CN115037590A

CN115037590A - Network virtualization system structure and virtualization method

Info

Publication number: CN115037590A
Application number: CN202210298990.6A
Authority: CN
Inventors: 匡立伟; 尹山; 徐安然; 李文超
Original assignee: Fiberhome Telecommunication Technologies Co Ltd
Current assignee: Fiberhome Telecommunication Technologies Co Ltd
Priority date: 2022-03-25
Filing date: 2022-03-25
Publication date: 2022-09-09
Anticipated expiration: 2042-03-25
Also published as: CN115037590B; WO2023179180A1

Abstract

The invention relates to a network virtualization architecture and a virtualization method. The network virtualization system structure part mainly comprises a physical space, a virtual space and an application space, wherein: the physical space comprises computing resources, storage resources and transmission resources, the fusion of the computing, storage and transmission resources is realized, and the physical space also reports equipment resource state data to the virtual space; the virtual space realizes the virtualization of each resource and describes the resource in a virtualization model so as to present the resource to the application space; the virtual space also issues a demand instruction to specific physical equipment or virtual equipment for execution after receiving the demand issued by the application space; the application space obtains a virtualization model described by the virtual space and provides the virtualization model for various service scenes; the application space also accepts the requirements, integrates the requirements and issues the requirements to the virtual space. The invention can support the deep fusion of computing, storing and transmitting resources and the overall optimal overall scheduling of resources.

Description

Network virtualization system structure and virtualization method

Technical Field

The present invention relates to the field of communications technologies, and in particular, to a network virtualization architecture and a virtualization method.

Background

The development of digital economy needs information technology and communication technology support, and the digital transformation requirement of the industry also pushes the traditional communication network architecture to evolve towards a 'cloud network convergence' architecture. A communication network architecture constructed by a traditional mode of physical equipment and professional network management is difficult to meet the digital transformation requirements of various industries on resource demand, flexible management and control, safety and reliability. Based on virtualization technology, a computing, storing and transmitting resource pool is constructed, a novel cloud network fusion framework is designed, an artificial intelligence technology is introduced to realize whole network business arrangement and end-to-end efficient resource cooperative control, the safety and the reliability of the virtual resource pool are guaranteed based on an endogenous safety technology, dynamic diversified business requirements of various industries are met, and the method becomes a network framework evolution direction.

In recent years, a series of researches have been conducted by the industry and academia on the technology of "cloud network fusion". Until now, related research results can support cloud network collaboration, but an architecture for supporting cloud network convergence is still lacking. The information technology includes two large technical systems, one is an information processing technical system represented by cloud computing, and the other is an information transmission technical system represented by a network. The information processing technology focuses on researching efficient algorithms to achieve large-scale high-performance data processing and mining valuable information from data, and the technology comprises distributed computing, parallel algorithms, data mining, knowledge discovery and the like. The information transmission technology is focused on researching connection protocols and high-performance devices, and supports data message transmission with higher speed, larger capacity and longer distance, and the technologies comprise network protocols, digital signal processing, optical amplification technologies, error correction coding technologies and the like. Scholars and engineers of information processing technology systems are biased towards perfecting network technology so that the network technology can support parallel distributed data processing, thereby providing more efficient computing and storage resources for users. Scholars and engineers in information transfer technology systems are biased towards enhancing computing storage hardware and software to support more stable and robust communication network construction. The two cloud network coordination modes can be abstractly summarized into 'perfecting transmission resources, enabling the network to better coordinate with the cloud', 'enhancing computing storage resources, and enabling the cloud to better coordinate with the network'. Neither of the two coordination methods puts the computing storage transmission resources in an equal position, which also results in a shallow coordination mode at present, and the cloud and the network are still in an isolated state. This isolation state causes the user to still need to make two types of resource requests when using the computing storage transmission resource, the first type is to make the computing storage resource request to the cloud computing center, and the second type is to make the transmission resource request to the communication network. This shallow cloud collaboration technique has two significant drawbacks. Firstly, the user needs to coordinate the technical detail indexes of the two types of resources at the same time, so that the user is difficult to concentrate on the business process. Secondly, whether the network is a better cloud coordination network or the cloud is a better cloud coordination network, the cloud network resources are difficult to realize overall scheduling, and the computing, storing and transmitting resources locally reach the optimal utilization rate, but the overall optimization cannot be guaranteed.

The cloud network resources are required to be put in the same position to really realize the cloud network convergence, and an efficient system structure is constructed to quantitatively describe the computing storage transmission resources. Two major difficulties are faced in building such an architecture. Firstly, the technology which is widely applied to the cloud computing center and the communication network at present cannot be completely abandoned, so that the smooth evolution of the cloud network cannot be realized, and huge waste of the established resources can be caused. Secondly, the cloud network resources cannot be improved and enhanced in a splitting manner, and then a global controller is deployed on an independent cloud computing center and a communication network to realize the shallow-level cooperation of computing, storing and transmitting. In the current related art research, a system structure for supporting the deep fusion of cloud network resources is not constructed from the global perspective, and an efficient processing method for realizing the uniform quantitative description of computing, storing and transmitting resources is not provided.

In view of the above circumstances, how to overcome the defects in the prior art, and solve the technical problems of difficult heterogeneous resource fusion, weak overall network cooperation degree, and poor resource allocation efficiency, are the problems to be solved in the technical field.

Disclosure of Invention

Aiming at the defects or the improvement requirements of the prior art, the invention provides a network virtualization system structure and a virtualization method, comprehensively analyzes and summarizes the related technologies of a cloud computing center and a communication network, constructs a novel system structure based on the network virtualization technology, uniformly describes computing, storing and transmitting resources, and supports the integration of the resources in a physical space, a virtual space and an application space. The network virtualization system structure provided by the invention not only accords with the technical trend of smooth evolution of the current cloud network architecture, but also can support deep fusion of computing, storing and transmitting resources, simultaneously supports a user to simultaneously apply for required resources in a unified quantization space, and supports global optimal overall scheduling of the resources.

The embodiment of the invention adopts the following technical scheme:

in a first aspect, the present invention provides a network virtualization architecture, comprising a physical space, a virtual space, and an application space, wherein:

the physical space comprises computing resources, storage resources and transmission resources, the fusion of the computing, storage and transmission resources is realized, and the physical space also reports equipment resource state data to the virtual space;

the virtual space realizes the virtualization of each resource and describes the resource in a virtualization model so as to present the resource to the application space; the virtual space also issues a demand instruction to specific physical equipment or virtual equipment for execution after receiving a demand issued by the application space;

the application space obtains a virtualization model described by the virtual space and provides the virtualization model for various service scenes; the application space also accommodates the requirements, integrates the requirements and issues the requirements to the virtual space.

Furthermore, the virtual space comprises an orchestrator, a controller, and virtual computing resources, virtual storage resources, and virtual transmission resources corresponding to the physical space; the application space comprises user application, computing capacity requirements, storage capacity requirements and transmission capacity requirements; the interfaces of the physical space and the virtual space comprise an equipment state interface and a management and control instruction interface; the interfaces between the application space and the virtual space comprise a cloud network capacity interface and a user requirement interface.

Further, the physical space, the virtual space, the application space, the device state interface, the management and control instruction interface, the cloud network capability interface, and the user requirement interface form a closed loop, specifically:

the integration of computing capacity requirements, storage capacity requirements and transmission capacity requirements is realized in the application space, and after a user puts forward resource requirements, the application space is matched to find corresponding virtual resources, and the corresponding virtual resources are sent to the virtual space through a user requirement interface;

the virtual space realizes the fusion of virtual computing resources, virtual storage resources and virtual transmission resources, and after receiving user requirements, the virtual space is processed by the orchestrator and the controller to form a specific operation instruction which is issued to specific physical equipment or executed on the virtual equipment through a control instruction interface;

the physical space realizes the fusion of three types of entity resources, namely computing resources, storage resources and transmission resources, wherein one part of state data in a service disk of the physical space is processed in the computing storage disk, and the other part of the state data is reported to the virtual space through an equipment state interface;

through the reported equipment resource state data, the virtual space can master the performance conditions of computing resources, storage resources and transmission resources in the physical space in real time, the computing, storage and transmission resource capacities of the virtual space are uniformly described in the virtualization model, and are presented to the application space through the cloud network capacity interface, and the application space is provided for various service scenes.

