A computing power service chain management and control system architecture for the Zhirong identification network

Abstract

The present invention provides a computing power service chain management and control system architecture for the Zhirong identification network. The architecture includes: a computing power resource perception module, a computing power standardization modeling module, a resource scheduling strategy generation module, and a computing power service chain and service chain graph forwarding module; the computing power resource perception module receives computing power requests initiated by users and perceives the computing power information of the network component layer; the computing power standardization modeling module uses a multi-attribute decision algorithm to model the computing power information and outputs the computing power evaluation status of each computing node; the resource scheduling strategy generation module generates the deployment and scheduling strategy of the computing power service chain and the service chain graph according to the computing power service and computing power evaluation status; the computing power service chain and service chain graph forwarding module sends the deployment and scheduling strategy of the computing power service chain to the network component layer. The present invention provides a unified parsing structure for different types of computing power services in different application scenarios with computing power requirements, so as to serve as an effective support for the subsequent computing power service management and control mechanism to generate customized strategies.

Description

Technical Field

[0001] The present invention relates to the field of Internet application technologies, and in particular to a computing power service chain management and control system architecture for an intelligent financing identification network. Background Art

[0002] Since its conception, the internet has experienced rapid development at all levels. Today, it plays an increasingly important role in all aspects of society. Simultaneously, the growing number of internet users and the ever-expanding scale of networks place increasing demands on network resources and performance. However, existing mainstream networks, also known as traditional networks, are centered around the IP protocol suite. The allocation of resources such as IP address segments and network bandwidth is often tied to geographic location, making them very rigid. This clearly cannot meet the explosively growing communication needs of the internet and cannot provide adequate QoS (Quality of Service). To address this increasingly rigid interconnection, the National Engineering Research Center for Mobile Private Networks at Beijing Jiaotong University has proposed a "three-layer, two-domain" network architecture based on the existing integrated network model. This architecture separates control and forwarding, and decouples identity from location. This increases network flexibility and scalability, and introduces universal service identification, significantly improving network utilization efficiency. On this basis, in order to further adapt to the requirements of the next-generation Internet architecture, we have innovatively proposed a "three-layer, three-domain" intelligent identification network. Through the intelligent integration of multi-space and multi-dimensional resources across the entire network, we can achieve on-demand supply of personalized services and effective support for flexible networking, providing efficient, differentiated and customized communication network services for different industries and users, which are generally applicable to various network application scenarios.

[0003] To address network quality issues brought on by the ever-increasing scale of users and traffic, existing technologies have proposed an NFV (Network Functions Virtualization) solution from a service perspective. The NFV solution virtualizes network functions and decouples them from hardware devices via software. It uses a control plane to orchestrate and flexibly combine virtual network functions on demand, enabling flexible management and control. From a network perspective, SDN (Software Defined Network) decouples the data plane from the control plane, enabling flexible management and control of the data plane by the control plane. Based on this, the solution proposes a new network system—the Intelligent Identity Network—that organically integrates services and network resources, collaboratively scheduling to provide differentiated and customized network services to various users in different industries and application scenarios, thereby improving network service quality and flexible management and control capabilities.

[0004] The shortcomings of the aforementioned existing NFV solutions include: while they provide differentiated and customized network services tailored to various user needs, they do not consider services based on computing resources. As existing networks increasingly demand computing resources, the cost of computing resource infrastructure continues to decrease, and edge computing and cloud computing have made significant progress, heterogeneous computing power (CPU, GPU, DPU, FPGA) and edge node / center node computing power must all be considered.

[0005] To enhance user service customization and rationally allocate resources for network functions, existing technologies, within the context of network function virtualization, have proposed a Service Function Chain (SFC) solution. This solution constructs an SFC as an abstract sequence of virtual network functions with a given execution order. It also proposes the core components of the service function chain architecture, discussing which network instances each SFC virtual network function will be deployed on and the order in which traffic will flow through each network node.

[0006] The shortcomings of the SFC solution in the above-mentioned existing technology include: although the solution can use virtualization technology to orderly combine network functions and abstract them into service function chains, the services provided are all network functions such as firewalls and DNS (Domain Name System), and the heterogeneous computing power and other resources of network nodes are not fully considered. Therefore, it is worth considering to standardize the computing power at each location and the heterogeneous computing power into a service chain. Summary of the Invention

[0007] The embodiments of the present invention provide a computing power service chain management and control system architecture for the Zhirong identification network to achieve the orchestration of computing power services and the effective management and control of multi-dimensional resources.

[0008] In order to achieve the above-mentioned purpose, the present invention adopts the following technical solutions.

[0009] A computing power service chain management and control system architecture for the Zhirong Identity Network, including: a computing power resource perception module, a computing power standardization modeling module, a resource scheduling strategy generation module, and a computing power service chain and service chain graph forwarding module;

[0010] The computing power resource perception module is used to receive computing power requests initiated by users, parse the computing power requests, generate various identifiers required for orchestrating the computing power service chain based on various parameters selected by users, perceive computing power information at the network component layer, transmit the computing power information to the computing power standardization modeling module, and transmit various identifiers required for orchestrating the computing power service chain to the resource scheduling strategy generation module;

[0011] The computing power standardization modeling module is used to model the computing power information using a multi-attribute decision-making algorithm, obtain the computing power evaluation status of each computing node, and transmit the computing power evaluation status to the resource scheduling strategy generation module;

[0012] The resource scheduling strategy generation module is used to decompose the computing power request based on the computing power service chain and the service chain graph according to the various identifiers required for orchestrating the computing power service chain, parse the computing power request into multiple minimum units, generate the deployment and scheduling strategy of the computing power service chain and the service chain graph according to the parsed computing power service and the computing power evaluation status, and transmit the deployment and scheduling strategy of the computing power service chain and the service chain graph to the computing power service chain and service chain graph forwarding module;

[0013] The computing power service chain and service chain graph forwarding module is used to send the deployment and scheduling strategies of the computing power service chain and service chain graph to the network component layer, realize the deployment and scheduling of the computing power service chain, and provide routing and scheduling solutions to the network component layer.

