), or other general purpose processor, digital signal processor (Digital Signal Processor, DSP), application specific integrated circuit (Application Specific Integrated Circuit, ASIC), off-the-shelf programmable gate array (Field-Programmable Gate Array, FPGA) or other programmable logic device, discrete gate or transistor logic device, discrete hardware components, or the like. A general purpose processor may be a microprocessor or the processor may be any conventional processor or the like.

The memory may be used to store computer programs and/or modules, and the processor may perform various functions of the apparatus/terminal device by executing or executing the computer programs and/or modules stored in the memory, and invoking data stored in the memory. The memory may mainly include a storage program area and a storage data area, wherein the storage program area may store an operating system, an application program (such as a sound playing function, an image playing function, etc.) required for at least one function, and the like; the storage data area may store data (such as audio data, phonebook, etc.) created according to the use of the handset, etc. In addition, the memory may include high-speed random access memory, and may also include non-volatile memory, such as a hard disk, memory, plug-in hard disk, smart Media Card (SMC), secure Digital (SD) Card, flash Card (Flash Card), at least one disk storage device, flash memory device, or other volatile solid-state storage device.

In this embodiment, the modules/units of the apparatus/terminal device integration may be stored in a computer readable storage medium if implemented in the form of software functional units and sold or used as a separate product. Based on such understanding, the present invention may implement all or part of the flow of the method of the above embodiment, or may be implemented by hardware related to the instructions of a computer program, where the computer program may be stored in a computer readable storage medium, and when executed by a processor, the computer program may implement the steps of each of the method embodiments described above. Wherein the computer program comprises computer program code, which may be in the form of source code, object code, executable files or in some intermediate form, etc. The computer readable medium may include: any entity or device capable of carrying computer program code, a recording medium, a U disk, a removable hard disk, a magnetic disk, an optical disk, a computer Memory, a Read-Only Memory (ROM), a random access Memory (RAM, random Access Memory), an electrical carrier signal, a telecommunications signal, a software distribution medium, and so forth.

It should be noted that the content of the computer readable medium can be appropriately increased or decreased according to the requirements of the legislation and the practice of the patent in the jurisdiction. While the preferred embodiments have been described, it will be understood by those skilled in the art that various changes and modifications may be made without departing from the spirit and scope of the invention, and it is intended that the scope of the invention shall be limited only by the claims appended hereto.

While the invention has been described in detail in the foregoing general description and with reference to specific embodiments thereof, it will be apparent to one skilled in the art that modifications and improvements can be made thereto. Accordingly, such modifications or improvements may be made without departing from the spirit of the invention and are intended to be within the scope of the invention as claimed.

Claims

1. The NFV resource scheduling method is characterized by comprising the following steps:

acquiring original request data and dynamic network topology information;

acquiring a trained resource allocation scheduling model;

inputting the fusion characteristics into the resource allocation scheduling model so as to acquire scheduling processing action information;

the obtaining service request features according to the original request data includes:

service request characteristics in the original request data are acquired through a multi-head self-attention mechanism, wherein,

position coding is introduced into a self-attention mechanism, multi-head self-attention is used, different linear mapping is used for projecting an original request to different dimensions, self-attention calculation is respectively carried out, and multi-head serial connection is finally carried out, so that a final service request characteristic is obtained;

the step of obtaining topology features according to the dynamic network topology information includes:

the topological feature in the dynamic network topological information is obtained through a double-component graph convolutional neural network, wherein,

feature extraction of dynamic topology is performed, double-component graph convolution is introduced to explicitly model the correlation of nodes and links, and automatic feature extraction is performed on the network topology of non-Euclidean domain.

2. The NFV resource scheduling method of claim 1, wherein after the inputting the fusion feature to the resource allocation scheduling model to obtain scheduling process action information, the NFV resource scheduling method further comprises:

3. The NFV resource scheduling method of claim 2, wherein the evaluation information includes an evaluation of the scheduling processing action information and index information.

4. The NFV resource scheduling method of claim 3, wherein after acquiring the evaluation information returned by the NFV environment after scheduling by the scheduling processing action information, the NFV resource scheduling method further comprises:

5. An NFV resource scheduling apparatus, wherein the NFV resource scheduling apparatus includes:

the model calculation module is used for inputting the fusion characteristics into the resource allocation scheduling model so as to acquire scheduling processing action information; wherein,,

6. An NFV resource scheduling system, characterized in that the NFV resource scheduling system comprises a plurality of NFV resource scheduling apparatuses according to claim 5.

7. An electronic device comprising a memory, a processor, and a computer program stored in the memory and capable of running on the processor, wherein the processor implements the NFV resource scheduling method of any one of claims 1-4 when executing the computer program.

8. A computer readable storage medium storing a computer program, wherein the computer program is capable of implementing the NFV resource scheduling method according to any one of claims 1 to 4 when executed by a processor.