CN114039693A

CN114039693A - Single clock synchronous concurrent network and data circulation method thereof

Info

Publication number: CN114039693A
Application number: CN202111301929.4A
Authority: CN
Inventors: 刘文龙
Original assignee: Xi'an Sidao Aviation Technology Co ltd
Current assignee: Xi'an Sidao Aviation Technology Co ltd
Priority date: 2021-11-04
Filing date: 2021-11-04
Publication date: 2022-02-11
Anticipated expiration: 2041-11-04
Also published as: CN114039693B

Abstract

According to the single-clock synchronous concurrent network and the data circulation method thereof, the global unified data packet is formed to circulate in the network, so that each switch in the network can receive the global unified data packet, and the technical contradiction between more distribution and more concentration of the distributed network in the prior art can be solved; meanwhile, the reference standard of the generation time of the unified data object is consistent through single clock alignment, the design of a whole system with a single time domain is realized, and the problem of multiple time domains of multiple distribution systems in the prior art is solved. The invention can provide a more accurate time-triggered data circulation mechanism without occupying computing resources, and the global unified data packet can express all contents of all external systems, analyze the contents into the contents and decompose the contents into the threshold values and states of all devices in the external systems, thereby solving the difference of function modeling and solving the coupling between task function models in the existing distributed network in the aspects of time, space, communication and protocol.

Description

Single clock synchronous concurrent network and data circulation method thereof

Technical Field

The invention belongs to the technical field of data driving, and particularly relates to a single clock synchronization concurrent network and a data circulation method thereof.

Background

In modern equipment design and development, the essence of modeling system functions is "behavior modeling" for "state data" and "event data", i.e., when, where, and in what external states, what actions are performed after a "thinking decision" (arithmetic process). The state data, the time data and the thinking decision have data characteristics and can be converted into a digital system for processing.

In a digitizing system, whether temporal, spatial, state, or event, specific "data" is ultimately used for expression. In the equipment design of the traditional digital system, the architecture of the traditional digital system is usually three layers (an execution layer, a device processing layer and an information processing layer) and two networks (the device processing layer adopts various buses and the information processing layer adopts a medium-high speed network), and an application program of the traditional digital system is directly coupled with the physical hardware and the protocols no matter the device layer or the information processing layer. The executive layer equipment has a "more physically distributed" characteristic, while the information processing process (computing equipment) has a "more information centralized" characteristic, and the "central maintenance system" of some large equipment even needs all data of all subsystems, equipment and sensors; i.e. centralized management of the distributed system.

In the distributed system, because a plurality of buses/networks coexist, the networks or the buses eliminate spatial differences, and time, state data and event data are respectively described, so that a scheduling strategy and event processing are emphasized, and the reason is that the resource selection sequence is in a problem during top-level design. Each molecular system usually runs according to a local clock of the molecular system, and no matter communication, scheduling or synchronization of a global clock, the molecular systems have a time division and master-slave mode characteristic physically or a time division and master-slave mode characteristic logically, and the molecular systems are subjected to cascade layering through the same mode, so that the whole system has a multi-time domain characteristic.

In the prior art, the design of digitizing systems is highly dependent on the characteristics of computing resources, communication resources, storage resources. Because the three types of resources are respectively provided by different manufacturers, different departments and even different industries, the equipment personnel usually firstly evaluate the communication resources required by system interaction, then evaluate the computing resources and finally consider the storage resources in the overall design. Communication resource characteristics determine or constrain the design of computational resources, memory resources, such that functional modeling couples various uncertainties due to communication resource characteristics in addition to severe coupling over time. Therefore, once one part of the design is changed, other parts need to be redesigned, so that the development period of the equipment personnel is increased, and the full-service development time is increased.

Disclosure of Invention

In order to solve the above problems in the prior art, the present invention provides a single clock synchronization concurrent network and a data circulation method thereof. The technical problem to be solved by the invention is realized by the following technical scheme:

in a first aspect, the invention provides a single clock synchronization concurrent network comprising a plurality of switches, each external system connected to the synchronization concurrent network through the switches,

the synchronous concurrent network is in a tree network topology, each switch is connected with other switches according to the hierarchy in the tree network, and the switches at the same hierarchy are not connected;

in the single-clock synchronous concurrent network, each switch is aligned through a single clock, so that the generation time nodes of the same data are kept the same in the single-clock synchronous concurrent network;

each switch synchronously receives the uniform data packet of the next level direct connection switch and the global uniform data packet forwarded from the root switch by the previous level in parallel and issues the global uniform data packet to each external system so that the external systems screen uniform data objects required by the external systems from the global uniform data packet;

and each unified object in the global unified data packet forwarded by each switch is consistent in the generation time node of the synchronous concurrent network.

