Wide-area hyper-convergence service customization system for distributed cloud

Abstract

This invention provides a wide-area hyperconverged service customization system for distributed clouds, comprising: constructing a wide-area hyperconverged service customization network architecture for distributed clouds based on three planes: forwarding, control, and service; creating a new service-centric protocol stack layered structure, rearranging the network capabilities of the service customization network; the new protocol stack layered structure includes an application layer, service expression layer, service transport layer, service interconnection layer, packet bearer layer, data link layer, and physical layer; and simultaneously drawing on the end-to-end trip service concept of third-party travel and transportation platforms, constructing a supply and demand matching mechanism between the application side and the network side. This invention introduces a service customization network into the distributed cloud, realizing service networking capabilities across wide area networks and throughout the distributed cloud, deterministic transmission, differentiated requirements of various applications, on-demand network service capabilities, and efficient collaboration between the network and applications.

Description

Technical Field

[0001] The present invention relates to the field of communication technologies, and in particular to a wide-area hyper-converged service customization system for distributed clouds. Background Art

[0002] With the integration of information technology (IT) and computer technology (CT), cloud-network integration has become a consensus and an important development trend. Its core concept is to integrate computing and storage functions in the network, and use software-defined methods to achieve collaborative management and scheduling of multi-dimensional resources such as network computing and storage. In particular, with the rise of edge computing and computing power network (CPN) technologies in recent years, the concept of "distributed cloud" has been proposed and strengthened, and has been listed by Gartner as one of the top ten strategic technology development trends in 2019 and 2020, and cloud-network integration has been further strengthened. The core concept of distributed cloud is to disperse the service capabilities of cloud computing from a high-level single point in the network to multiple low-level points in the network to form distributed cloud service capabilities across a wide range. However, in a distributed cloud, the connection between the cloud edge needs to cross the wide area network. The traditional Transmission Control Protocol / Internet Protocol (TCP / IP) technology system has problems such as "difficult to predict transmission quality" due to statistical multiplexing of forwarding resources, "network manageable but not controllable" due to distributed routing control, and "weak coordination between application networks" due to end-to-end application transmission. As a result, although the cloud, edge, and end can be connected through unified application management and control, it is difficult to achieve efficient data flow through the network, making it impossible for the cloud, edge, and end to truly achieve organic integration.

[0003] Service Customized Network (SCN) is a new model of future network architecture independently proposed by the team of Academician Liu Yunjie. The concept of SCN was proposed in 2014. It combines software-defined networking, network function virtualization, and cloud computing technologies to try to build an open, scalable, efficient and flexible network architecture with differentiated service capabilities that can schedule information resources. Network infrastructure providers can flexibly schedule network resources such as computing, storage, and bandwidth according to the personalized needs of users, build virtual networks with different service capabilities, and then build a new industrial ecological chain.

[0004] Introducing service-customized networks in distributed clouds can change the traditional "resource-centric" modeling method. Through the "service-centric" idea, multi-objective joint scheduling of global resources can be carried out to achieve hyper-convergence and integration of network, computing, and storage. However, introducing service-customized networks in distributed clouds faces many technical challenges, which are specifically manifested in:

[0005] (1) Service semantics are not reflected in the entire TCP / IP layer. The service quality capabilities provided by "patching" cannot fundamentally change the problems of "difficult to predict transmission quality" and "the network can be managed but not controlled".

[0006] (2) "Network application collaboration" is the key issue in realizing hyper-converged networks. Traditional networks are oriented to hosts, identify applications through deep packet inspection and import corresponding network paths. However, this is difficult for the application side to adopt because the network side needs to invade the inside of the application. Network path control can also be directly completed by the application initiating source routing, but the network side needs to open control permissions to the application side, which is difficult for the network side to accept. The network and application are relatively independent, and the network side and the application side belong to different entities. There is a huge gap between the network and applications. Summary of the invention

[0007] In view of this, an embodiment of the present invention provides a wide-area hyper-converged service customization system for distributed cloud to eliminate or improve one or more defects existing in the prior art, and to solve the problems existing in the prior art when integrating service customization networks in distributed clouds, such as difficult to predict transmission quality, manageable but uncontrollable networks, and inability to efficiently collaborate between networks and applications.

[0008] The present invention provides a distributed cloud-oriented wide-area hyper-converged service customization system, which is characterized by comprising:

[0009] The application layer encapsulates, encrypts and decrypts data packets of each application;

[0010] A service expression layer, which performs multi-dimensional service identification addressing on each application and its attributes of the application layer, wherein the attributes refer to the characteristics attached to each application during communication;

[0011] A service transmission layer, wherein the service transmission layer uses an open interface preset by a white box switch to control the transmission of a service customized network access side;

[0012] A service interconnection layer, wherein the service interconnection layer uses a software-defined network controller to realize the connection and communication between the service customization network and various wide area networks;

[0013] A packet bearer layer, wherein the packet bearer layer adopts a segment routing router to forward the data packet according to the network path specified at each node of the system; the packet bearer layer encapsulates the forwarding capability into a NaaS form, matches the forwarding capability with the needs of each application, and builds a consumption matching mechanism;

[0014] A data link layer, wherein the data link layer adds a frame header and a frame trailer to the head and the tail of the data packet, respectively, to perform frame sealing, and uses a time-sensitive network switch to perform end-to-end transmission of the data packet;

[0015] The physical layer uses an Ethernet optoelectronic interface for communication; the physical layer provides a multi-dimensional resource pool constructed by a core cloud server and an edge cloud server, and the multi-dimensional resource pool includes at least network resources, computing resources, storage resources, and time resources.