[0014] Preferably, the computing power resource perception module is specifically used to decompose the computing power service according to the computing power application scenario, divide a computing power service into multiple subtasks with computing power requirements, and adapt the subtasks to the multi-dimensional resources of computing power, network and storage, arrange the subtasks in order to form a computing power service chain, and realize the orderly scheduling of multi-dimensional resources of physical nodes and virtual nodes. According to the correlation between each computing power subtask, the computing power service chain graph is represented as a directed graph of GS (Task, Eage), where Task is a set of subtasks constituting a computing power service, and Eage is a set of correlations between subtasks;

[0015] The various identifiers required for the generated orchestration computing power service chain include the service identifier SID, the service demand behavior description SBD, the group identifier FID, the group function behavior description FBD and the component identifier NBD. First, the SBD is generated by parsing the user computing power request, and the user's abstract request is concretized to generate a unique SID for the entire network based on the SBD.

[0016] Preferably, the computing power standardization modeling module is specifically used to perform standardized modeling of computing power based on computing power information, comprehensively considering the different computing powers of heterogeneous hardware devices, different algorithms and different computing power requirements of different scenarios and the real-time status of the network, and using a multi-attribute decision-making method to evaluate the computing power of each member of the network component, and output the computing power evaluation status of each computing node. The multi-attribute decision-making method includes obtaining decision information, including: attribute weights and attribute values, aggregating the decision information in a certain way, and sorting and selecting the best solutions.

[0017] Preferably, the computing power standardization modeling module is specifically used to set different weights for various computing power status information according to different computing power application scenarios and computing power service requests. The modeling mechanisms set for different computing power service types according to the multi-attribute decision-making method include: benefit type, cost type and interval type. For different computing power status information, the corresponding relationship between the standardized modeling types of computing power information is shown in Table 4-5:

[0018] Table 4-5 Correspondence between computing power status information and modeling type

[0019]

[0020]

[0021] Preferably, the resource scheduling strategy generation module is specifically configured to establish a resource scheduling strategy model for the network nodes that each service flow passes through based on the node computing power information parameters, network topology, and bandwidth information in the computing power evaluation state. The network parameters in the resource scheduling strategy model are shown in Table 4-4:

[0022] Table 4-4 Network parameters

[0023]

[0024]

[0025] For any task i , starting from the first starting task of the service to Task i The time required for the execution to complete is:

[0026]

[0027] in Task i The time required to perform the task calculation, is the transmission time of the predecessor task, the specific value is determined by the network topology, link bandwidth, transmission rate, and the output data length Op of the predecessor task j and routing policy decisions for transmission; For its waiting time,

[0028] The state of node k, sn(k), is 1 when there is a task to be executed on this node; it is 0 when there is no task to be executed on this node. i Deployed on Node i When going up, Node i The corresponding state sn(Node i ) changes from 0 to 1;

[0029] The execution node of each task i Through algorithm selection, in Task i All corresponding predecessor tasks Bf i After all tasks are completed, execute the Task i , looking for the execution node Node i When When the node k in S(t) makes the value of sn(k) 0, all nodes are selected as Node t ;

[0030] When Task i The last unfinished predecessor task is completed and the output data is transferred to the specified Node i At this moment, Task i The waiting time is exactly 0, when this node i After the tasks that are being executed and the tasks to be executed are completed, the waiting time value reaches If there is a fluctuation in node computing power resources, that is, the computing power p(k) of some nodes changes, the nodes to be deployed for tasks that have not yet been deployed should be replanned and re-arranged;

[0031] Assume that each computing node has a certain computing fee, based on the computing power of the node, measured by the fee rate sc(k), then for the task Task i , and its computational cost on the node is:

[0032] C i =sc(Nodei)×p(Node j ) / c j

[0033] Given a service GS, the total time required for its execution in the network is:

[0034] T=max{T i |Tasks i ∈Ed G}

[0035] The total computational cost is:

[0036]

[0037] Computing resource scheduling, or computing task deployment, is modeled as a multi-objective optimization problem to minimize service execution time and service computing costs, namely:

[0038]

[0039]

[0040]

[0041]

[0042]

[0043] Among them, α and β represent the weights of service execution time and service calculation cost respectively; constraint (1) is represented by the state of Zhirong identification node, which is a binary variable; constraint (2) ensures that all tasks in the predecessor task set must execute Tasks i The post-task can be executed only after , which is a binary variable; Constraint (3) represents the node computing resource constraint, which must ensure that the total computing resources required for the tasks deployed on a node should not exceed its maximum computing capacity. The binary variable x i；n =1 indicates that task i is at node Node j ∈n; Constraint (4) represents the transmission constraint of the task execution path, which must ensure that the bandwidth required by the link through which the service GS is executed should not exceed the maximum available bandwidth resource of the link. The binary variable y i；l =1 means executing task i data packet transmission path Link i ∈L;

[0044] By solving the multi-objective optimization problem, the deployment and scheduling strategy of the computing power service chain and service chain graph is generated.