Further, the process of implementing automatic fusion scheduling of computing storage and transmission resources by the physical space, the virtual space, and the user space includes:

making a request, and submitting request description in an application space;

applying a space analysis request to obtain calculation, storage and transmission resource quantization indexes, and describing the resource quantization indexes in a virtualization model;

after the resource quantization indexes are uniformly described, the resource quantization indexes are converted into specific operation instructions by the orchestrator and the controller of the virtual space;

establishing a virtual model based on a virtualization technology, instantiating the virtual model, verifying an operation instruction, and issuing equipment after the operation instruction is accurate and correct;

the computing, storing and transmitting equipment receives and executes various operation instructions.

Further, the process of implementing the fusion of three types of entity resources, i.e., the computing resource, the storage resource, and the transmission resource, in the physical space specifically includes:

one or more computing storage disks are arranged in a physical space, and the physical space also comprises a master control disk and a plurality of service disks;

installing and deploying basic software, wherein the basic software comprises one or more of an operating system, data tool software, a digital twin modeling framework, an intelligent training reasoning framework and an algorithm analysis processing framework; the data tool software calls a function provided by the operating system to access one or more of a central processing unit, a memory, a disk and peripheral equipment; a digital twin modeling framework, an intelligent training reasoning framework and an algorithm analysis processing framework call data tool software to obtain various data;

the service disk is used for acquiring equipment state data, and if the acquired data is related to the network element, the data is uploaded to the computing storage disk for edge processing; if the network element is related to other network elements, uploading the network element to a master control panel, and transferring the network element to a control platform by the master control panel; after receiving the equipment state data, the computing storage disk is allocated according to the data processing requirement, and if the simulation modeling operation is carried out, the data are forwarded to the digital twin modeling framework; if the edge artificial intelligence reasoning is carried out, the data are forwarded to an intelligent training reasoning framework; and if alarm correlation analysis or performance degradation analysis operation is carried out, forwarding the data to an algorithm analysis processing framework.

Further, the virtualization model includes a three-layer three-plane network virtualization model and a seventh order tensor model, wherein:

the three-layer three-side network virtualization model comprises three types of resources, a three-layer space and three technologies, wherein the three types of resources comprise computing resources, storage resources and transmission resources, the three-layer space comprises a physical space, a virtual space and an application space, and the three technologies comprise a virtualization technology, an intelligent operation and maintenance technology and an endogenous safety technology; the three types of resources are positioned in the three-layer space, and the three technologies act on the three-layer space and the three types of resources;

the seven-order tensor model is used for uniformly describing and quantitatively representing three types of resources in a three-layer space, and the three types of technologies are used for processing various elements in the seven-order tensor model and comprise the following steps: the virtualization technology is used for realizing virtualization of physical space computing resources, storage resources and transmission resources, the intelligent operation and maintenance technology is used for realizing cross-domain overall scheduling of the virtualized resources, and a safe and reliable virtualization system is built based on the endogenous safety technology.

Further, the seventh-order tensor model is described as T epsilon R ^{I1×I2×I3×I4×I5×I6×I7} Where R denotes the real number domain, I1, I2, I3, I4, I5, I6, I7 denote seven orders of the tensor model, representing time, physical space, virtual space, application space, computational resources, storage resources, transfer resources, respectively.

In a second aspect, the present invention further provides a virtualization method, which virtualizes each resource by using a virtualization technology, where the virtualization technology includes a functional fidelity technology and a functional simulation technology, the functional fidelity technology includes a scheduling fidelity technology and a virtual fidelity technology, and the functional simulation technology includes a functional mapping simulation technology and a functional fitting simulation technology; and for a specific physical entity resource, selecting according to the priority of the scheduling fidelity technology, the virtual fidelity technology, the function mapping simulation technology and the function fitting simulation technology when selecting the virtualization technology.

Further, when the virtualization technology is selected, three transition modes, namely P transition, F transition and FP transition, are also included, specifically:

if the operation mechanism of a certain physical entity cannot be known, a function fitting simulation technology is needed to be adopted for virtualization; when the operation mechanism of the physical entity can be completely mastered, virtualization is carried out by adopting a function mapping simulation technology in a P transition mode; when the virtualized entity constructed based on the function mapping simulation technology can really bear the functions of the physical entity, virtualization is carried out by adopting a virtual fidelity technology in an F transition mode; and if the conditions of P transition and F transition are met simultaneously, performing virtualization by adopting a scheduling fidelity technology in an FP transition mode.

Further, when the physical entity is virtualized and modeled, the method further comprises optimizing the constructed virtualized instance, specifically:

determining a physical entity deployed in a real network, and uploading equipment state data of the physical entity and configuration data of the physical entity to a virtual space;

according to the state data and the configuration data of the relevant equipment uploaded by the physical entity, combing the running mechanism of the entity, and selecting a corresponding virtualization technology to construct a virtualization model;

calling a virtualization model according to the application space service scene requirement, transmitting real-time data, creating a virtualization instance, and constructing a virtual network;

after the application space executes various application scene specific tasks, the virtualization effect is fed back, and the virtualization model is optimized according to the virtualization effect.

Compared with the prior art, the invention has the beneficial effects that:

(1) the invention provides a novel network virtualization system structure, which realizes cross-domain unified scheduling and control of computing resources, storage resources and transmission resources, supports end-to-end global arrangement of services, supports intelligent operation and maintenance of resources, and realizes deep fusion of a cloud computing environment and a communication network through a set of integral network virtualization system structure. The invention adopts a novel tensor-based three-layer three-surface network virtualization model and a seven-order tensor model, uniformly describes, calculates, stores and transmits resources in a high-order multi-dimensional tensor space, clearly defines a physical space, a virtual space and an application space, introduces a virtualization technology, an intelligent operation and maintenance technology and an endogenous safety technology, and solves the problems of poor resource demand allocation capability, difficult resource cross-domain coordination, time-consuming and low efficiency of resource operation and maintenance of the current network system architecture.

(2) The invention provides two major network virtualization technologies, namely four minor network virtualization technologies, and a high-precision virtualization model can be constructed by selecting a proper virtualization technology according to the characteristics of different physical entities. Based on the virtualization models, the virtualization network is created and service verification or resource operation is carried out for the user requirements of each application space, and automatic processing of network services, automatic maintenance of network resources and automatic optimization of network performance can be realized. The invention also adopts a virtualization instance automatic optimization mechanism, optimizes the virtualization instance through the physical space, the virtual space and the application space cross-layer collaborative feedback technology, and improves the simulation degree of the virtualization model to the physical entity.

(3) The computing storage transmission resource fusion method provided by the invention can realize the deep fusion of resources in a physical space, a virtual space and an application space respectively. The physical space supports the splitting of data between a single compute storage disk and a single transport service disk. The virtual space deploys a service orchestrator and a resource controller, supports global service orchestration and cross-domain resource control, supports creation, combination and optimization of a virtualized instance, and realizes that network capacity is opened as required through the virtualized instance. In the application space, a user can realize simple and accurate request for calculating, storing and transmitting resources without knowing too much technical background, so that the user can pay more attention to the service, the perception of the user is improved, and the accuracy and consistency of the resource request can be ensured.

Drawings

In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required to be used in the embodiments of the present invention will be briefly described below. It is obvious that the drawings described below are only some embodiments of the invention, and that for a person skilled in the art, other drawings can be derived from them without inventive effort.

Fig. 1 is a schematic diagram of a network virtualization architecture for computing storage transport resource fusion according to embodiment 1 of the present invention;

fig. 2 is a schematic diagram of a network virtualization technology tree according to embodiment 3 of the present invention;

fig. 3 is a schematic diagram of the optimization of the physical entity virtualization modeling and virtualization example provided in embodiment 4 of the present invention;

fig. 4 is a schematic diagram illustrating fusion of computing storage transmission resources in a physical virtual application space according to embodiment 5 of the present invention;

fig. 5 is a schematic diagram of resource automatic fusion scheduling based on virtualization technology according to embodiment 6 of the present invention;

fig. 6 is a schematic diagram of fusion of computing storage and transfer resources in a physical space according to embodiment 7 of the present invention.