[0016] In some embodiments of the present invention, the service expression layer performs multi-dimensional service identification addressing on each application and its attributes of the application layer, further comprising:

[0017] On the basis of having a multi-dimensional service identifier, the multi-dimensional resources in the multi-dimensional resource pool are hierarchically scheduled, the target resource cluster is obtained according to the first-level scheduling indicators, the target node is obtained according to the second-level scheduling indicators, and the tasks of the corresponding application are distributed to the target node; the first-level scheduling indicators include geographic location, usage cost and cloud server information; the second-level scheduling indicators include resource proportion, power consumption and dependency association parameters between resources and corresponding nodes.

[0018] In some embodiments of the present invention, the communication requirements of each application are analyzed at the application layer to generate a requirement association matrix, and a representation paradigm of the communication requirements is constructed, and the communication requirements are expressed through the service expression layer.

[0019] In some embodiments of the present invention, the relevant factors affecting resource scheduling are abstracted into a multidimensional vector, the multidimensional vector is orthogonally decomposed to obtain decision subsets of different dimensions, and a multi-level decision tree is constructed, and the multi-level decision tree is used to generate scheduling decisions; the relevant factors include resource proportion, power consumption, dependency association parameters between resources and corresponding nodes, and resource energy consumption.

[0020] In some embodiments of the present invention, based on time slot constraints, a real-time queuing model is constructed with the jitter and delay of the network path as the target, and the real-time queuing model is used to generate a queuing plan for corresponding application data packets; the time slot constraints include data packet size, quantity, packet sending time and time resources.

[0021] In some embodiments of the present invention, a hybrid method of single-path scaling and multi-path sharing is adopted to elastically change the network path bandwidth to receive burst data.

[0022] In some embodiments of the present invention, network status information obtained through feedback is collected and analyzed in real time, a network optimization model is constructed, and the service customized network is optimized; the network status information includes link load, link bandwidth, link delay, link jitter, link packet loss rate, and link utilization.

[0023] In some embodiments of the present invention, a hybrid method of probe dialing and in-band telemetry is used to observe the actual delay and jitter of the network path, and metadata is injected into and extracted from production environment messages to perceive the network status information in real time.

[0024] In some embodiments of the present invention, a network actual consumption model is constructed based on the network status information obtained through real-time perception, the consumption records of the network paths are recorded, and a post-hoc settlement model is constructed in combination with a preset network service quality evaluation mechanism.

[0025] In some embodiments of the present invention, a canonical form game of the cloud provider to which the core cloud server and the edge cloud server belong is constructed based on game theory with the goal of maximizing its own interests. The overall interest calculation formula of the cloud provider is:

[0026]

[0027] Wherein, u(oi) represents the revenue obtainable by the cloud provider; pi represents the possibility of obtaining the corresponding revenue.

[0028] The beneficial effects of the present invention are at least:

[0029] The present invention provides a wide-area hyper-convergence service customization system for distributed cloud. Starting from the three planes of forwarding, control and business, a wide-area hyper-convergence service customization network architecture for distributed cloud is constructed, and the three technical levels of new network bearing, network operating system and cloud network hyper-convergence are combined to realize deterministic transmission, application differentiation requirements and on-demand network service capabilities. By constructing a new service-centric protocol stack layered structure, a wide-area hyper-convergence service customization network protocol stack for distributed cloud is designed, and the TCP end-to-end transmission control is broken from top to bottom by the application layer to realize application-level, cross-wide-area, distributed on-demand interconnection. On the basis of application differentiation requirements and deterministic service capabilities, the interface between the layers of the new protocol stack is further designed, and the network service capabilities are abstracted by introducing the concept of network capability ticket NaaSTicket in the service interconnection layer and the packet bearing layer, and a trust and anchoring relationship between the network and the application is established, so as to realize efficient collaboration between the network and the application, and matching between the network path and the application, and truly open up the relationship between the network and the application.

[0030] Additional advantages, purposes, and features of the present invention will be described in part in the following description, and will become apparent to those skilled in the art after studying the following, or may be learned from the practice of the present invention. The purposes and other advantages of the present invention may be achieved and obtained by the structures specifically indicated in the specification and the accompanying drawings.

[0031] Those skilled in the art will appreciate that the objectives and advantages that can be achieved with the present invention are not limited to the above specific description, and the above and other objectives that can be achieved by the present invention will be more clearly understood from the following detailed description. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] The drawings described herein are used to provide a further understanding of the present invention, constitute a part of the present application, and do not constitute a limitation of the present invention. In the drawings:

[0033] Figure 1 A schematic diagram of a novel protocol stack layered structure of a wide-area hyper-converged service customization system for distributed cloud in one embodiment of the present invention.