[0045] Preferably, the computing power service chain and service chain graph forwarding module is specifically used to generate various identifiers required for computing power service deployment according to the deployment and scheduling strategies of the service chain and service chain graph. The identifiers include: SID, SBD, FID, FBD and NBD identifiers, and determine the network components used for computing power routing based on the generated NID and NBD. The network component information is parsed into a flow table for routing forwarding that can be recognized by the ONOS controller of the control plane of the routing layer. The data packet format SIH of the Zhirong Identification Network Data Plane is used on the data plane. The SIH is compatible with the computing power service through the extended header, and the deployment and scheduling strategies of the service chain and service chain graph are implemented through the controller and orchestrator using the extended FID protocol.

[0046] It can be seen from the technical solutions provided by the above-mentioned embodiments of the present invention that the present invention designs a service chain and service chain graph structure for computing power, providing a unified parsing structure for different types of computing power services in different application scenarios with computing power requirements, so as to serve as an effective support for the subsequent computing power service management and control mechanism to generate customized strategies.

[0047] Additional aspects and advantages of the present invention will be set forth in part in the following description, will become apparent from the following description, or may be learned by practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0049] Figure 1 A schematic diagram illustrating a computing power service provided by an embodiment of the present invention;

[0050] Figure 2 A schematic diagram of a computing power service chain management and control architecture for the Zhirong Identity Network provided by an embodiment of the present invention;

[0051] Figure 3 A schematic diagram of a computing power service chain and service chain graph scheduling provided by an embodiment of the present invention;

[0052] Figure 4 A schematic diagram of the format definition of a SID and SBD provided in an embodiment of the present invention;

[0053] Figure 5 A schematic diagram of the format of FID and FBD provided by an embodiment of the present invention;

[0054] Figure 6 A schematic diagram of a data packet format (SIH) for the Zhirong Identity Network data plane provided by an embodiment of the present invention;

[0055] Figure 7 A schematic diagram of the format of an FID-CSFC data packet provided in an embodiment of the present invention;

[0056] Figure 8 A schematic diagram of the structure of a network component layer provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0057] The embodiments of the present invention are described in detail below, examples of which are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and are not to be construed as limiting the present invention.

[0058] It will be understood by those skilled in the art that, unless expressly stated otherwise, the singular forms "a", "an", "said" and "the" used herein may also include the plural forms. It should be further understood that the term "comprising" used in the description of the present invention refers to the presence of the features, integers, steps, operations, elements and / or components, but does not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components and / or groups thereof. It should be understood that when we refer to an element as being "connected" or "coupled" to another element, it may be directly connected or coupled to the other element, or there may be intermediate elements. In addition, "connected" or "coupled" as used herein may include wireless connections or couplings. The term "and / or" used herein includes any unit and all combinations of one or more associated listed items.

[0059] It will be understood by those skilled in the art that, unless otherwise defined, all terms (including technical and scientific terms) used herein have the same meaning as commonly understood by those skilled in the art in the art to which the present invention pertains. It should also be understood that terms such as those defined in common dictionaries should be understood to have meanings consistent with their meanings in the context of the prior art and, unless defined as such herein, will not be interpreted in an idealized or overly formal sense.

[0060] To facilitate understanding of the embodiments of the present invention, several specific embodiments will be further explained below with reference to the accompanying drawings. However, each embodiment does not constitute a limitation on the embodiments of the present invention.

[0061] The embodiment of the present invention provides a computing power service chain management and control system architecture for the Zhirong identification network, which connects the intelligent service layer, resource adaptation layer and network component layer from top to bottom to perform differentiated and customized management of the computing power services requested by users, policy issuance and data forwarding, so as to realize flexible and efficient scheduling of computing power services and multi-dimensional resources such as computing power and network.

[0062] The embodiment of the present invention designs a computing power-oriented service chain, which aims to decompose the computing power service according to the computing power application scenario, divide a computing power service into multiple subtasks with computing power requirements, and adapt the subtasks to the multi-dimensional resources of computing power, network and storage, so that the subtasks with computing power requirements are arranged in order to form a computing power service chain, and realize the orderly scheduling of multi-dimensional resources of physical nodes and virtual nodes. For example, in the unmanned driving application scenario, the service can be divided into three ordered subtasks of environment recognition, path planning and vehicle control. The composition of the computing power service chain is shown in Table 4-1. Among them, the service type is a computing power service in the application scenario with computing power requirements; the source IP and destination IP are the IP addresses of the entry and exit nodes for scheduling the computing power service; the number of subtasks is the number of tasks required to be executed in an orderly manner to complete a computing power service; the computing power requirement is the size of the computing power resources required to run each subtask, including CPU (central processing unit), Memory, etc.