Detailed Description

In order to make the objects, technical solutions and advantages of the present invention more apparent, the present invention is described in further detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and do not limit the invention.

The present invention is a system structure of a specific function system, so the functional logic relationship of each structural module is mainly explained in the specific embodiment, and the specific software and hardware implementation is not limited.

In addition, the technical features involved in the embodiments of the present invention described below may be combined with each other as long as they do not conflict with each other. The invention will be described in detail below with reference to the figures and examples.

Example 1:

the core of the cloud network integration is to realize the integration of cloud computing center resources and network resources. The cloud computing resources comprise computing resources and storage resources, and a central processing unit, an artificial intelligence chip and the like belong to the computing resources. Storage arrays, hard disks, memories, etc. all belong to storage resources. The network resources mainly refer to transmission resources, including optical fibers, network cables, communication equipment single disks, optical modules, optical devices, network protocols and the like. In other words, the core of the cloud network convergence is to realize the convergence of three types of resources of computing, storing and transmitting, and the cloud network convergence aims to efficiently, flexibly, intelligently and safely provide the computing, storing and transmitting resources for users. In current cloud computing centers and communication networks, three types of resources, computing, storage, and transmission, exist in physical space in the form of software and hardware entities. And the digital application of each industry adopts the form of abstract capability to use computing, storing and transmitting resources in a user space. The physical resources in the physical space are converted and provided to the user in the form of computing capacity, storage capacity and transmission capacity, and therefore the computing, storage and transmission resources need to be simulated and modeled based on a virtualization technology in the virtual space, and a scheduler and a controller are deployed for scheduling. The virtualization technology realizes abstract modeling of entity resources, the resources are efficiently and flexibly provided for users in a user space, but the problems of complexity and safety of resource operation and maintenance are also caused, and an intelligent operation and maintenance technology and an endogenous safety technology need to be introduced for support.

Based on the description and aiming at the problems of difficult heterogeneous resource fusion, weak whole network cooperation degree and poor resource allocation efficiency faced by the traditional communication network, the embodiment of the invention provides a novel network virtualization system structure, which realizes the deep fusion and cooperative scheduling of computing resources, storage resources and transmission resources, supports the fusion and scheduling of the three types of resources in physical space, virtual space and application space, and realizes the high-efficiency, flexible, intelligent and safe resource supply targets based on virtualization technology, intelligent operation and maintenance and endogenous safety technology.

Embodiment 1 of the present invention provides a network virtualization architecture for computing storage and transport resource fusion, which includes a physical space, a virtual space, and an application space, where: the physical space realizes the fusion of computing, storing and transmitting resources and reports the state data of equipment resources to the virtual space; the virtual space realizes the virtualization of each resource and describes the resource in a virtualization model so as to present the resource to the application space; the virtual space also issues a demand instruction to specific physical equipment or virtual equipment for execution after receiving the demand issued by the application space; the application space obtains a virtualization model described by the virtual space and provides the virtualization model for various service scenes; the application space also accepts the requirements, integrates the requirements and issues the requirements to the virtual space.

The preferred embodiment proposes two types of virtualization technologies, namely, a functional fidelity virtualization technology (functional fidelity technology) and a functional simulation virtualization technology (functional simulation technology), which are virtualization technologies used for implementing virtualization of each resource in the virtual space. The input and output of the virtualization model constructed based on the functional fidelity technology are consistent with the physical entity model, the internal processing mechanism is consistent with the physical entity, and the hardware operation is simulated only by a software method. The input and output of the virtualization model constructed based on the functional simulation technology are not completely consistent with the physical entity model, and the virtualization model is mainly used for verifying, monitoring and optimizing the functions and performance of the physical entity by simulating the running mode of approaching the physical entity through software.

The functional fidelity virtualization technology proposed in this embodiment is further subdivided into two specific technologies, namely, scheduling fidelity technology and virtual fidelity technology. The scheduling fidelity technology is to virtualize a plurality of logical network slices in a physical network, wherein each slice meets the service requirement of a specific type. The virtual fidelity technology is that network element functions based on special hardware are virtualized and then deployed on general hardware to form a virtual network element pool.

The function simulation virtualization technology provided by the embodiment is also subdivided into two specific technologies, namely, a function mapping simulation technology and a function fitting simulation technology. The function mapping simulation technology translates and maps the communication mechanism of the physical space network entity, and then constructs a corresponding digital twin model in a virtual space. The function fitting simulation technology mainly aims at network functions which are difficult to accurately describe communication mechanisms, and performs fitting based on input and output data to construct a digital twin model. In a specific implementation process, a selection can be made from the virtualization technologies according to the characteristics of the physical entity to create a virtualized resource.

As shown in fig. 1, the virtualization model in the present embodiment includes a three-layer three-plane network virtualization model and a seventh order tensor model. Specifically, a Tensor-based Three-Layer Three-sided network virtualization Model (as shown in the left side of fig. 1), which is called a sensor based Three Layer and Three Plane Model in english, and is abbreviated in englishIs TL ³ P ³ The three-layer three-side network virtualization model comprises three types of resources, a three-layer space and three technologies, wherein the three types of resources are located in the three-layer space, the three technologies simultaneously act on the three-layer space and the three types of resources, the three types of resources comprise computing resources, storage resources and transmission resources, the three-layer space comprises a physical space, a virtual space and an application space, and the three technologies comprise a virtualization technology, an intelligent operation and maintenance technology and an endogenous safety technology.

The network virtualization architecture of this embodiment further includes a seventh-order tensor model (as shown in the right side of fig. 1) corresponding to the three-layer three-sided network virtualization model, where the seventh-order tensor model performs unified description and quantitative representation on three types of resources located in the three-layer space, and the three types of technologies are used to process each element in the seventh-order tensor model. The method comprises the following steps: the virtualization technology is used for realizing virtualization of physical space computing resources, storage resources and transmission resources, the intelligent operation and maintenance technology is used for realizing cross-domain overall scheduling of the virtualized resources, and a safe and reliable virtualization system is built based on the endogenous safety technology.

In the preferred embodiment, the seventh order tensor model is described as T ∈ R ^{I1×I2×I3×I4×I5×I6×I7} Where R denotes the real number domain, I1, I2, I3, I4, I5, I6, I7 denote seven orders of the tensor model, representing time, physical space, virtual space, application space, computational resources, storage resources, transfer resources, respectively.

The preferred embodiment also provides three resource fusion methods, including a physical space resource fusion method, a virtual space resource fusion method, and an application space resource fusion method. The traditional communication equipment comprises a plurality of service single disks and only has a single transmission function. The preferred embodiment provides a physical space resource fusion method, in which a computation storage single disk is added in a communication device, part of service single disk data is uploaded to the computation storage single disk, the service single disk data is directly processed at the device side, and part of the service single disk data is uploaded to a management and control platform. The virtual space resource fusion method provided in the preferred embodiment realizes cross-domain overall scheduling of virtual computing resources, virtual storage resources, and virtual transmission resources, and can optimally schedule various virtual resources according to a template issued by a global service orchestrator. The application space resource fusion method provided by the preferred embodiment can enable an industrial user to directly request to calculate, store and transmit resources based on business needs without possessing a bottom-layer resource capability technical background, a network virtualization architecture as a whole converts the user needs into resource needs, scheduling, allocation simulation and verification are carried out based on a virtualization resource pool, and the corresponding physical entity equipment is issued through a controller accurately and unmistakably to realize allocation of the calculation, storage and transmission resources.