[0034] Figure 2 A schematic diagram of a customized network architecture for wide-area hyper-converged services for distributed cloud in one embodiment of the present invention.

[0035] Figure 3 The figure is a flow chart of the NaaS Ticket supply, demand and matching mechanism with service capability abstraction in one embodiment of the present invention. DETAILED DESCRIPTION

[0036] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the embodiments and the accompanying drawings. Here, the illustrative embodiments of the present invention and their descriptions are used to explain the present invention, but are not intended to limit the present invention.

[0037] It should also be noted that, in order to avoid obscuring the present invention due to unnecessary details, only structures and / or processing steps closely related to the solutions according to the present invention are shown in the accompanying drawings, while other details that are not closely related to the present invention are omitted.

[0038] It should be emphasized that the term “include / comprises” when used herein refers to the presence of features, elements, steps or components, but does not exclude the presence or addition of one or more other features, elements, steps or components.

[0039] It should also be noted that, unless otherwise specified, the term “connection” herein may refer not only to a direct connection but also to an indirect connection involving an intermediate.

[0040] Hereinafter, embodiments of the present invention will be described with reference to the accompanying drawings. In the accompanying drawings, the same reference numerals represent the same or similar components, or the same or similar steps.

[0041] It should be emphasized here that the step marks mentioned below are not intended to limit the order of the steps, but it should be understood that the steps can be executed in the order mentioned in the embodiment, or in a different order from the embodiment, or several steps can be executed simultaneously.

[0042] The present invention takes into account the above-mentioned huge challenges faced by the potential development trend of the existing technology in integrating service customization networks in distributed clouds. Specifically, there are problems such as difficult to predict transmission quality, manageable but uncontrollable networks, and inability to efficiently collaborate between networks and applications. Therefore, the present invention proposes a wide-area hyper-convergent service customization system for distributed clouds, breaking through the layered design and system architecture of future network protocol stacks, while drawing on the end-to-end itinerary service concept of third-party travel and transportation platforms, and exploring and studying the supply and demand matching mechanism between the application side and the network side. While building cloud network technology for distributed clouds, various information and communication technology (ICT) services and resources (algorithms, programs, data, content, etc.) are intelligently and dynamically distributed and connected on-demand at multiple levels of the operator's "cloud edge" and even across multiple operators, realizing hyper-convergent integration of network, computing, and storage in a wide area.

[0043] TCP / IP simplifies the Open System Interconnection (OSI) model into a four-layer model, with the IP protocol cluster as the network waist. The IP "best effort" technology gene leads to the coexistence with TCP "end-to-end transmission control". The service semantics are not reflected in the entire TCP / IP layer. At the same time, the service quality capabilities provided by "patching" cannot effectively achieve the target capabilities of deterministic transmission, differentiated requirements and on-demand use required by the service customization network. Figure 1 As shown, the present invention is based on the OSI seven-layer model, replaces the network layer with the service layer, constructs a new protocol stack layered structure, and redesigns the functions of each layer. The new layered structure includes:

[0044] The application layer has the capabilities of the application layer, presentation layer, and part of the session layer in the OSI model. It is mainly responsible for data encapsulation, encryption, and decryption of data packets of each application.

[0045] The service expression layer has the capabilities of some of the session layer and some of the transport layer in the OSI model. It performs multi-dimensional service identification addressing on each application and its attributes in the application layer, where attributes refer to the characteristics attached to the application during communication.

[0046] The service transport layer has some of the capabilities of the transport layer in the OSI model and uses the open interface preset by the white box switch to control the transmission on the access side of the service customized network.

[0047] The service interconnection layer has the capabilities of some network layers in the OSI model and uses a software-defined network controller (SDN controller) to achieve interconnection and interoperability of service-customized networks.

[0048] The packet bearer layer has some capabilities of the network layer and data link layer in the OSI model, and uses segment routing routers to forward packets according to the network path specified by each application data packet. At the same time, the forwarding capability is encapsulated into a NaaS form, and the forwarding capability is matched with the needs of each application to build a consumption matching mechanism.

[0049] The data link layer has some capabilities of the data link layer in the OSI model. It is responsible for adding frame headers and frame trailers to the head and tail of data packets, sealing the frames, and using time-sensitive network switches to transmit data packets end-to-end.

[0050] The physical layer has the capabilities of the physical layer in the OSI model, uses Ethernet optical and electrical interfaces for communication, and provides a multi-dimensional resource pool constructed by core cloud servers and edge cloud servers, where the multi-dimensional resource pool includes at least network resources, computing resources, storage resources, and time resources.

[0051] Based on the new hierarchical structure provided by the present invention, the system layout and redesign of the key functions such as packet loss retransmission, order preservation and deduplication, flow control, congestion control, and traffic engineering in the data link layer, network layer, and transport layer are realized, the new data link layer is strengthened and the new service transport layer function is weakened, and the supply of "on-demand customization" capabilities of the network is systematically supported. At the same time, in order to realize the "on-demand customization" driven by each application, the communication requirements between each application are analyzed and processed, and the capabilities of each layer of the network are connected from top to bottom based on functional layering.