[0063] Table 4-1 Composition of the computing power service chain

[0064]

[0065]

[0066] A schematic diagram of a computing power service analysis provided by an embodiment of the present invention is as follows: Figure 1 As shown. Based on the expansion of the computing power service chain, the present invention designs a computing power-oriented service chain diagram, which aims to describe complex services of computing power application scenarios with multiple basic tasks (i.e., service chain entry nodes). However, the divided computing power subtasks often have correlations. If the correlations between these tasks are not considered, the computing power service cannot be completed normally. Therefore, when scheduling and deploying computing power services, it is necessary to consider the attributes of a single computing power task and the correlations between multiple subtasks in the service at the same time to be closer to actual applications. A single computing power subtask mainly considers the computing power resources it consumes; and the correlation between computing power subtasks mainly lies in the execution order of the dependencies between tasks. In order to construct a computing power service chain diagram, the computing power task parameters are defined as shown in Table 4-2. Among them, the computing power service chain diagram is represented as a directed graph of GS (Task, Eage), Task is a set of subtasks that constitute a computing power service, and Eage is a set of correlations between subtasks; since a complex computing power service may be composed of multiple starting tasks and terminating tasks, St is defined. G For the starting task set, Ed G is the set of terminated tasks; due to the order of the service chain, each subtask may have its predecessor and successor tasks, and Bf is defined i is the set of predecessor tasks of task i, Af iis the set of subsequent tasks for task i. In addition, computing task parameters also include the computing resources required for the task, input data length, output data length, etc. The composition of the computing service chain diagram is shown in Table 4-3.

[0067] Table 4-2 Computing Task Parameters

[0068]

[0069]

[0070] Table 4-3 Composition of the computing power service chain diagram

[0071]

[0072]

[0073] The embodiment of the present invention provides a computing power service chain management and control architecture for the Zhirong identification network. Figure 2 As shown, the computing power service chain management and control architecture includes: a computing power resource perception module, a computing power standardization modeling module, a resource scheduling strategy generation module, and a computing power service chain and service chain graph forwarding module.

[0074] From top to bottom, the system is divided into three layers. The intelligent service layer is the entry point for user requests. It analyzes user requests and generates computing power service chain orchestration strategies. The resource adaptation layer perceives computing power and other resource information and transmits the orchestration strategies generated by the intelligent service layer to the network component layer. The network component layer is the physical part of deployment and planning. The architecture runs through these three layers from top to bottom, realizing an intelligent, flexible, and efficient computing power service chain orchestration solution.

[0075] The process of executing the above computing power service chain management and control architecture can be divided into the following steps:

[0076] The user initiates a computing power request and transmits it to the intelligent engine of the intelligent service layer through the API (Application Programming Interface). The intelligent engine then parses the computing power request. The intelligent engine also parses the initiated computing power service and transmits it to the resource adaptation layer. Based on the parameters selected by the user, it generates the SID (Service ID), SBD (Service Behavior Description), FID (Family ID), FBD (Family Behavior Description), and NBD (Node Behavior Description) identifiers required to orchestrate a computing power service chain.

[0077] The computing power resource perception module in the resource adaptation layer actively perceives the computing power information of the network component layer and stores it in the database. This computing power information includes various status information related to computing, storage, and forwarding.

[0078] The computing power resource perception module delivers the collected computing power information to the computing power standardization modeling module in the intelligent engine. The computing power standardization modeling module uses a multi-attribute decision algorithm to model the computing power information. After modeling, it will output the computing power evaluation status of each computing node.

[0079] After computing power assessment, the computing power standardization modeling module inputs the assessment results into the resource scheduling strategy generation module, also a smart engine. Based on the computing power service chain and service chain diagram, the resource scheduling strategy generation module generates deployment and scheduling strategies for the computing power service chain and service chain diagram based on the computing power assessment results. Simultaneously, the resource scheduling strategy generation module decomposes computing power requests. Through task parsing, computing power requests are broken down into indivisible, minimal units. The parsed computing power services are combined with the acquired computing, storage, and network resource information and input into the strategy generation algorithm module to generate appropriate adaptation and scheduling strategies. The algorithm primarily guides the deployment and scheduling of the computing power service chain.

[0080] The resource scheduling policy generation module distributes the adaptation and scheduling policies for the generated service chains and service chain graphs to the computing service chain and service chain graph forwarding module in the resource adaptation layer. This computing service chain and service chain graph forwarding module can be the ONOS controller and the Kubernetes orchestrator, which are the entities that directly manage the network component layer. The ONOS routing controller uses the packet header format identified by the FID-CSFC to provide end-to-end data connectivity for the computing service chain. The Kubernetes orchestrator primarily allocates algorithmically processed tasks to pods based on policy results, facilitating their scheduling.

[0081] The controller and orchestrator issue flow tables to the network component layer, deploy containers, and provide routing and scheduling solutions. This policy is primarily issued through HTTP requests. Policies are written into JSON files and issued via southbound interfaces to implement the deployment and scheduling of the computing service chain. The network component layer consists of a cluster of general-purpose x86 servers. The controller and orchestrator are installed on the same server to provide centralized control within this network. Additionally, the remaining nodes are equipped with BMv2 software switches for cross-node container communication, connecting to pods via Linux bridges.

[0082] To deploy the computing power service chain, the computing power resource perception module and the computing power standardization modeling module specifically process the following: The computing power standardization modeling module quantifies and models the underlying heterogeneous computing power resources, forming a unified, quantified computing power resource that is understandable and quickly usable by the business layer. Business operations require the computing power requirements of the platform or device. At the same time, different types of businesses also require personalized requirements such as storage capacity and network services. The computing power resource perception module first generates a service behavior description (SBD) based on the parsed user computing power request, concretizing the user's abstract request. Based on the SBD, it then generates a network-wide unique service identifier (SID) to implement customized services. The SID is known to the user and allows the user to monitor the progress of the service, such as terminating a service prematurely. The module then enters the computing power perception modeling module.