Based on the above fusion method provided in this embodiment, a communication device manufacturer develops and produces a virtual model including a virtual computing resource model, a virtual storage resource model, and a virtual transmission resource model while developing and producing a physical entity device. The virtualization models are managed and controlled by a network virtualization system structure in a unified mode, after an operator communication network is deployed, data collected and reported in real time by a physical entity are fused with the virtualization models to form a series of virtualization examples, the virtualization examples in the virtual space correspond to entity equipment in the physical space, and the real-time presentation, history backtracking and forward-looking prediction of the operation mechanism and the health state of the physical entity equipment are supported. Based on the method provided by the embodiment, after the physical entity equipment has the conditions of performance degradation and the like, the communication network management and control platform establishes a performance optimization scheme based on an intelligent technology, establishes a virtualization instance in a virtual space to verify the performance optimization scheme, and can directly send an operation instruction to the physical entity equipment if the optimization effect can be achieved. In addition, many operations in a communication network take a relatively long time, but once a network failure or a performance degradation phenomenon occurs, it is required that the network can be quickly recovered. Based on the method provided by the embodiment, the virtualized instance is accurately matched with the existing network health state, the potential hidden danger or deterioration condition of the network can be predicted based on the virtualized instance, a corresponding processing scheme is formed and verified, when the communication network really has problems, the prepared processing scheme can be directly adopted, the scheme creation time is saved, and the communication network is ensured to be in real time troubleshooting or optimization.

Example 2:

based on the network virtualization architecture for computing storage and transport resource fusion provided in embodiment 1, this embodiment 2 specifically describes a three-layer three-plane network virtualization model and a seventh-order tensor model of the network virtualization architecture in more detail.

Continuing with fig. 1, for the three-layer three-sided network virtualization model of the present embodiment, the entire model is represented as a cube, three layers are directly in front of the cube, the lowest layer is a physical space, and physical devices including a computing device (e.g., a server), a storage device (e.g., a disk array), and a transmission device (e.g., an optical transmission device, a data communication device) are deployed. The middle layer is a virtual space, and a service orchestrator, a cross-domain controller, a single-domain controller, virtual computing resources, virtual storage resources and virtual transmission resources are deployed on the basis of the virtual technology. The top layer is the application space, containing a series of user applications, such as distributed machine learning. The right side of the cube comprises three faces, the three faces correspond to three technologies, and the virtualization technology, the intelligent operation and maintenance technology and the endogenous safety technology are respectively arranged from front to back. The three technologies run through three layers from top to bottom, and represent that the three technologies simultaneously act on a physical space, a virtual space and an application space. The virtualization technology carries out simulation modeling on various physical equipment operation mechanisms in the physical space, constructs corresponding virtual resources and deploys the virtual resources in the virtual space. The virtualization technology models various service scenes in an application space, abstracts main characteristics of various applications, and recommends an efficient resource allocation mode for a user according to the main characteristics. The intelligent operation and maintenance technology can realize the automatic processing of various resource planning stages, opening stages and maintenance stages in the time dimension, realize the intelligent processing of service arrangement and resource control based on the technologies such as artificial neural network, machine learning, knowledge map and the like, and support the intelligent correlation analysis of network root and derivative alarm, the intelligent positioning elimination of network faults and the intelligent prediction of performance degradation trend. The invention constructs a safe and reliable network virtualization system structure based on an endogenous safety technology to ensure the safety of various physical virtual resources and service scenes of a physical space, a virtual space and an application space.

The right side of fig. 1 is a seven-order tensor network virtualization architecture model (seven-order tensor model) provided by the embodiment of the present invention, and the model can quantitatively express each element in the three-layer three-sided network virtualization model and establish a relationship between each element. In this embodiment, three types of resources (i.e., computing resources, storage resources, and transmission resources) located in three spaces (i.e., a physical space, a virtual space, and a user space) are uniformly described and represented in a seven-order tensor model, and the three types of technologies (i.e., virtualization technology, intelligent operation and maintenance technology, and intrinsic safety technology) process elements in the tensor model, so as to implement efficient, flexible, intelligent, and safe resource provisioning.

Specifically, the seventh-order tensor model of the embodiment is described as T ∈ R ^{I1×I2×I3×I4×I5×I6×I7} Where R denotes the real number domain, I1, I2, I3, I4, I5, I6, I7 denote seven orders of the tensor model, representing time, physical space, virtual space, application space, computational resources, storage resources, transfer resources, respectively. The tensor comprises a plurality of dimensions at each order, the seven-order seven-dimension can determine the specific position of the object in the seven-order tensor model, the tensor element value of the position is a real number, and the real number corresponds to the specific eigenvalue of the object.

The following table 1 illustrates how tensor elements of the seventh-order tensor model take values in a machine learning application scenario, a controller system and a communication device calculation single disk.

Table 1 tensor element values example of the seventh order tensor model

Serial number	Name(s)	Stage I1	Stage I2	Stage I3	Stage I4	Stage I5	Stage I6	Stage I7	Numerical value
										1	Machine learning	2	0	0	1	8	2	4	1
2	Controller	3	0	1	0	4	1	1	2
										3	Calculating single disc	5	1	0	0	32	64	0	2

In table 1 above, the second to fourth rows represent three objects, the third to ninth columns represent seven orders of the tensor model, and the tenth column is an object specific numerical value. The seventh-order tensor model provided by this embodiment is a general model, and specific indexes of the seventh-order dimensionality values of the tensor can be defined according to actual application requirements. For example, when the current physical device of the communication network reports full-scale performance data every fifteen minutes in a minute-level data reporting mode, a period with a length of fifteen minutes can be represented by each integer of the first-order dimension values of the tensor model. All objects in the seventh-order tensor model are ordered according to the initial running time, the earliest running time corresponds to zero coordinates of the first-order dimension of the tensor model, and then every time the dimension is increased by one number, the representation spans one period, namely, the dimension value of 1 represents the fifteenth minute, and the dimension value of 2 represents the thirty-th minute. Following the above rules, the three objects in Table 1: the corresponding time of the machine learning, the controller and the single-disk calculation are respectively thirty minutes, forty-five minutes and fifteen minutes in the first hour. The object machine learning is a business scene located in an application space, so the value of the order I4 is 1, and the values of the orders I2 and I3 are all zero. In the implementation process, three-order dimensional indexes of a physical space, a virtual space and an application space can be defined according to specific requirements. In table 1, dimension 1 of the application space I4 is defined as a machine learning service scene, and dimension 2 is defined as a virtual reality application scene, so the value in row 2 and column 6 of table 1 is 1. Similarly, the controller is located in virtual space, and the tensor I3 order dimension 1 represents the controller object. Row 4, column 4, numerical value 1 of table 1 represents the tensor model I2 order corresponding compute single disk object. The tensor model I5 level represents the number of computing resource CPU cores, and the numerical value 8 in row 2 and column 7 of table 1 represents that the computing resource required for machine learning is 8-core CPU. Similarly, row 2, column 8, numerical value 2 of table 1 indicates that the storage resource required for machine learning is 2T hard disk space, and row 2, column 9, numerical value 4 of table 1 indicates that the transmission resource required for machine learning is 2 network slices. Row 2, column 10 of table 1 corresponds to tensor element values, in this example, the value 1 represents a machine learning application. The objects in line 3 of table 1 are controllers and the tensor element values 2 represent the deployment of 2 sets of controllers in virtual space. In the specific implementation process, the deployment of 2 sets of controllers adopts the main/standby mode to prevent the network service interruption caused by the failure of the controllers, and generally, the resources and the running states configured by the two sets of controllers are kept consistent, so that the dimensions corresponding to the stages of the tensor models in the two sets of controllers are the same. The other steps of the controller and the calculation unit are the same as the definition of the machine learning, and are not described herein again.

Table 1 of the present embodiment describes how to correspond the requirements of various objects to the dimensions of the tensor model in the implementation process by way of example. The machine learning object represented in the second row of table 1 may be formally represented in the tensor model as T (2,0,0,1,8,2,4) ═ 1. The controller object represented in the third row of table 1 may be represented formally in the tensor model as T (3,0,1,0,4,1,1) ═ 2. The computed single-disk object represented by the fourth row of table 1 may be formally represented in the tensor model as T (5,1,0,0,32,64,0) ═ 2. In the above-mentioned tensor element numerical table representation, T represents a tensor, seven values in parentheses represent the dimensional values of seven tensor orders, respectively, and the right side of the equation is the tensor element value. For example, the fifth element in the brackets of T (2,0,0,1,8,2,4) represents tension of the fifth order I5 dimension 8, with the actual meaning representing the computing resource being an 8-core Central Processing Unit (CPU). In table 1, the machine learning and controller is a party needing to compute, store and transmit resources, the machine learning object needs to consume resources to implement artificial intelligence learning training, and the controller consumes resources to execute various tasks such as communication network path computation. The computational single disk in table 1 is a resource provider capable of providing both computational resources and storage resources. The seventh-order tensor model provided by this embodiment can perform specific quantitative description on the calculation, storage, and transmission resources required by various objects in a physical space, a virtual space, and an application space, and can also perform quantitative description on the resource capacity values provided by various objects, thereby implementing unified quantitative description on the calculation, storage, and transmission resource requirements and supply capacities in one tensor model.