[0052] In some embodiments, a systems theory approach is used at the application layer to analyze the communication requirements of each application, generate a requirement association matrix, and construct a representation paradigm for communication requirements, and express the communication requirements through a service expression layer.

[0053] At the same time, data flow modeling is carried out, and byte encoding mapping is performed on the service transport layer, service interconnection layer, and data link layer layer by layer; a new network coding theory (information theory) is used to achieve spatial and temporal control of network coding to drive the differentiated allocation of network space-time resources in the new physical layer.

[0054] like Figure 2 As shown, on the basis of the above-mentioned new hierarchical structure, in order to realize the target capabilities of service customization network with deterministic transmission, differentiated requirements and on-demand use, the architecture of wide-area hyper-converged service customization system for distributed cloud is proposed, which can be divided into three major technical planes: new network bearer, network operating system and cloud-network hyper-convergence. The network forwarding plane, control plane and business plane are systematically reconstructed. Specifically, the new network bearer technology plane corresponds to the forwarding plane, the network operating system technology plane corresponds to the control plane, and the cloud-network hyper-convergence technology plane corresponds to the business plane.

[0055] In the new network bearer technology plane, multiple domains are included, and a new network data platform that is programmable, deterministic, customized, and high-performance is built using segment routing routers, time-sensitive network switches (TSN), and flexible Ethernet (Flex Ethernet, FlexE) optical interfaces. Among them, segment routing routers are used to perform programmable forwarding according to the network path specified by each application data packet, while reserving network resources; time-sensitive network switches are used to provide microsecond-level deterministic services for end-to-end transmission of data packets through frequency clock synchronization, cycle mapping, and traffic shaping technologies; Ethernet optical interfaces are used for communication, breaking the one-to-one mapping relationship between the original network layer and the original physical layer, and realizing flexible and refined management of interface resources. At the same time, Ethernet optical interfaces guarantee the physical underlying performance of high-pass bandwidth, elastic pooling, and flexible encapsulation, and support optical signal PAMx modulation system.

[0056] The network operating system technology plane includes an SDN (Software Defined Network) controller. The SDN controller is an operating system (OS) of the network. The SDN controller does not control the network hardware but runs as software, which is conducive to network automation management. In the present invention, the SDN controller, as a centralized network brain, can collect the status of the service customization network in real time, including network topology, interface status, queue buffer, etc., and form the global dynamic load information of the service customization network.

[0057] At the same time, white-box devices are used to transmit data based on algorithm-based open interfaces. Among them, white-box device management is mainly based on white-box switches with software and hardware decoupling. The underlying hardware and upper-layer software can be customized according to the needs of each application. Compared with the traditional switch software and hardware bundled purchase and monopoly use, it can significantly reduce the purchase cost of the switch. In addition, in terms of software functions, secondary development can be carried out based on open source software to reduce development cycle and cost. In the present invention, white-box device management is adopted to customize the forwarding logic of the data plane in a software-defined manner, support the rapid development of various routing protocols, and improve the agility, customizability and unified and flexible management of the network.

[0058] At the same time, the network operating system performs unified abstract description modeling for each application, and realizes lightweight reporting of the northbound interface of the data plane through the algorithm open interface design, supporting efficient applications under the requirements of large-scale heterogeneous algorithms. As a technical plane that effectively connects the network resources of the forwarding plane with the network requirements of the business plane, the network operating system draws on the resource scheduling algorithm for user-state applications (processes, threads) and user-state applications themselves (coroutines) in the computer operating system, schedules the transmission resource configuration required by application services (five-tuple, connection, service / sub-service), solves the trust boundary problem between the two states, and realizes a general intelligent and scalable network management and control capability middle platform.

[0059] In the cloud-network hyper-convergence technology plane, there are multi-layer core cloud servers and multi-layer edge cloud servers, which form a distributed cloud. Among them, core cloud servers are mostly located in the core position, and edge cloud servers are close to user-side mobile terminal devices, which are divided into different layers according to the distance from the user.

[0060] The cloud-network hyper-convergence technology plane realizes the vision of wide-area hyper-convergence integration with functions such as multi-dimensional service identification addressing, integrated resource multi-level scheduling, and large-scale multi-cloud exchange. Specifically:

[0061] At the service expression layer, multi-dimensional service identification addressing is performed on each application and its attributes at the application layer. Multi-dimensional service identification addressing is divided into basic identification and non-basic identification; attributes refer to the characteristics attached to each application during communication. Basic identification is used to distinguish different applications natively, and service attributes are calibrated to provide addressing capabilities for two-dimensional service space. For example, basic identification includes identification of information such as programs, content, and applications.

[0062] In some embodiments, when the basic identifier cannot cover some non-basic content, the metadata is used for association, and non-basic identifiers are formulated to identify each application. Exemplarily, the non-basic identifiers include identifying information such as IP addresses and GPS.