[0083] By invoking various tools in the computing resource awareness module, you can achieve real-time awareness of ubiquitous computing power information. For general-purpose servers, you can call Zabbix to monitor computing power information such as CPU and memory. You can also use ICMP commands and shell scripts to monitor network information and floating-point computing power. Specific resource status information includes CPU utilization, available memory capacity, available storage capacity, GPU and NPU floating-point computing power, maximum network port transmission rate, link latency, and link bandwidth, as shown in Table 4-4.

[0084] Table 4-4 Resource status information

[0085] CPU utilization Available memory capacity Available storage capacity Floating-point computing capability Network port transmission rate Link delay Link bandwidth cpu_util memory_ava storage_ava FLOPS eth_max latency_max bw_max

[0086] The perceived data will be stored in the database. The computing power standardization modeling module will then use the data in the database to perform standardized computing power modeling, taking into account the different computing power of heterogeneous hardware devices, different algorithms, different computing power requirements in different scenarios, and the real-time status of the network. The computing power of each member of the network component is evaluated to prepare for computing power routing and resource adaptation. The specific node computing power evaluation mechanism adopts a multi-attribute decision-making method and is mainly composed of two parts:

[0087] 1) Obtain decision information, including attribute weights and values. 2) Aggregate this information in a specific way, and rank and prioritize the options. There are many ways to aggregate information, such as calculating a weighted average.

[0088] Depending on different computing power application scenarios and computing power service requests, computing power status information will have different weights. Therefore, the modeling mechanisms of different computing power service types vary according to the multi-attribute decision-making method. The multi-attribute decision-making algorithm can generally be divided into benefit-based, cost-based, fixed-based, deviation-based, interval-based, and deviation-interval-based models. The corresponding standardized modeling types of computing power information for different computing power status information are shown in Table 4-5.

[0089] Table 4-5 Correspondence between computing power status information and modeling type

[0090]

[0091]

[0092] Through computing power awareness modeling, the network topology identifies several nodes with optimal computing power in a given scenario. Through topology awareness and discovery, the Zhirong Identity Controller can perceive the network topology and, combined with network-wide computing power and other resource information, plan an optimal computing power service forwarding path through its intelligent engine. After planning is complete, the Zhirong Identity Controller delivers the policy results.

[0093] The processing process of the resource scheduling strategy generation module based on the computing power service chain and service chain graph provided by the embodiment of the present invention includes the following: The present invention aims to jointly schedule computing power resources and network resources based on the service chain and service chain graph in computing power application scenarios and build a resource scheduling strategy model. In the computing power service chain and service chain graph, tasks are deployed and executed in several Zhirong identification nodes in the network based on node computing power information and network status information. The input and output data of each task are forwarded between these nodes, forming several corresponding data streams.

[0094] Figure 3 This is a schematic diagram of a computing power service chain and service chain graph scheduling provided by an embodiment of the present invention. Computing power service chain and service chain graph scheduling involves scheduling computing power resources and network resources, namely, the deployment of tasks and the forwarding of task data flows. Therefore, the decision-making process for computing power service chain and service chain graph scheduling requires modeling the network nodes that each service flow passes through, understanding the node computing power information parameters, as well as network information such as network topology and bandwidth. Specific network parameter descriptions are shown in Table 4-4. The resource scheduling strategy model is constructed as follows:

[0095] Table 4-4 Network parameters

[0096]

[0097]

[0098] For any task i , starting from the first starting task of the service to Task i The time required for the execution to complete is

[0099]

[0100] in Task i The time required to perform the task calculation, is the transmission time of the predecessor task, the specific value is determined by the network topology, link bandwidth, transmission rate, and the output data length Op of the predecessor task j and routing policy decisions for transmission; For its waiting time, In particular, when Task i ∈St G hour,

[0101] The state of node k, sn(k), is 1 when there is a task to be executed on this node; it is 0 when there is no task to be executed on this node. i Deployed on Node i When going up, Node i The corresponding state sn(Node i ) changes from 0 to 1. Therefore, S(t) is a vector that changes with the deployment and execution of tasks.

[0102] The execution node of each task i Should be selected by algorithm, in Task i All corresponding predecessor tasks Bf i After all tasks are completed, you can execute the Task. i , looking for the execution node Node i When When the node k in S(t) makes the value of sn(k) 0, theoretically all nodes can be selected as Node i .

[0103] When Task i The last unfinished predecessor task is completed and the output data is transferred to the specified Node i At this moment, Task i The waiting time is exactly 0, when this node i After the tasks that are being executed and the tasks to be executed are completed, the waiting time value reaches If there is a fluctuation in node computing power resources, that is, the value of the computing power p(k) of some nodes changes, the nodes to be deployed for tasks that have not yet been deployed should be re-planned and re-arranged.

[0104] Therefore, if you want to make the task i The execution completion time value is the minimum, and the network link conditions and node computing power information need to be considered simultaneously.