In a second aspect, the present invention provides a data circulation method for a single clock synchronization concurrent network, which provides the single clock synchronization concurrent network of the first aspect, and the data circulation method for the synchronization concurrent network includes:

the root switch receives first uniform data packets sent by the direct-connected switches of the next level in parallel, when the first uniform data packets contain scheduling clock signals, uniform data objects in the first uniform data packets are sequenced according to a preset sequencing rule, global uniform data packets of the current scheduling period are generated, and the global uniform data packets are forwarded to the switches of each level;

the other switches that are not root switches,

receiving a global unified data packet of a current scheduling period, an external system data packet sent by a directly connected external system and a second unified data packet sent by a next-level switch in parallel; the global unified data packet of the current scheduling period is issued to a directly connected external system, so that the external system screens unified data objects required by the external system from the global unified data packet;

analyzing the second unified data packet, and determining that the next scheduling period comes when the received second unified data packet or the external system data packet carries a scheduling clock signal; determining the generation time of a unified data object in each external system data packet according to the clock of the second unified data packet, and determining the generation time of each unified data object in each second unified data packet according to the link delay calculated during the alignment of the single clock; and sequencing the unified data objects in the second unified data packet and the external system data packet according to a preset sequencing rule to generate a third unified data packet and uploading the third unified data packet to the switch directly connected with the previous layer.

According to the synchronous concurrent network and the data circulation method thereof, the global unified data packet is formed to circulate in the network, so that each switch in the network can receive the global unified data packet, and the technical contradiction between more distribution and more concentration of the distributed network in the prior art can be solved; meanwhile, the invention enables the reference standard of the generation time of the unified data object to be consistent through single clock alignment, realizes the design of a whole system with a single time domain, solves the problem of multiple time domains of a plurality of distribution systems in the prior art, does not occupy computing resources, and can provide a more accurate time-triggered data circulation mechanism.

The present invention will be described in further detail with reference to the accompanying drawings and examples.

Drawings

FIG. 1 is a diagram of a design model of a data-based system provided by an embodiment of the present invention;

fig. 2 is a schematic diagram of a connection between a synchronous concurrent network and an application layer device according to an embodiment of the present invention;

fig. 3 is a schematic flowchart of a data circulation method of a synchronous concurrent network according to an embodiment of the present invention;

FIG. 4 is a schematic diagram of partitioning a global data virtual space according to an embodiment of the present invention;

fig. 5 is a schematic diagram of a switch synchronization concurrency mode provided by an embodiment of the present invention;

FIG. 6 is a tree network topology formed after root arbitration according to an embodiment of the present invention;

fig. 7 is a schematic diagram of switch synchronization concurrency and clock synchronization provided by an embodiment of the present invention;

fig. 8 is a data flow level diagram of a single clock synchronous concurrent network according to an embodiment of the present invention.

Detailed Description

The present invention will be described in further detail with reference to specific examples, but the embodiments of the present invention are not limited thereto.

Before describing the present invention, the conceptual processes developed by the present invention will be described first.

In the digitization process, all digital representations can be categorized into two categories: one type is periodic "state data" and one type is temporary, bursty "event data". The manner of ignoring the data itself is different, and no matter whether the decision algorithm code of the layer sensor, the actuator or the pure software is executed, all the functional design behaviors can be uniformly regarded as a periodic cycle of consumption data, decision calculation and production data. For periodic 'state data', decision calculation is always fixed and invariable, namely, the solution is performed once every cycle, modeling design is relatively simple, and an algorithm is emphasized. For the 'event data', due to unknown characteristics such as temporality and burstiness, the expression of the 'event data' needs to relate to various information such as time, space, threshold values and threshold values, complex logic processing is often needed, high coupling on the time, the space, the threshold values and the threshold values during behavior modeling is caused, and in a distributed system, factors such as physical communication media, protocols, transmission delay and transmission disorder are also related, so that functional models distributed at different nodes present complex coupling relations, and not only are the description difficult, but also all the cross-linking relations are difficult to completely analyze at the initial stage of design.