[0063] In some embodiments, based on the multi-dimensional service identification, a hybrid scheduling framework is designed to coordinate the scheduling of multiple heterogeneous resource clusters across regions, and all nodes and / or clusters of the system are uniformly managed and hierarchically scheduled. Among them, hierarchical scheduling is specifically embodied as follows: the target resource cluster is obtained according to the first-level scheduling indicators, the target node is obtained according to the second-level scheduling indicators, and the corresponding application is distributed to the selected target nodes. Capacity is divided between the first-level and second-level scheduling. The first-level scheduling indicators include geographic location, usage cost and cloud server information; the second-level scheduling indicators include resource proportion, power consumption and dependency association parameters between resources and corresponding nodes.

[0064] At the same time, the cloud network hyper-convergence technology plane also builds a large-scale multi-cloud exchange and interconnection system. In view of the inconsistent capabilities, functions, interfaces and heterogeneous cloud scenarios of each cloud platform such as public cloud, private cloud and industry cloud, a distributed software service platform is designed to improve the overall stability and scalability of the system and enhance the management capabilities of high-performance multi-cloud gateways. Therefore, the cloud network hyper-convergence technology plane connects the computing and storage resources in the distributed cloud, supports the intelligent and dynamic distribution of various application services in the distributed cloud, and thus realizes the heterogeneous ubiquitous integrated cloud network business capability prototype.

[0065] Based on the three major technical planes of new network bearer, network operating system and cloud-network hyper-convergence described above, we can further explore the dual-plane form of the future network, with the basic bearer network (including forwarding plane and control plane) as the first plane and the distributed cloud business plane as the overlay service second plane.

[0066] In a macro sense, a distributed cloud consists of core cloud servers and edge cloud servers. Core cloud servers are mostly located in the core position of the network, with high network connectivity and rich computing and storage resources, but the average distance from the user-side mobile terminal is relatively long, resulting in increased access latency. Edge cloud servers are mostly located at the edge of the network. Relatively speaking, the network connectivity is low and the computing and storage resources are relatively few, but the average distance from the edge cloud server to the user-side mobile terminal is close, resulting in lower access latency. Core cloud servers at different levels can offload the core cloud to form a distributed remote core cloud service capability, and can achieve organic collaboration with the core cloud server. Specifically, the core cloud server can provide full service capabilities, and multi-level edge cloud servers can collaborate to achieve dynamic caching of the service capabilities of the core cloud server.

[0067] From the perspective of the network, the above structure is provided by the edge cloud server to provide access control capabilities, and the core cloud server can provide traffic optimization capabilities between edge cloud servers. This structure can be built independently on the operator's bearer network, and by utilizing the rich network connectivity of the core cloud server, it can cover the entire Internet across different operators, thereby forming the second plane of the Internet.

[0068] As the first plane, the basic bearer network can provide capabilities for the distributed cloud interconnection of the second plane based on the service customization concept and the network operating system, including elastic and scalable bandwidth and predictable service quality. The "overlay service plane" and the "basic bearer plane" can evolve into a dual-plane form of the future network in a loosely coupled but organically coordinated manner.

[0069] In the existing technology, the network and applications are relatively independent, and there are still many problems in achieving effective collaboration. The trust and anchoring relationship between the network and applications is the technical key to the development of wide-area hyper-converged service customized networks for distributed clouds, including two important mechanisms: one is to decouple service-related network functions from the underlying network infrastructure through virtualization; the other is to encapsulate network resources into service entities and expose and utilize them through abstract interfaces. This technology is called Network-as-a-Service (NaaS).

[0070] Based on the above principles, the present invention encapsulates the differentiated requirements of the service customized network, deterministic transmission and other forwarding capabilities into NaaS form in the packet bearer layer, introduces the concept of network capability ticket (NaaS Ticket) to abstract the network service capabilities, and constructs a contractual model between passengers (services) and transportation tools (bearers) through the concept of tickets. Thus, a consumption matching mechanism for network service capabilities is established to achieve organic coordination between network capabilities on the network side and service requirements on the application side, thereby truly opening up the relationship between the network and applications.

[0071] like Figure 3 As shown, the present invention constructs a NaaS Ticket supply and demand and matching mechanism with abstract service capabilities, which mainly includes: a network ticket supply mechanism with elasticity determined by differences, an application ticket consumption mechanism with on-demand declaration of a posteriori, and a cross-domain timely and dynamic platform ticket matching mechanism.

[0072] Aiming at the customized needs of communication between applications, the focus is on abstracting service capabilities such as network path delay, jitter, and bandwidth, and building a network ticket supply mechanism with differential determination and elasticity. The specific process includes the following steps S101 to S104:

[0073] Step S101: Use dimensional analysis to establish the intrinsic relationship between network spatiotemporal resources and network path capabilities.

[0074] Step S102: Real-time online planning is used to implement differentiated scheduling of network paths under resource-constrained conditions.

[0075] Step S103: using packet time slot queuing to ensure that network path delay and jitter are bounded.

[0076] Step S104: elastically changing the network path bandwidth based on a combination of single-path scaling and multi-path sharing.

[0077] In step S101, the service capabilities of the network path, such as latency, jitter, and bandwidth, are essentially the demand for spatiotemporal resources such as forwarding queues and links. Therefore, in order to establish the intrinsic relationship between network spatiotemporal resources and network path capabilities, the dimensional analysis method is used to quantify the network capabilities and construct a pricing model for resource consumption.