[0105] Because the task uses the computing power of the node, each computing power node has a certain computing fee, which is measured by the fee rate sc(k) based on the computing power of the node. i , and its computational cost on the node is:

[0106] C i =sc(Nodei)×p(Node j ) / c j

[0107] Therefore, given a service GS, the total time required for its execution in the network is:

[0108] T=max{T i |Tasks i ∈Ed G}

[0109] The total computational cost is:

[0110]

[0111] In summary, computing resource scheduling, i.e., computing task deployment, is modeled as a multi-objective optimization problem to minimize service execution time and service computing costs, namely:

[0112]

[0113]

[0114]

[0115]

[0116]

[0117] Among them, α and β represent the weights of service execution time and service calculation cost respectively; constraint (1) is represented by the state of Zhirong identification node, which is a binary variable; constraint (2) ensures that all tasks in the predecessor task set must execute Tasks i The post-task can be executed only after , which is a binary variable; Constraint (3) represents the node computing resource constraint, which must ensure that the total computing resources required for the tasks deployed on a node should not exceed its maximum computing capacity, and the binary variable x i；n =1 indicates that task i is at node Node j ∈N; Constraint (4) represents the transmission constraint of the task execution path, which must ensure that the bandwidth required by the link through which the service GS is executed should not exceed the maximum available bandwidth resource of the link. The binary variable y i；l =1 means executing task i data packet transmission path Link i∈L.

[0118] By solving the multi-objective optimization problem, the deployment and scheduling strategy of the computing power service chain and service chain graph is generated.

[0119] The processing process of the computing power service chain and service chain graph forwarding module provided by the embodiments of the present invention includes: The present invention designs a forwarding module based on the computing power service chain and service chain graph for the Zhirong identity network, which includes two parts: identity generation and routing forwarding. The Zhirong identity controller generates the various identities required for computing power service deployment based on the generated strategy results, and distributes the generated strategy through the controller and orchestrator.

[0120] The identification generation module mainly generates five types of identifications, and generates computing power service deployment strategies based on these identifications. The first is the computing power service identification SID and the service description identification SBD. SID and SBD are generated by parsing user requests from the smart service layer. The format definitions of SID and SBD are as follows Figure 4 The meanings of the relevant fields are as follows: Service ID uniquely identifies a service request; Service Name includes computing power services and other services; Source Node represents the IP address from which the user initiates the request; Number of Subtasks represents the number of tasks that a computing power request can be broken down into; Computing Power Requirement identifies the computing power resources required to complete the computing power request; Number of Chain Nodes represents the number of Zhirong identification nodes that the computing power service passes through; Node Selection Strategy and Task Allocation Strategy represent the key elements required to deploy a computing power service.

[0121] Typical SIDs and SBDs are put into JSON in the following format:

[0122]

[0123]

[0124] Next is the ethnic group identifier FID and ethnic group description identifier FBD, which are generated by parsing SBD. The formats of FID and FBD are as follows: Figure 5 As shown, the meanings of the relevant fields are as follows: the group ID is a number representing the number of a computing power service; the group attributes include the computing power group; the group type represents the specific form of the computing power request; the number of group members represents the number of Zhirong identification nodes in the group; the service scheduling path, the number of subtask allocations and the node resources represent the specific resources required to constitute a computing power service.

[0125] The FID and FBD are put into JSON, and the typical form is as follows:

[0126]

[0127]

[0128] Finally, the component description identifier (NBD) is parsed from the FID and FBD, representing the specific deployment status in the network. This includes the Zhirong node IP address, the number of containers allocated to each node, and other related information. The NBD is encapsulated in JSON and sent to the controller and orchestrator. The typical NBD format is as follows:

[0129]

[0130]

[0131] The routing and forwarding module determines the network components used for computing power routing based on the generated NID and NBD, and parses this information into a flow table that can be recognized by the ONOS controller of the control plane of the routing layer for routing and forwarding. The data plane uses the packet format (SIH) of the Zhirong identification network data plane, as shown below. Figure 6 As shown in the figure, the computing power service is compatible with the extended header. To achieve this goal, it is first necessary to register the computing power service with the Zhirong Identification Controller and generate a family identifier (FID) for the computing power service to represent different computing power service families.

[0132] Through the extended FID protocol, the forwarding of the computing service chain is realized, and the modification of the data packet is completed using the programmable data plane of p4. The format of the FID-CSFC data packet provided by the embodiment of the present invention is as follows Figure 7 As shown, a SFC-like data packet header FID-CSFC and an FID-CSFC protocol are defined to implement flexible forwarding of data packets.

[0133] The network component layer of the computing power service adopts a special network solution. Figure 8 A schematic diagram of the structure of a network component layer provided by an embodiment of the present invention is shown in FIG. Figure 8The figure shows a generic server node on a data plane. Kubelet is the pod orchestration agent responsible for managing all pods running on the node. A pod is an abstract atomic unit representing one or more containers and their shared resources. The Weave plugin is a Kubelet plugin used to allocate networks for pods. As shown in the figure, the virtual switch and pods are connected via a Linux bridge. When the server node is initialized, all components except the pod are configured. When the control plane orchestrator issues an instruction to create a container, the Kubelet agent receives the instruction, allocates resources to create the pod, and deploys the corresponding container in the pod. It then calls the Weave plugin to assign an IP address to the created pod and set its gateway to the Linux bridge. The control plane's SDN controller then sends the corresponding flow table to the virtual switch. Data packets sent from containers within the pod will then reach the virtual switch through its gateway, the Linux bridge, and be forwarded according to the flow table rules.