Because the coupling extends from the physical layer to the internal code processing process of the functional model at one time, the complex coupling relationship also affects the fault-tolerant strategy and the scheduling strategy of the system when the design is changed or the system is expanded/reduced, so that the design process of modern complicated and intelligent equipment becomes abnormally complicated. In the traditional design, the relation among models of all distributed nodes is described by adopting a plurality of intelligent tools through a complex demand graph, a logic graph, a state graph, a sequence diagram, a use diagram and the like through a methodology of MBSE (multi-class rule-based) so as to achieve the aim of auxiliary design. However, the methodology and intelligent tools do not change or simplify the complex relationships between models, and do not achieve the purpose of simplifying the design.

Referring to FIG. 1, FIG. 1 is a design model of a data-enabled system. For a digital electronic system, the system is a behavior simulation of an objective world with continuous time, continuous space and continuous state, the continuous objective world is divided into enough small time segments (operation periods) to abstract state data and event data into information data, the data information is fused through a distributed network, and automatic intelligent control is realized through a decision algorithm: when the system task function is designed, the execution process of each task function model can be regarded as that external information data is obtained through a certain data exchange mode, a corresponding decision algorithm is executed according to the information data, and corresponding data is output to an external executor or other task function models:

from the methodology, the modeling of the system task function is essentially the 'behavior modeling' aiming at 'state data' and 'event data', namely, what action is executed after 'thinking decision' (operation process) when, where and under what external state. In traditional behavioral modeling, the processing patterns can be unified to two kinds of data:

time-triggered modeling based on periodic "state data": the data is a continuously changing quantity in the objective world, the digital design only discretizes the data and periodically appears, the processing time is predictable and determinable, and the process is actually a decision calculation process triggered by time;

hybrid trigger modeling based on bursty "event data": the event data is triggered based on time, triggered based on events, or triggered based on a mixture of time and events, and processing time is unpredictable and uncertain.

The biggest difficulty in processing the event data is that the description of the event data is mixed with factors such as time, threshold value, transmission delay and the like, and a plurality of events may have a chronological relationship in time (including a causal relationship among the plurality of events). Just because of these uncertain characteristics of "event data", it creates a high degree of coupling and is difficult to unify between multiple models of an application. The change of the event data in the value range does not affect the modeling mode, and the nature of the modeling of the upper application or the time processing mode is influenced.

In a digital system, whether time, space, state or event, a specific 'data' is finally used for expression, and in a traditional design, a distributed system eliminates spatial difference through a network or a bus, and describes time and state data and event data respectively, so that a scheduling strategy and event processing are emphasized. The reason for this is that the order of resource selection is problematic at the top level design: the design of the digital system highly depends on the characteristics of computing resources, communication resources and storage resources, and because the three types of resources are respectively provided by different manufacturers, different departments and even different industries, equipment personnel usually evaluate the communication resources required by system interaction firstly, then evaluate the computing resources and finally consider the storage resources in the overall design. The communication resource characteristics determine or constrain the design of computational resources, memory resources, such that functional modeling couples various uncertainties due to the communication resource characteristics in addition to severe coupling over time. Therefore, the present invention needs to provide a single clock synchronization concurrent network and a data transmission method thereof in view of the above design defects.

The following continues to describe in detail a single clock synchronous concurrent network provided by the present invention.

The invention provides a single clock synchronization concurrent network which comprises a plurality of switches,

the external system may be an equipment system, a communication system, a user system, etc., and all systems to which the single clock synchronization concurrent network of the present invention can be connected belong to the external system described in the present invention, and the present invention does not limit the external system.

The invention takes 2 as an example to explain the application process of the single clock synchronous concurrent network. Referring to fig. 2, fig. 2 is divided into three parts, namely, an execution layer equipment facility, a single clock synchronization concurrent network and a computer internal concurrent network. The single clock synchronous concurrent network not only is physically interconnected with distributed equipment of an execution layer, but also defines all links related to data exchange as generalized communication, including IO conversion, various buses/network communication and communication between computer internal threads/threads, and realizes internal data exchange of the equipment by the same exchange mechanism.

Starting from the existing multi-layer distributed network, in the data transmission process, the links executed by each layer in the existing network are unified to three aspects of data production, data consumption and decision calculation. The exchange links of the producer and the consumer are all born by a single clock synchronization concurrent network.