[0078] In step S102, the resource supply of the bearer network is limited. In order to ensure differentiated demands and efficient resource utilization, real-time online planning (operations research) is used to implement differentiated scheduling of network paths under resource constraints.

[0079] In some embodiments, the relevant factors affecting resource scheduling are abstracted into multi-dimensional vectors, and the multi-dimensional vectors are orthogonally decomposed to obtain decision subsets of different dimensions, and a multi-level decision tree is constructed. The multi-level decision tree is used to coordinate the decision sub-modules of different dimensions to complete the entire decision process, and finally generate a multi-level hybrid distributed scheduling decision to meet the communication needs of various applications. Among them, the relevant factors include resource proportion, power consumption, dependency association parameters between resources and corresponding nodes, and resource consumption.

[0080] In step S103, based on the resource abstraction of step S102 and the resource scheduling of step S103, time dimension resources, i.e. time slots, are reserved in advance for network traffic. In order to achieve deterministic quality of service, packet time slot queuing (operations research) is used to ensure bounded network path delay and jitter.

[0081] In some embodiments, in complex topology and massive traffic scenarios, based on time slot constraints, with the jitter and delay of the network path as the target, a real-time queuing model is constructed, and the queuing scheme of each application data packet is generated using the real-time queuing model to solve the problems of queuing network and complex queuing. Among them, the time slot constraints include data packet size, quantity, packet sending time and time resources. At the same time, according to the requirements of the above-mentioned deterministic service classification and grading, deterministic grading can be achieved through the dual-factor conversion of jitter and delay.

[0082] In step S104, considering the network micro-burst problem, that is, the network receives a lot of burst data in milliseconds, making the burst data rate tens or hundreds of times higher than the average rate, or even exceeding the port bandwidth, a hybrid method of single-path scaling and multi-path sharing is used to elastically change the network path bandwidth, and support appropriate micro-bursts through commitment and tolerance dual-rate adjustment.

[0083] Referring to the standard specifications and end-to-end modeling of Quality of Service (QoS) of the Telecommunication Standardization Department ITU-T, a consumption mechanism of application tickets for on-demand declaration of a posteriori application in a service customization network is constructed. The specific process includes the following steps S201 to S204:

[0084] Step S201: using distributed system theory to mine the intrinsic relationship between system throughput concurrency and end-to-end QoS of inter-application communication.

[0085] Step S202: adopting feedback control theory to realize on-demand expression of end-to-end QoS of applications and automatic decomposition of network performance NP.

[0086] Step S203: Establish a network service capability evaluation mechanism based on a combination of probe dialing and in-band telemetry.

[0087] Step S204: network path consumption accounting is implemented based on actual traffic observation and service capability evaluation.

[0088] In step S201, in order to further analyze the impact of statistical service quality QoS on system throughput, the distributed system theory (system theory) is used to explore the intrinsic relationship between system throughput concurrency and end-to-end QoS of inter-application communication. Collaborative multi-cloud server distributed system splits services into small-grained microservices that can be deployed on different independent machines, and establishes a budget model for system utility. At the same time, in order to ensure the stability of online services and prevent a single cloud server from abnormalities and causing an avalanche effect, a cluster method is used to ensure the high availability of the system.

[0089] In step S202, feedback control theory (control theory) is used to achieve steady-state quantification of the processing performance of each node. Through the pre-trained business demand budget model and the real-time collection and analysis of feedback obtained network status information, a network optimization model is constructed to learn the effective optimization mechanism for each data in the service customization network to solve the problems of data transmission QoS guarantee and link load balancing, optimize the service customization network, and achieve on-demand expression of end-to-end QoS of each application and automatic decomposition of network performance NP. Among them, the network status information includes link load, link bandwidth, link delay, link jitter, link packet loss rate, and link utilization.

[0090] In step S203, in order to enable the network operating system technology plane to accurately perceive the state of the network, a hybrid method of probe dialing and in-band telemetry is used to observe the actual delay and jitter of the network path, and metadata is injected into and extracted from production environment messages to perceive network status information in real time. At the same time, based on the active measurement of high-precision network information and communication flow information, a network service capability QoS evaluation mechanism is constructed by declaring deviation variance.

[0091] In step S204, based on the observation of actual traffic in step S203, an actual network consumption model is constructed according to the network status information obtained by real-time perception, the consumption record of the network path is recorded, and a posteriori settlement model is established in combination with the network service capability QoS evaluation mechanism obtained in step S203.

[0092] Combining the network ticket supply mechanism and the application ticket consumption mechanism, drawing on the end-to-end itinerary service concept of the third-party travel transportation platform, a cross-domain timely and dynamic platform ticket matching mechanism is constructed. The specific process includes the following steps S301~S304:

[0093] Step S301: Use game theory to realize trusted co-governance of the alliance in an environment where multiple subjects participate in the application network.

[0094] Step S302: construct a communication model between applications across multiple autonomous domains based on the idea of ​​transit and multi-use.

[0095] Step S303: Ensure timely application non-blocking calls of short flows based on a two-phase commit mechanism.

[0096] Step S304: Implementing network service capability tracking after application cross-domain migration based on a consistency update method.