[0134] The computing power-oriented service chain and service chain graph structure designed by the present invention is divided into fields such as service type, source IP, destination IP, number of subtasks, and computing power requirements of each subtask; other methods: the fields can be expanded, and the computing power resource types in the computing power requirements can be expanded.

[0135] In the scheduling scheme for computing power and network resources in the resource adaptation layer, the present invention integrates computing power resources with network resources and performs resource allocation management at the same time; other methods: based on the computing power request, each computing power sub-task is first deployed in the underlying physical network node and virtual network, and then, based on the relevance of the computing power sub-tasks, network resources are considered in turn to orchestrate computing power services and forward data packets.

[0136] In summary, the embodiments of the present invention provide a computing service orchestration and resource adaptation mechanism in general network scenarios, compared to Network Function Virtualization (NFV) technology. Based on the constraints of node computing and storage resources, they consider the deployment of virtual network function (firewall, DNS, etc.) instances and the routing planning of computing service flows based on the constraints of network resources such as latency and bandwidth. This provides a unified parsing structure for different types of computing services in different application scenarios with computing power requirements, effectively supporting the generation of customized strategies for subsequent computing service management and control mechanisms.

[0137] The present invention proposes a computing power service chain management and control mechanism for the intelligent and integrated identification network, which decomposes the computing power demand initiated by users into multiple sub-tasks (including virtual network functions) with multi-dimensional resource requirements such as computing power and network and related dependencies. At the same time, it integrates computing power resources such as computing and storage and network resources such as latency and bandwidth, and collaboratively manages service requirements and routing planning to achieve the orchestration of computing power services and the flexible management and control of multi-dimensional resources.

[0138] Those skilled in the art will appreciate that the accompanying drawings are merely schematic diagrams of an embodiment, and the modules or processes in the accompanying drawings are not necessarily required to implement the present invention.

[0139] From the above description of the embodiments, it can be seen that those skilled in the art can clearly understand that the present invention can be implemented by means of software plus the necessary general-purpose hardware platform. Based on this understanding, the technical solution of the present invention, or the portion that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a storage medium such as ROM / RAM, a magnetic disk, or an optical disk, and includes a number of instructions for enabling a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in various embodiments of the present invention or certain parts of the embodiments.

[0140] Each embodiment in this specification is described in a progressive manner. The same or similar parts between the embodiments can be referred to each other. Each embodiment focuses on the differences from other embodiments. In particular, for the device or system embodiments, since they are basically similar to the method embodiments, the description is relatively simple. For the relevant parts, refer to the partial description of the method embodiments. The device and system embodiments described above are merely schematic, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the scheme of this embodiment. A person of ordinary skill in the art can understand and implement it without making any creative efforts.

[0141] The above description is merely a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in the present invention should be included in the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be based on the scope of protection of the claims.

Claims

1. A computing power service chain management and control system for the Zhirong identification network, characterized by: include: Computing power resource perception module, computing power standardization modeling module, resource scheduling strategy generation module, and computing power service chain and service chain graph forwarding module; The computing power resource perception module is used to receive computing power requests initiated by users, parse the computing power requests, generate various identifiers required for orchestrating the computing power service chain based on various parameters selected by users, perceive computing power information at the network component layer, transmit the computing power information to the computing power standardization modeling module, and transmit various identifiers required for orchestrating the computing power service chain to the resource scheduling strategy generation module; The computing power standardization modeling module is used to model the computing power information using a multi-attribute decision-making algorithm, obtain the computing power evaluation status of each computing node, and transmit the computing power evaluation status to the resource scheduling strategy generation module; The resource scheduling strategy generation module is used to parse the computing power request based on the computing power service chain and the service chain graph according to the various identifiers required for orchestrating the computing power service chain, parse the computing power request into multiple minimum units, generate the deployment and scheduling strategy of the computing power service chain and the service chain graph according to the parsed computing power request and the computing power evaluation status, and transmit the deployment and scheduling strategy of the computing power service chain and the service chain graph to the computing power service chain and service chain graph forwarding module; The computing power service chain and service chain graph forwarding module is used to send the deployment and scheduling strategies of the computing power service chain and service chain graph to the network component layer, realize the deployment and scheduling of the computing power service chain, and provide routing and scheduling solutions to the network component layer.

2. The computing power service chain management and control system architecture for the Zhirong Identity Network according to claim 1 is characterized in that: The computing power resource perception module is specifically used to decompose computing power services according to computing power application scenarios, divide a computing power service into multiple subtasks with computing power requirements, and adapt the subtasks to computing power, network and storage multi-dimensional resources. The subtasks are arranged in order to form a computing power service chain, and the multi-dimensional resources of physical nodes and virtual nodes are orderly scheduled. According to the correlation between each computing power subtask, the computing power service chain diagram is represented as a directed graph of GS (Task, Eage), where Task is a set of subtasks that constitute a computing power service, and Eage is a set of correlations between subtasks. The various identifiers required for the generated orchestration computing power service chain include the service identifier SID, the service demand behavior description SBD, the group identifier FID, the group function behavior description FBD and the component identifier NBD. First, the SBD is generated by parsing the user computing power request, and the user's abstract request is concretized to generate a unique SID for the entire network based on the SBD.