In the physical connection of fig. 2, any connection between a "producer" and a "consumer", whether one-to-one, one-to-many, many-to-one, or many-to-many, has a "point-to-point" characteristic in the data exchange, that is, between a switch and a switch, and between a switch and an external system.

The invention compresses the data exchange of the point-to-point connection, which is related to physics and protocol, to the edge end of the physical connection port for processing, realizes the conversion of the protocol and the data at the edge of the connection, and exchanges the data in the single clock distributed network, thus not only eliminating the physics correlation, but also realizing the simple adaptation reconstruction system with the original system or equipment.

On the basis of the single-clock concurrent distributed network, the data circulation method of the single-clock concurrent distributed network provided by the invention is continuously introduced.

The data circulation method is applied to a completely data content-oriented networking structure taking a single-clock concurrent distributed network formed by a switch as a core. The external system, namely the application layer, connected with the single clock concurrent network only focuses on the data content in the network topology.

As shown in fig. 3, the data transmission method of the single clock synchronization concurrent network provided by the present invention includes the following steps:

s301, a root switch receives a first uniform data packet sent by a next-level direct-connected switch in parallel, when the first uniform data packet contains a scheduling clock signal, uniform data objects in the first uniform data packet are sequenced according to a preset sequencing rule to generate a global uniform data packet of a current scheduling period, and the global uniform data packet is forwarded to the switch of each level;

the scheduling clock signal of the invention corresponds to the scheduling cycle time length, the scheduling cycle time length can be changed according to the requirements of an external system, and various time lengths are set.

S302, receiving a global unified data packet of a current scheduling period, an external system data packet sent by a directly connected external system and a second unified data packet sent by a next-level switch in parallel by other switches which are not root switches;

the data priority of a plurality of uniform data objects included in the external system data packet is divided into the following parts from high to low: strict real-time data objects related to clock signals and emergency handling events, real-time data objects of periodic status data, and non-real-time data objects of the non-emergency class.

In the actual implementation process of the present invention, strict real-time data objects, such as scheduling, sudden stop, fatal error, etc., are "instant transmission" in both external network, i.e. the network formed by the external system, and internal, i.e. the single clock synchronization concurrent network transmission of the present invention, and are of a first priority; the periodic state data is a real-time data object, and is output data generated by various sensor data and a model per se in decision calculation, and is of a second priority; the unified data objects of the non-emergency class are non-real-time data objects, including stream data, non-fatal errors, various user-defined events, and the like, and are the third priority.

S303, other switches which are not root switches issue the global unified data packet of the current scheduling period to a directly connected external system, so that the external system screens the unified data objects required by the external system from the global unified data packet;

it is worth to be noted that, in the synchronous grid-connected network of the present invention, each unified data object in the global unified data packet is synchronously "streamed" through each port, so as to implement a "data circulation" mechanism, and to implement information integration in the path service process of the network itself. In an application system facing a large bandwidth requirement, multi-channel concurrency is adopted, namely, the multi-channel concurrency is issued to each lower-layer switch in parallel, so that each external system can subscribe in real time, and compared with the single-line bandwidth improvement, the multi-channel concurrent subscription method has the advantages of better instantaneity, lower cost and higher reliability.

S304, the second unified data packet is not analyzed for other switches of the root switch, and when the received second unified data packet or the external system data packet carries a scheduling clock signal, the next scheduling period is determined to arrive; determining the generation time of a unified data object in each external system data packet according to the clock of the second unified data packet, and determining the generation time of each unified data object in each second unified data packet according to the link delay calculated during the alignment of the single clock;

the external system may regard the transmission time of the unified data object as the generation time of the unified data object.

And S305, sequencing the unified data objects in the second unified data packet and the external system data packet of the other switches which are not root switches according to a preset sequencing rule, and generating a third unified data packet to be uploaded to the switch directly connected to the previous level.

According to the data circulation method of the single clock synchronous concurrent network, the global unified data packet is formed to circulate in the network, so that each switch in the network can receive the global unified data packet, and the technical contradiction between more distribution and more concentration of the distributed network in the prior art can be solved; the invention can not only occupy computing resources, but also provide a more accurate time-triggered data circulation mechanism, the global unified data packet can express all information of all external systems, and the external systems analyze the content and decompose the data into the threshold and the state of each device in the external systems, thus realizing the unified processing mode of 'event data' and 'state data' and solving the difference of function modeling, therefore, the invention can solve the coupling among task function models in the existing distributed network in time, space, communication, protocol and other aspects.