The tensor model provided by the embodiment is constructed, and the method has the greatest beneficial effect of laying a foundation for subsequently establishing the complex quantity relationship among the three types of resources of calculation, storage and transmission. Or, only by describing the calculation, storage and transmission resources quantitatively and realizing uniform representation in a space, the quantity relationship among various resources can be established, and the establishment of the quantity relationship is the core for realizing the optimal scheduling of the resources. Based on the tensor model provided by the invention, the computing, storing and transmitting resources are subjected to image extraction modeling by a virtualization technology and are sequentially expressed in a unified space in a quantization form; the artificial intelligence technology establishes the quantitative relation among calculation, storage and transmission resources based on tensor space elements, and realizes the intelligent operation and maintenance of the cloud computing center and the communication network in the whole life cycle of planning, construction, maintenance, optimization and operation; the endogenous security technology establishes a quantitative relation among calculation, storage and transmission resources based on tensor space, clearly evaluates the influence of various potential safety hazards possibly on user service application, formulates a related security plan, and starts the security plan once a security problem occurs to ensure that a user can still use the calculation, storage and transmission resources.

Example 3:

based on the network virtualization architecture for computing storage and transport resource fusion provided in embodiment 1, this embodiment 3 describes the virtualization technology used in the architecture in more detail.

As shown in fig. 2, the proposed network virtualization technology tree for this embodiment includes a root, two branches, and four leaves. The tree root represents a network virtualization technology, and the two branches represent a functional fidelity technology and a functional simulation technology. The four leaves are respectively a scheduling fidelity technology, a virtual fidelity technology, a function mapping simulation technology and a function fitting simulation technology.

In the specific implementation process, in order to virtualize computing, storing and transmitting resources of a physical space and construct a corresponding virtual entity, the embodiment provides a method of "fidelity is prior and simulation is second", and for a specific physical entity resource, a virtualization technology is selected in four leaves in a left-to-right sequence, that is, a sequence of (scheduling fidelity technology, virtual fidelity technology, function mapping simulation technology and function fitting simulation technology) is selected. For a concrete physical entity resource, firstly selecting a scheduling fidelity technology for virtualization, and if the scheduling fidelity technology is difficult to implement, selecting a virtual fidelity technology for abstract modeling. And if the fidelity technology cannot be applied to the physical entity resource, selecting a function mapping technology to construct a virtual entity, and if the fidelity technology cannot be implemented, selecting a function fitting technology to construct the virtual entity. In the field of virtualization technologies, current technical research is independently developed, and is focused on functional fidelity technologies or functional simulation technologies, and a unified virtualization technology tree does not clearly express various virtualization technologies, which also leads to a lack of an effective selection method for virtualizing a physical entity. The patent provides a 'fidelity optimization and simulation secondary' method to effectively solve the problem.

In addition, the virtualization technology represented by the four leaves in fig. 2 can make a transition along the sequence from right to left after a certain condition is met, namely, the transition along the sequence of (r) and (c). This embodiment proposes three transition modes, which are P transition, F transition, and FP transition. In the implementation process, if the operation mechanism of the physical entity is not known, a function fitting simulation technology is needed to be adopted for virtualization; when the operation mechanism of the physical entity can be completely mastered, virtualization is carried out by adopting a function mapping simulation technology in a P transition mode; when the virtualized entity constructed based on the function mapping simulation technology can really bear the functions of the physical entity, virtualization is carried out by adopting a virtual fidelity technology in an F transition mode; and if the conditions of P transition and F transition are met simultaneously, performing virtualization by adopting a scheduling fidelity technology in an FP transition mode. For example, for purchased optical devices, the internal operating mechanism of the optical devices is generally unknown. If a manufacturer later adopts a mode of independently developing the optical device, the operation mechanism of the optical device can be completely mastered, virtualization is carried out by adopting a function mapping simulation technology in a P transition mode, and the simulation Performance of a virtualized entity can be greatly improved from function fitting simulation to function mapping simulation, so that the P transition is called, and P is the first letter of Performance English word Performance. If the virtualized entity constructed based on the Function mapping simulation technology can really bear the functions of the physical entity through various technical breakthroughs, the virtualization is carried out based on the virtual fidelity technology through F transition, the functions of the virtual entity are actually expanded from the Function mapping simulation to the virtual fidelity technology, the virtual entity works like the physical entity, and F is the first letter of the Function English word Function. And if the performance improvement and the function expansion can be met, constructing a virtual entity based on a scheduling fidelity technology through FP transition. As shown in fig. 2, the network virtualization technology tree proposed in this embodiment includes four types of virtualization technologies, a virtualization technology selection method, and three transition methods. The foregoing illustrates a selection method and three transition methods for virtualization technologies, and the following describes four types of virtualization technologies.

For the virtualization technology, a communication device entity in a physical space constructs a corresponding virtual entity through the virtualization technology, and deploys the virtual entity in the virtual space. If the virtual entity's functionality corresponds to the physical device entity functionality, then functional fidelity techniques need to be employed. Functional simulation techniques may be employed if the virtual entity does not necessarily fully map the operating mechanisms of the physical device entities.

The scheduling fidelity technology of this embodiment adopts a function mapping manner to collect and sequence requests of users for calculating, storing and transmitting resources, formulate a resource scheduling priority level, and allocate resources from high to low according to the priority level to meet the user requirements. The bottom left corner of fig. 2 is a schematic diagram of the scheduling fidelity technique, the two bottom circles represent physical resources, the top circle represents user resource requirements, and the middle box represents a scheduler. The scheduler knows the availability of all physical resources, and schedules according to the resource requirements submitted by the users, and allocates proper resources for the users. In the lower left corner of fig. 2, the scheduler allocates the physical resources represented by the left circle to fulfill the user's demand.

Table 2 scheduling fidelity virtualization technique delay paths example

Table 2 illustrates scheduling fidelity virtualization techniques with latency path resource requirements as an example. If the user needs to complete one operation within 15 ms, wherein the calculation time is 5 ms, the data transmission time of the sending end and the receiving end needs to be less than 5 ms. Based on the three-layer three-surface network virtualization model and the seven-order tensor model constructed in the embodiment, physical entities on all transmission paths of the sending end and the receiving end are searched, and the time delay of the physical entities on each path for processing data is accumulated to obtain the total time delay of the path. For example, one path is connected over 400 km of fiber, with a transmission delay of 2 milliseconds. Two optical amplifiers are involved, with a propagation delay of 0.2 microseconds. Assuming that the digital communication device requires a delay of 300 microseconds, and other factors such as initial dispersion compensation fiber, congestion factors, etc., the propagation delay of this path is about 5 milliseconds. When a user puts forward a 10 millisecond resource requirement, the scheduler can distribute the path and the corresponding sending end-receiving end to the user, and the requirements of the user on calculation delay and transmission delay are met.

The second diagram at the bottom left of fig. 2 depicts virtual fidelity techniques, with the bottom three circles representing physical devices and the top two pentagons representing virtual devices. The physical equipment is constructed based on special hardware, realizes the processing of electric signals and optical signals and realizes the transmission of data messages. The virtual device is based on general hardware, processing performance is improved by using the cluster, and a data transmission function of the physical device is simulated by a software program. For example, a client side device cpe (client premium equipment) may construct a virtualized client side device vcpe (virtualized client premium equipment) through a virtualization technology, a Broadband Access server bras (Broadband Remote Access server) may construct a virtualized Broadband Access server vbars (virtualized Broadband Remote Access server) through a virtualization technology, and a virtualized client side device cpe (client premium equipment) may construct a virtualized Broadband Access server through a virtualization technology. The virtual fidelity technology can realize the mapping of the operation mechanism of the physical equipment and the virtual equipment, and ensure that the functions of the physical equipment and the virtual equipment are kept consistent. The virtual device constructed by the virtual fidelity technology can be in a box-type physical device form, and virtual device software is operated on a general central processing unit. The virtual device software can also be directly deployed on the cloud data center virtual machine. The specific shape of the virtual device can be determined according to the application scene of the user in the implementation process.