3. The computing power service chain management and control system architecture for the Zhirong Identity Network according to claim 2 is characterized in that: The computing power standardization modeling module is specifically used to perform standardized modeling of computing power based on computing power information, comprehensively considering the different computing powers of heterogeneous hardware devices, different algorithms and different computing power requirements of different scenarios and the real-time status of the network, and using a multi-attribute decision-making method to evaluate the computing power of each member of the network component, and output the computing power evaluation status of each computing node. The multi-attribute decision-making method includes obtaining decision information, including: attribute weights and attribute values, aggregating the decision information in a certain way, and sorting and selecting the best solutions.

4. The computing power service chain management and control system architecture for the Zhirong Identity Network according to claim 3 is characterized in that: The computing power standardization modeling module is specifically used to set different weights for various computing power status information according to different computing power application scenarios and computing power service requests. The modeling mechanisms set for different computing power service types based on the multi-attribute decision-making method include: benefit type, cost type, and interval type. For different computing power status information, the corresponding relationship between the standardized modeling types of computing power information is as follows: Resource information: cpu_util, Modeling type: Cost, Model expression: Resource information memory_ava, modeling type: benefit type, model expression: Resource information storage_ava, modeling type: benefit type, model expression: Resource information FLOPS, modeling type: benefit type, model expression: Resource information eth_max, modeling type: benefit type, model expression: Resource information latency_max, modeling type: interval type, model expression: Resource information bw_max, modeling type: interval type, model expression:

5. The computing power service chain management and control system architecture for the Zhirong Identity Network according to claim 3 or 4 is characterized in that: The resource scheduling strategy generation module is specifically used to establish a resource scheduling strategy model for the network nodes that each service flow passes through based on the node computing power information parameters, network topology, and bandwidth information in the computing power evaluation state. The network parameters in the resource scheduling strategy model are: G(N,L) is the network topology that provides computing resources; Task is a collection of all related tasks that constitute a service; Node i Task i ∈The node where Task is deployed; Link i To execute the task i ∈Task path; b i Task i ∈Network resources required by Task; c i Task i ∈Task required computing resources; sn(k) is the state of node k, indicating whether there is a task to be executed; st(j) is the status of predecessor task j, indicating whether the predecessor task has been executed; sc(k) is the usage fee rate of node k; S(t) is the state set vector of computing nodes at time t; p(k) is the computing power of node k; b(l) is the available bandwidth resource of link l; For any task i , starting from the first starting task of the service to Task i The time required for the execution to complete is: in Task i The time required to perform the task calculation, is the transmission time of the predecessor task, the specific value is determined by the network topology, link bandwidth, transmission rate, and the output data length Op of the predecessor task j and routing policy decisions for transmission; For its waiting time, The state of node k, sn(k), is 1 when there is a task to be executed on this node; it is 0 when there is no task to be executed on this node. i Deployed on Node i When going up, Node i The corresponding state sn(Node i ) changes from 0 to 1; The execution node of each task i Through algorithm selection, in Task i All corresponding predecessor tasks Bf i After all tasks are completed, execute the Task i , looking for the execution node Node i When When the node k in S(t) makes the value of sn(k) 0, all nodes are selected as Node i ; When Task i The last unfinished predecessor task is completed and the output data is transferred to the specified Node i At this moment, Task i The waiting time is exactly 0, when this node i After the tasks that are being executed and the tasks to be executed are completed, the waiting time value reaches If there is a fluctuation in node computing power resources, that is, the computing power p(k) of some nodes changes, the nodes to be deployed for tasks that have not yet been deployed should be replanned and re-arranged; Assume that each computing node has a certain computing fee, based on the computing power of the node, measured by the fee rate sc(k), then for the task Task i , and its computational cost on the node is: C i =sc(Node j )×p(Node j ) / c j Given a service GS, the total time required for its execution in the network is: T=max{T i |Tasks i ∈Ed G } The total computational cost is: Computing resource scheduling, or computing task deployment, is modeled as a multi-objective optimization problem to minimize service execution time and service computing costs, namely: Among them, α and β represent the weights of service execution time and service calculation cost respectively; constraint (1) is represented by the state of Zhirong identification node, which is a binary variable; constraint (2) ensures that all tasks in the predecessor task set must execute Tasks i The post-task can be executed only after , which is a binary variable; Constraint (3) represents the node computing resource constraint, which must ensure that the total computing resources required for the tasks deployed on a node should not exceed its maximum computing capacity, and the binary variable x i；n =1 indicates that task i is at node Node j ∈N; Constraint (4) represents the transmission constraint of the task execution path, which must ensure that the bandwidth required by the link through which the service GS is executed should not exceed the maximum available bandwidth resource of the link. The binary variable y i；l =1 means executing task i data packet transmission path Link i ∈L; By solving the multi-objective optimization problem, the deployment and scheduling strategy of the computing power service chain and service chain graph is generated.

6. The computing power service chain management and control system architecture for the Zhirong Identity Network according to claim 5 is characterized in that: The computing power service chain and service chain graph forwarding module is specifically used to generate various identifiers required for computing power service deployment according to the deployment and scheduling strategies of the service chain and service chain graph. The identifiers include: SID, SBD, FID, FBD and NBD identifiers. The network components used for computing power routing are determined according to the generated NID and NBD, and the network component information is parsed into a flow table for routing forwarding that can be recognized by the ONOS controller of the control plane of the routing layer. The data plane data packet format SIH of the Zhirong Identification Network Data Plane is used in the data plane. The SIH is compatible with computing power services through an extended header, and the extended FID protocol is used to implement the deployment and scheduling strategies of the service chain and service chain graph through the controller and orchestrator.

Images