[0097] In step S301, under the wide-area hyper-converged service customization network architecture for distributed cloud, the on-demand customization of service requirements often involves multiple cloud providers to which cloud servers belong. Game theory (operation research) is used to solve the Byzantine Generals Problem in a multi-subject participation environment, and the alliance's trusted co-governance is achieved through dynamic bidding rational incentives. In a multi-cloud provider system, each cloud provider acts as an agent, and each agent aims to maximize its own interests. The concept of Normal-Form Game is introduced. Assuming that there are N agents in the system, the service behaviors that each agent can do are included in A, and the service benefits obtained by each agent are recorded as u. The triple (N, A, u) is constructed, and the utility function is constructed.

[0098] In some embodiments, the overall benefit calculation formula of all cloud providers is as shown in formula (1):

[0099]

[0100] Among them, u(oi) represents the benefits that each cloud provider can obtain; pi represents the possibility of obtaining the corresponding benefits.

[0101] The overall utility function can be decomposed into several sub-utility functions, and in some embodiments, a decision tree is used for solving the problem. The branching is performed according to the value of each random variable, and finally the utility function value on each leaf of the decision tree is obtained to obtain the optimal decision, maximize the benefits of each cloud provider, and realize the on-demand customization capability of the service customization network.

[0102] In step S302, traffic transmission across the wide area network requires the construction of a communication model between applications across multiple autonomous domains, based on the idea of ​​transit and multi-transportation, that is, the transmission needs to go through another or several transit stations from the origin to the destination, and the delay decomposition problem of the cross-domain network path is solved by comprehensive credit cost and other factors.

[0103] In step S303, in order to make all nodes based on the distributed system architecture maintain the information consistency of the routing strategy, a two-phase submission mechanism algorithm is designed. Specifically, a network operating system as a coordinator is introduced to uniformly control the routing strategies of all nodes (hereinafter referred to as participants) and finally instruct these nodes whether to actually submit the routing strategy. That is, the participants notify the coordinator whether the routing strategy is implemented or not, and the coordinator decides whether each participant should continue to transmit or terminate the transmission based on the feedback information of all participants, so as to avoid the retention of the first data packet during the connection establishment period and ensure the timely non-blocking application call of the short flow.

[0104] In step S304, after steps S301 to S303, the network service capability is followed after the cross-domain migration of the application based on the consistency update method, so as to ensure the dynamic maintenance of the communication requirements between applications.

[0105] At the same time, the design of the interface between the application layer and the service expression layer in the new layered structure provided by the present invention is significantly different from the interface between the application layer and the transport layer in the TCP / IP technical system. This is because under the condition that each application uses TCP / IP as the basis for network interconnection, the interface semantics with the transport layer is mainly reflected as "reachable", while the service customization network provided by the present invention requires "on-demand use".

[0106] In some embodiments, the present invention reshapes the semantics of the network programming part in the traditional portable operating system application programming interface (Portable Operating System Interface of UNIX Application Programming Interface, POSIX API), and replaces the traditional socket front and back ends through API and / or using a software development kit (Software Development Kit, SDK), allowing applications to actively express their needs for the network. This is similar to a "ticket buying" process, which goes through multiple levels of intermediate parsing and mapping, and finally pays the fare and gets on the bus with the ticket to obtain the required network capabilities.

[0107] The embodiment of the present invention also provides a computer-readable storage medium on which a computer program is stored, and when the computer program is executed by a processor, the steps of the aforementioned edge computing server deployment method are implemented. The computer-readable storage medium can be a tangible storage medium, such as a random access memory (RAM), a memory, a read-only memory (ROM), an electrically programmable ROM, an electrically erasable programmable ROM, a register, a floppy disk, a hard disk, a removable storage disk, a CD-ROM, or any other form of storage medium known in the technical field.

[0108] In summary, the present invention provides a wide-area hyper-convergence service customization system for distributed cloud. Starting from the three planes of forwarding, control and business, a wide-area hyper-convergence service customization network architecture for distributed cloud is constructed, and the three technical levels of new network bearing, network operating system and cloud network hyper-convergence are combined to realize deterministic transmission, application differentiation requirements and on-demand network service capabilities. By constructing a new service-centric protocol stack layered structure, a wide-area hyper-convergence service customization network protocol stack for distributed cloud is designed, and the TCP end-to-end transmission control is broken from top to bottom by the application layer to realize application-level, cross-wide-area, distributed on-demand interconnection. On the basis of application differentiation requirements and deterministic service capabilities, the interface between the layers of the new protocol stack is further designed, and the network service capabilities are abstracted by introducing the concept of network capability ticket NaaS Ticket in the service interconnection layer and the packet bearing layer, and the trust and anchoring relationship between the network and the application is established, and the efficient collaboration between the network and the application and the matching between the network path and the application are realized, which truly opens up the relationship between the network and the application.

[0109] It should be understood by those skilled in the art that the exemplary components, systems and methods described in conjunction with the embodiments disclosed herein can be implemented in hardware, software or a combination of the two. Whether it is performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of the present invention. When implemented in hardware, it can be, for example, an electronic circuit, an application specific integrated circuit (ASIC), appropriate firmware, a plug-in, a function card, etc. When implemented in software, the elements of the present invention are programs or code segments used to perform the required tasks. The program or code segment can be stored in a machine-readable medium, or transmitted on a transmission medium or a communication link via a data signal carried in a carrier.