The third diagram below in FIG. 2 describes a functional mapping simulation technique. If the internal mechanisms of a physical space communication device can be understood in detail, the internal mechanisms can be translated into a mathematical model through a function mapping simulation technique. Table 3 illustrates the implementation steps of the functional mapping simulation technique using the erbium-doped fiber amplifier, which includes seven steps. Firstly, various parameter values, such as a gain spectrum and a functional spectrum, of the erbium-doped fiber amplifier are obtained. The method comprises the following steps of setting incident pump and signal light power, calculating the distribution of a fundamental mode of pump light and signal light, and calculating the incident pump and signal photon flux. And step five, calculating energy level distribution and calculating the transverse distribution of the gain coefficient. And the sixth step is loop iteration, the pumping and signal light power distribution of the tail end of the optical fiber of the amplifier is calculated from the incident pumping and signal light power through instructions such as updating and judging, and the source code of the simulation model is output in the step. And step seven, deploying the simulation model on the virtualization platform, collecting network data, executing data and model fusion operation, calculating through the simulation model, and outputting a virtualization simulation result.

TABLE 3 erbium doped fiber amplifier functional mapping simulation

The last diagram below in fig. 2 is a schematic diagram of a function-fitting simulation virtualization technology. If the internal mechanism of the physical space communication equipment cannot be known under certain conditions, input and output data of the communication equipment can be measured through instruments in a laboratory, a model is built based on a function fitting simulation technology, and a real function in a physical entity is approximated through a fitting function. The physical entity of the communication equipment in the figure comprises five functional modules, the first functional module is branched after execution, the upper flow needs to be processed by two functional modules, the lower flow needs to be processed by one functional module, and the two branches are summarized and processed by the last functional module after processing is finished, so that the result is output. Input data and output data are input and fitted to obtain a simulation function through a function fitting simulation technology, wherein the two circle functional modules of the upper branch are fitted to form a triangular functional module and a hexagonal functional module, and the lower branch is fitted to form a hexagonal functional module. This figure is a schematic diagram for visually illustrating the process flow of the functional fitting simulation technique. For example, if osnr (optical Signal Noise ratio) virtualization modeling is performed, values such as optical power and dispersion can be measured as input through an instrument in a laboratory, values of optical Signal to Noise ratio can be measured as output, and a deep Neural network (dnn) (deep Neural network) model is selected for fitting to obtain an osnr virtualization simulation model.

The network virtualization technology tree provided by the embodiment comprises four types of virtualization technologies, a virtualization technology selection method and three transition methods, and based on the technologies and the methods, the virtualization of a physical entity can be realized most effectively, the virtualized entity is updated according to the situation change, the performance and the function are optimized, and a foundation is laid for the fusion of subsequent computing, storing and transmitting resources.

Example 4:

based on the network virtualization architecture for computing storage and transport resource fusion provided in embodiment 1 and the network virtualization technology described in embodiment 3, as shown in fig. 3, embodiment 4 provides a method for modeling virtualization of a physical entity and optimizing a virtualization instance, where the method further includes optimizing the constructed virtualization instance when modeling virtualization of the physical entity.

The method of this embodiment includes four steps.

First, a physical entity deployed in a real network is defined, and device state data of the physical entity, configuration data of the physical entity, and the like are uploaded to a virtualization space (corresponding to "acquisition" in fig. 3). In the lower part of fig. 3, physical entity examples are shown, including optical transmission equipment and optical fibers, various single-disc models and parameters of the optical transmission equipment, various module and device model parameters are uploaded to the virtual space, and parameters such as the type of the optical fiber, the dispersion coefficient of the optical fiber, the attenuation coefficient of the optical fiber are uploaded to the virtual space.

And secondly, combing an entity operation mechanism according to the related equipment state data and configuration data uploaded by the physical entity, and selecting a corresponding virtualization technology to construct a virtualization model (constructing the virtualization model based on the network virtualization technology provided in embodiment 3, corresponding to modeling in fig. 3). In fig. 3, if the internal process flow of the physical entity can be accurately known, a function mapping simulation technique can be used to construct the virtualization model. If the internal processing flow of the physical entity cannot be known, a function fitting simulation technology can be adopted to construct the virtualization model.

And thirdly, calling a virtualization model according to the requirements of the user space service scene, transmitting real-time data, creating a virtualization instance, and constructing a virtual network (corresponding to the call in fig. 3). An example of a virtualized network is shown in fig. 3, which includes six network nodes, and this virtualized network may support an application scenario such as distributed machine learning.

And fourthly, after the application space executes various application scene specific tasks, feeding back a virtualization effect, and optimizing the virtualization model according to the virtualization effect (corresponding to the virtualization effect feedback, the combination strategy adjustment, the virtual algorithm optimization and the acquisition updating mode in fig. 3). The optimization of the virtualization model is divided into three types of situations, the first type of situation updates the data acquisition mode, the data acquisition reporting frequency is increased or new data indexes are added, and the virtualization effect of the model is improved. The second type of situation optimizes the virtual algorithm, such as enhancing or perfecting the functional mapping method, using a better fitting function, thereby achieving the goal of improving the simulation accuracy. And a third situation adjusts a combination strategy, which is used for complex service scene simulation, and improves the model virtualization effect by calling different virtualization models, for example, virtual computing resources are increased while virtual storage resources are reduced.

The upgrading optimization of the virtualization model and the virtualization instance can be realized through the four steps. The method for optimizing the physical entity virtualization modeling and the virtualization instance upgrading is a dynamic and automatic method. The traditional method is static and manual, the calling is passively waited after the virtualization model is built, then the next round of adjustment is passively waited, and the adjustment strategy is single. In the embodiment, after the virtualization model is built, the virtualization instances are generated in the virtual space, and all the instances provide calculation, storage and transmission capabilities for users after the virtualization spaces are fused. After the capabilities are provided for the user, upgrading optimization is actively carried out according to user feedback, and the optimization method comprises the steps of updating a data acquisition mode, optimizing a virtual algorithm and adjusting a combination strategy, which are provided by the embodiment. The optimized virtual algorithm may adopt the virtualization technology transition method in embodiment 3 of the present invention in the implementation and execution process. The dynamic automatic upgrading optimization method in the embodiment can be used for perfecting the virtualization model in real time, creating the high-precision virtualization instance in real time, ensuring that virtual computing, virtual storage and virtual transmission resources created in the virtual space can meet the diversified business application requirements of users in real time, and ensuring the cloud network convergence service quality.

Example 5:

based on the network virtualization architecture for computing storage and transport resource fusion provided in embodiment 1, this embodiment 5 proposes three fusion manners of computing storage and transport resources, which are respectively implemented in a physical space, a virtual space, and an application space.

Fig. 4 is a schematic diagram illustrating the fusion of the computing storage and transfer resources in the physical virtual application space in the present embodiment.

The bottom of fig. 4 is a physical space p (physical space) that is an entity device for deploying computing, storing, and transmitting resources, that is, the physical space includes computing resources, storing resources, and transmitting resources. The physical space computing storage transmission resource fusion method proposed in this embodiment refers to adding a computing and storage unit disk (may be simply referred to as a computing storage disk) to a transmission communication device, uploading a part of state data on a service disk to the computing storage disk, and uploading a part of state data to a management and control platform. And if the state data on the service disk is only related to the communication equipment, uploading the calculation storage disk and processing the calculation storage disk on the equipment side. And if the state data on the service disk is related to other network elements, uploading the state data to a management and control platform for centralized processing. And the computing storage disk is provided with an artificial intelligence learning training framework, and the training and reasoning of the lightweight artificial intelligence model are supported.

In fig. 4, a virtual space v (virtua space), a deployment orchestrator, and a controller are provided in the middle, and in addition, virtual computing resources, virtual storage resources, and virtual transmission resources are also constructed based on the network virtualization technology provided in embodiment 3, and these three types of virtual resources are also deployed in the virtual space. The virtual resources correspond to the physical resources, and in the four types of network virtualization technologies proposed in embodiment 3, the actual instructions finally executed by the virtual resources, which are constructed based on the scheduling fidelity technology, the function mapping simulation technology, and the function fitting simulation technology, are all on the corresponding physical resources. Virtual resources built based on virtual fidelity techniques can execute actual instructions. In fig. 4, the virtual space and the physical space have two types of interfaces, wherein the device status is uploaded from the physical space to the virtual space through a device status interface S-flow (state flow). The management command is issued from the virtual space to the physical space through a management command interface M-flow (management and control flow).

The upper part of fig. 4 is an application space a (application space) including various user applications, such as a distributed machine learning application and a cloud virtual reality application. In addition, the application space stores the user's capability requirements for resources, including computing capability requirements, storage capability requirements, and transport capability requirements. In embodiments 1 and 2, the calculation, storage and transmission resource capacity is uniformly described in the three-layer three-sided network virtualization model and the seventh-order tensor model, and after the user calculation, storage and transmission capacity demand is submitted, the application space searches the tensor model, matches the required resources, and allocates the resources to the user. The interfaces of the application space and the virtual space comprise a cloud network capacity interface C-flow (capability flow) and a user demand interface R-flow (requirement flow).

In fig. 4, three spaces (physical space, virtual space, application space) and four types of interfaces (device status interface, management and control instruction interface, cloud network capability interface, and user requirement interface) form a closed loop. Specifically, the integration of computing capacity requirements, storage capacity requirements and transmission capacity requirements is realized in the application space, after a user puts forward resource requirements, the application space is used for matching, corresponding virtual resources are found, and the virtual resources are sent to the virtual space through a user requirement interface R-flow. The virtual space realizes the fusion of virtual computing resources, virtual storage resources and virtual transmission resources, after receiving user requirements, the virtual space is processed through the orchestrator and the controller to form specific operation instructions, and the specific operation instructions are issued to specific physical equipment or executed on the virtual equipment through the management and control interface M-flow. The physical space realizes the integration of three types of entity resources of computing resources, storage resources and transmission resources, one part of state data in a service disk of the physical space is processed in the computing storage disk, and the other part of the state data is reported to the virtual space through an equipment state interface S-flow. Through the reported equipment state data, the virtual space can master the performance conditions of three types of resources, namely computing resources, storage resources and transmission resources in the physical space in real time, so that optimization decision can be realized during resource allocation. In this embodiment, the resource calculation, storage, and transmission capabilities of the virtual space are uniformly described in the three-layer three-sided network virtualization model and/or the seven-order tensor model, and are presented to the application space through the cloud network capability interface C-flow, and the application space is provided for various service scenes.

The method provided by the embodiment realizes the deep fusion of computing, storing and transmitting resources in three types of spaces of physics, virtualization and users, and selects the most suitable fusion mode according to the business requirements in the specific implementation process. For example, in order to provide computing storage services for users more quickly, a physical space fusion mode may be adopted in the edge access communication device, and the computing and storage unit disk is embedded in the communication device, so that not only can communication delay be reduced, but also communication security risks may be reduced, and computing storage transmission capability is provided at the same time. As another example, for the requirement of storing and transmitting resources in sudden large-scale mass computing, if the current resources are difficult to meet the requirement of the user, the virtual space resource fusion technology can be utilized to efficiently arrange services and schedule virtual resources according to the priority of the user. And if the user requirement changes in the scheduling process, performing dynamic virtual resource scheduling based on an artificial intelligence technology, and issuing physical equipment through a management and control interface M-flow. In conclusion, the three fusion modes provided by the invention can realize efficient and flexible fusion of computing, storing and transmitting resources, and can intelligently and dynamically schedule resources to meet the needs of users.

Example 6:

based on the network virtualization architecture for computing storage and transport resource fusion provided in embodiment 1 and the three fusion modes provided in embodiment 5, this embodiment 6 provides an automatic resource fusion scheduling method based on a virtualization technology.

As shown in fig. 5, the method of this embodiment includes five steps, and a specific implementation of the five steps is described below by taking a distributed machine learning application scenario as an example.

The first step is to make a request and submit a request description in an application space. In the step, a user puts forward a distributed machine learning request, submits request description in an application space, and submits machine learning training data, and the data is selected according to needs.

And secondly, analyzing the user request by using a space to obtain a calculation, storage and transmission resource quantization index, and describing the resource quantization index in a three-layer three-surface network virtualization model and/or a seven-order tensor model. If the requirements of the distributed machine learning application scenario on computing, storage and transmission resources are shown in table 4, the user training time is about 2 hours, the hard disk storage space required by training data is 2T, the memory requirement is 256G, the distributed machine learning adopts a data center cluster mode, a bandwidth of 100Gbps is required between a server and a switch, and the transmission delay is required within 20 milliseconds. The five indices in table 4 will correspond to the fifth, sixth, and seventh dimensions of the tensor model.

TABLE 4 example of quantitative indicators of resource requirements for distributed machine learning application scenarios

Application name	Training time	Hard disk storage space	Memory requirements	Transmission bandwidth requirement	Latency requirement
						Distributed machine learning	2 hours	2T	256G	100Gbps	20 milliseconds

After the resource quantization index is uniformly described, the virtual space executes a third step, and the orchestrator and the controller thereof convert the resource quantization index into a specific operation instruction.

And fourthly, creating a virtual model based on the virtualization technology, instantiating the virtual model, verifying the operation instruction, and issuing the equipment after the operation instruction is accurate. In this step, the virtual model interacts with the physical entity through the device state interface S-flow to synchronize the current state of the device in real time. Therefore, if the operation instruction cannot be executed in the physical entity, the virtual entity will not pass the verification, and will not issue the operation instruction to the device.

Fifthly, the calculating, storing and transmitting equipment receives and executes various operation instructions. Aiming at the resource requirements in the table 4, the controller obtains a path meeting the time delay requirement, data are sent to a corresponding server through the path, the server provides a storage space and a memory meeting the requirements, a virtual machine is established and machine learning training is executed, and if the intermediate parameters of the artificial neural network model need to be synchronized in the training process, the controller distributes a transmission path. And returning the execution result to the user after the training is finished.

In the implementation process, five types of information are interacted among the five steps, namely a Request of a user, capability requirements of resource quantization indexes, Operations executed on the virtual equipment, Instructions executed on the computing, storing and transmitting physical equipment, and Feedback information of execution results. The five steps and the five types of interactive information form a closed loop, and the automatic fusion scheduling of the computing, storing and transmitting resources is realized through the physical space, the virtual space and the user space.

The resource automatic fusion scheduling method based on the virtualization technology provided by the embodiment adopts an intelligent mechanism to realize virtualization, efficient fusion, automatic scheduling and feedback optimization of computing, storing and transmitting resources. The intelligent automatic mechanism can solve the problems of long period, low efficiency and the like caused by manual scheduling, and can ensure that the resource allocation result meets the service requirement of an application space through a closed-loop interaction mechanism and ensure that the resource allocation strategy is highly consistent in a virtual space and a physical space through a simulation verification mechanism.

Example 7:

based on the network virtualization architecture for computing storage and transport resource fusion provided in embodiment 1 and the three fusion methods provided in embodiment 5, this embodiment 7 describes in more detail the fusion method for computing storage and transport resources in a physical space.

Fig. 6 is a schematic diagram illustrating fusion of computing storage and transmission resources in a physical space according to the present embodiment. In specific implementation, the present embodiment includes three processes.

Firstly, one or more computing storage disks are arranged in a physical space, and in addition, the physical space also comprises a master control disk and a plurality of service disks. Specifically, this embodiment requires one or more computing storage disks, and the specific number may be determined according to a service scenario. For example, if the transfer device contains multiple service disks and the generated device status data is large, multiple computing storage disks need to be developed and deployed. If the number of service disks included in the transmission device is small and the amount of generated device state data is not large, a block of computing storage disk can be developed and deployed. After the communication network is established, the traffic volume can be increased or decreased, but the device state data volume reported by the service disk does not change, so that the required number of the computing storage disks can be determined when the communication device is delivered. If the device state data volume is increased due to the fact that the device state data collection frequency is increased, a computing storage disc with higher performance can be developed for replacement. For example, the performance index data of the current device is acquired and reported once in five minutes, because of the requirement of network optimization, the subsequent requirement is adjusted to be acquired and reported once in one minute, and the processing capacity and the storage capacity of the currently configured computing storage disk are difficult to meet the requirement, so that a computing storage disk with higher performance can be developed for replacement and upgrade.