[0110] It should be clear that the present invention is not limited to the specific configuration and processing described above and shown in the figures. For the sake of simplicity, a detailed description of the known method is omitted here. In the above embodiments, several specific steps are described and shown as examples. However, the method process of the present invention is not limited to the specific steps described and shown, and those skilled in the art can make various changes, modifications and additions, or change the order between the steps after understanding the spirit of the present invention.

[0111] In the present invention, features described and / or illustrated for one embodiment may be used in the same or similar manner in one or more other embodiments, and / or combined with features of other embodiments or replace features of other embodiments.

[0112] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the embodiments of the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A wide-area hyper-converged service customization system for distributed cloud, characterized in that: include: The application layer encapsulates, encrypts and decrypts data packets of each application; A service expression layer, wherein the service expression layer addresses each application and its attributes of the application layer with a multi-dimensional service identifier, including: hierarchical scheduling of multi-dimensional resources in a multi-dimensional resource pool based on the multi-dimensional service identifier, obtaining a target resource cluster based on the first-level scheduling index, obtaining a target node based on the second-level scheduling index, and distributing tasks of corresponding applications to the target node; the first-level scheduling index includes geographic location, usage cost and cloud server information; the second-level scheduling index includes resource proportion, power consumption and dependency association parameters between resources and corresponding nodes; the attributes refer to the characteristics attached to each application during communication; A service transmission layer, wherein the service transmission layer uses an open interface preset by a white box switch to control the transmission of a service customized network access side; A service interconnection layer, wherein the service interconnection layer uses a software-defined network controller to realize the connection and communication between the service customization network and various wide area networks; A packet bearer layer, wherein the packet bearer layer adopts a segment routing router to forward the data packet according to the network path specified at each node of the system; the packet bearer layer encapsulates the forwarding capability into a NaaS form, matches the forwarding capability with the needs of each application, and builds a consumption matching mechanism; A data link layer, wherein the data link layer adds a frame header and a frame trailer to the head and the tail of the data packet, respectively, to perform frame sealing, and uses a time-sensitive network switch to perform end-to-end transmission of the data packet; The physical layer uses an Ethernet optoelectronic interface for communication; the physical layer provides a multi-dimensional resource pool constructed by a core cloud server and an edge cloud server, and the multi-dimensional resource pool includes at least network resources, computing resources, storage resources, and time resources.

2. The distributed cloud-oriented wide-area hyper-converged service customization system according to claim 1, characterized in that: The communication requirements of each application are analyzed at the application layer to generate a requirement association matrix, and a representation paradigm of the communication requirements is constructed, and the communication requirements are expressed through the service expression layer.

3. The distributed cloud-oriented wide-area hyper-converged service customization system according to claim 1, characterized in that: The relevant factors affecting resource scheduling are abstracted into multidimensional vectors, the multidimensional vectors are orthogonally decomposed to obtain decision subsets of different dimensions, and a multi-level decision tree is constructed, and the scheduling decision is generated using the multi-level decision tree; the relevant factors include resource proportion, power consumption, dependency association parameters between resources and corresponding nodes, and resource energy consumption.

4. The distributed cloud-oriented wide-area hyper-converged service customization system according to claim 1, characterized in that: Based on time slot constraints, taking the jitter and delay of the network path as the target, a real-time queuing model is constructed, and the queuing plan for the corresponding application data packets is generated using the real-time queuing model; the time slot constraints include data packet size, quantity, packet sending time and time resources.

5. The distributed cloud-oriented wide-area hyper-converged service customization system according to claim 1, characterized in that: A hybrid method of single-path scaling and multi-path sharing is used to flexibly change the network path bandwidth to receive burst data.

6. The distributed cloud-oriented wide-area hyper-converged service customization system according to claim 1, characterized in that: Collect and analyze the network status information obtained through feedback in real time, build a network optimization model, and optimize the service customization network; the network status information includes link load, link bandwidth, link delay, link jitter, link packet loss rate, and link utilization.

7. The distributed cloud-oriented wide-area hyper-converged service customization system according to claim 6, characterized in that: A hybrid method of probe dialing and in-band telemetry is used to observe the actual delay and jitter of the network path, and metadata is injected into and extracted from production environment messages to perceive the network status information in real time.

8. The distributed cloud-oriented wide-area hyper-converged service customization system according to claim 7, characterized in that: A network actual consumption model is constructed based on the network status information obtained through real-time perception, the consumption records of the network paths are recorded, and a post-hoc settlement model is constructed in combination with a preset network service quality evaluation mechanism.

9. The distributed cloud-oriented wide-area hyper-converged service customization system according to claim 1, characterized in that: Based on game theory, a canonical game of the cloud provider to which the core cloud server and the edge cloud server belong is constructed with the goal of maximizing its own interests. The overall interest calculation formula of the cloud provider is: ； in, Indicates the benefits obtainable by the cloud provider; Indicates the possibility of obtaining corresponding benefits.

Images