After the communication device develops and deploys the computing storage disk, a second process is performed, as shown in the left side of fig. 6, installing and deploying basic software, including installing an operating system (e.g., a CentOS operating system, an Ubuntu operating system), installing data tool software (e.g., a database MySQL, a distributed file system), installing a digital twin modeling framework, an intelligent training inference framework, an algorithmic analysis processing framework, and the like. The basic software shown on the left side of fig. 6 is divided into hierarchical relationships, and the lower layer software is called by the upper layer software. Such as: the data tool software calls functions provided by the operating system to access the central processing unit, the memory, the disk, the peripheral equipment and the like. And the digital twin modeling framework, the intelligent training reasoning framework and the algorithm analysis processing framework call data tool software to acquire various data.

The third process is to implement data splitting and edge processing, as shown on the right of fig. 6. And the service disk 1, the service disks 2 and … and the service disk n acquire equipment state data, and if the acquired data are closely related to the network element, the data are uploaded to a calculation storage disk for edge processing. If the network element is related to other network elements, the network element is uploaded to a master control disk, and the master control disk is handed over to the management and control platform. After receiving the equipment state data, the computing storage disk is allocated according to the data processing requirement, and if the simulation modeling operation is carried out, the data is forwarded to the digital twin modeling framework; if the edge artificial intelligence reasoning is carried out, the data are forwarded to an intelligent training reasoning framework; and if the operation such as alarm correlation analysis or performance degradation analysis is carried out, the data are forwarded to an algorithm analysis processing framework.

Software developed based on the three processes is deployed in a service disk, a master control disk and a computing storage disk, achieves data distribution and task processing in a distributed cooperation mode, and achieves computing, storing and transmitting resource fusion with various single disks, operating systems, data tools and various frames in a physical space.

The physical space resource fusion mode provided by this embodiment can provide a new and more efficient product scheme, a computing storage disk is added in addition to a communication device service disk, and computing, storage and transmission infrastructure capabilities are provided for an upper layer virtual space and an application space in a physical space fusion mode. The novel product scheme can provide cloud computing and network capacity for users at network edge access positions, wherein the cloud computing capacity is borne by a computing storage disc, and the network capacity is borne by a service disc. In a novel product scheme, based on the method provided by the embodiment, a service disk analyzes and processes data, and if a user needs low-delay processing and the data scale is not particularly large, the data can be directly distributed to a computing storage disk by the service disk. If the data planning is large and exceeds the capacity of the computing storage disk, the data is distributed to the master control disk by the service disk, then uploaded to the control platform through the master control disk and delivered to the virtual space to be processed in the large-scale computing cluster. At present, many personal user applications show the characteristics of large quantity, less data, high time delay requirement, high safety requirement and the like, and information basic equipment is required to process data in real time at an access side, so that the time delay can be reduced on one hand, and potential safety hazards caused by network communication can be reduced on the other hand. The physical space fusion mode provided by the embodiment effectively solves the problem by deploying the computing storage disk on the communication equipment, adding a software tool and adopting a form of distributing data by the service disk, the computing storage disk, the master control disk and the management and control platform together, and can bring up a new product scheme.

The above description is intended to be illustrative of the preferred embodiment of the present invention and should not be taken as limiting the invention, but rather, the intention is to cover all modifications, equivalents, and alternatives falling within the spirit and scope of the invention. Those not described in detail in this specification are within the skill of the art.

Claims

1. A network virtualization architecture comprising a physical space, a virtual space, and an application space, wherein:

the application space obtains a virtualization model described by the virtual space and provides the virtualization model for various service scenes; the application space also accepts the requirements, integrates the requirements and issues the requirements to the virtual space.

2. The network virtualization architecture of claim 1 wherein the virtual space comprises an orchestrator, a controller, and virtual computing resources, virtual storage resources, virtual transport resources corresponding to the physical space; the application space comprises user applications, computing capacity requirements, storage capacity requirements and transmission capacity requirements; the interfaces of the physical space and the virtual space comprise an equipment state interface and a management and control instruction interface; the interface between the application space and the virtual space comprises a cloud network capacity interface and a user requirement interface.

3. The network virtualization architecture according to claim 2, wherein the physical space, the virtual space, the application space, the device state interface, the management and control instruction interface, the cloud network capability interface, and the user requirement interface form a closed loop, specifically:

the integration of computing capacity requirements, storage capacity requirements and transmission capacity requirements is realized in the application space, and after a user puts forward resource requirements, the application space is matched to find corresponding virtual resources, and the virtual resources are sent to the virtual space through a user requirement interface;

through the reported equipment resource state data, the virtual space can master the performance conditions of computing resources, storage resources and transmission resources in the physical space in real time, the computing, storage and transmission resource capacities of the virtual space are uniformly described in a virtualization model, and are presented to the application space through a cloud network capacity interface, and the application space provides various service scenes.

4. The network virtualization architecture of claim 3, wherein the process of implementing the automated converged scheduling of compute storage delivery resources by the physical space, the virtual space, and the user space comprises:

making a request, and submitting request description in an application space;

applying a space analysis request to obtain calculation, storage and transmission resource quantization indexes, and describing in a virtualization model;

5. The network virtualization architecture according to claim 3, wherein the process of merging the three types of entity resources, namely the computing resource, the storage resource and the transmission resource, implemented by the physical space specifically comprises:

installing and deploying basic software, wherein the basic software comprises one or more of an operating system, data tool software, a digital twin modeling framework, an intelligent training reasoning framework and an algorithm analysis processing framework; the data tool software calls a function provided by an operating system to access one or more of a central processing unit, a memory, a disk and peripheral equipment; a digital twin modeling framework, an intelligent training reasoning framework and an algorithm analysis processing framework call data tool software to obtain various data;

6. The network virtualization architecture of any one of claims 1-5, wherein the virtualization models comprise a three-layer three-sided network virtualization model and a seventh order tensor model, wherein:

7. The network virtualization architecture of claim 6, wherein the seventh order tensor model is described as T e R ^{I1×I2×I3×I4×I5×I6×I7} Wherein R represents a real number domain, I1,I2, I3, I4, I5, I6, I7 represent seven orders of the tensor model, representing time, physical space, virtual space, application space, computational resources, storage resources, transmission resources, respectively.

8. A virtualization method is characterized in that each resource is virtualized by adopting a virtualization technology, wherein the virtualization technology comprises a functional fidelity technology and a functional simulation technology, the functional fidelity technology comprises a scheduling fidelity technology and a virtual fidelity technology, and the functional simulation technology comprises a functional mapping simulation technology and a functional fitting simulation technology; and for a specific physical entity resource, selecting according to the priority of a scheduling fidelity technology, a virtual fidelity technology, a function mapping simulation technology and a function fitting simulation technology when selecting the virtualization technology.

9. The virtualization method according to claim 8, further comprising three transition modes, specifically, P transition, F transition, and FP transition, when selecting the virtualization technology:

if the operation mechanism of a certain physical entity cannot be solved, a function fitting simulation technology is needed to be adopted for virtualization; when the operation mechanism of the physical entity can be completely mastered, virtualization is performed by adopting a function mapping simulation technology in a P transition mode; when the virtualized entity constructed based on the function mapping simulation technology can really bear the functions of the physical entity, virtualization is carried out by adopting a virtual fidelity technology in an F transition mode; and if the conditions of P transition and F transition are met simultaneously, performing virtualization by adopting a scheduling fidelity technology in an FP transition mode.

10. The virtualization method according to any one of claims 8 to 9, wherein, when performing virtualization modeling on the physical entity, the virtualization method further comprises optimizing the constructed virtualization instance, specifically:

defining a physical entity deployed in a real network, and uploading equipment state data of the physical entity and configuration data of the physical entity to a virtual space;

calling a virtualization model according to the requirements of the application space service scene, transmitting real-time data, creating a virtualization instance and constructing a virtual network;