In an embodiment of the present invention, the preset ordering rule is:

sorting according to data priority;

when the data priorities are the same, sorting according to the generation time of the uniform data objects;

and when the generation time is the same, sequencing according to the interface sequence of the uniform data packet sent by the switch.

It is worth mentioning that: unified data objects from an external system, i.e., the application layer, contain some data of more urgent events, which need to be processed in time to avoid a possible large failure of the application layer, and therefore need to be sorted according to data priority. When the priorities are the same, which data object is generated first, the data object is transmitted preferentially, and therefore, the data object needs to be sorted according to the generation time. And the generation time of the data objects with the same priority is the same, and when the small probability event occurs, the data objects are sorted according to the sequence of the switch interfaces for transmitting the data objects, and the data objects with the switch interface serial numbers in the front are sorted in the front. Therefore, all data objects can be sequenced, and the data objects with high data priority are preferentially transmitted in the single-clock distributed network.

In one embodiment of the invention, the global unified data packet forms a global data virtual space;

the global data virtual space comprises a plurality of virtual areas, and one virtual area corresponds to an external system and sorts the generated uniform data objects in a scheduling period according to a preset sorting rule to form an external system data packet.

In actual operation, the global unified data packet corresponds to the unified data packet object of each functional module in the global data space. The structured global unified data packet directly enters an application layer from a single clock synchronous concurrent network, a special memory is not required to be established for management, a repeated copying process of a memory database transfer mode is omitted, and after the application layer extracts related data objects in the data, the memory occupied by the data packet is released. That is, in operation, the global data space of each application node (switch) does not require a dedicated memory area to manage the global data space, and all packets simply "flow" through each node in a round-robin fashion. Various data processing processes are triggered by the received event of communication in cooperation with a global single clock.

In an embodiment of the present invention, the receiving, by the root switch, a first uniform packet sent by the next-tier direct-connected switch in parallel includes:

the root switch receives the switching data of each unified data object in the first unified data packet sent by each next-level direct connection switch in parallel at the specific position of the global data virtual space;

the other switches which are not root switches receive the global unified data packet of the current scheduling period, the external system data packet sent by the directly connected external system and the second unified data packet sent by the next-level switch in parallel, and the method comprises the following steps:

and other switches which are not root switches receive the global unified data packet of the current scheduling period, the external system data packet sent by the directly connected external system and the specific position switching data of each unified data object in the global data virtual space in the second unified data sent by the next-level switch in parallel.

It is worth mentioning that: the output data of all the functional modules in all the external systems form a global virtual data space, and the data are organized in the memory according to continuous addresses. While the global virtual data space is defined, the number of communication packets and the total data flow transmitted by the system in a single scheduling period are fixed, and the number and the sequence of transmission are also fixed. The characteristic can ensure that the number of system communication, the communication flow and the communication sequence of the system are not changed after the system is redeployed in the functional module. If the system is an external network, the external network is required to support priority transmission, the external network characteristic cannot be ensured, and the redeployment characteristic after system expansion cannot be ensured. But the modeling mode of the system and the characteristics of the software module in the computer during expansion or contraction are not influenced, for example, the number of the functional modules can be increased by upgrading the number of the cores of the CPU. When the system is in internal process or thread communication, the system takes the global address of the global unified data packet as the priority to transmit:

each global unified data packet of the global virtual data space is a section of continuous memory space, each unified data object is presented in a theme form, each theme has a global unique address, the themes can be continuous or discontinuous in address, but all data are distributed into continuous data space when actual memory data are processed.

As shown in fig. 4, in the distributed system, each computer node reserves a complete data space in units of computer nodes. All unified data objects Upack of each functional model of the present invention constitute the model data package ModelUpack of the model, and each input data thereof corresponds to a corresponding data item of the global data space. In order to ensure the expandability of the system, when the data structure of the global data space is designed, each unified data object Upack is allocated with a continuous memory data (structural body or dynamic memory allocation) corresponding to the sub-system of the application layer, each system can have a plurality of sub-systems, and then all data packet objects of each sub-system form the global data space.

In one embodiment of the present invention, each switch includes a plurality of ports, and the plurality of ports are divided into an uplink port, a downlink port, a same-level port, and a user port;

each level switch receives an external system data packet of an external system directly connected with the level switch through a user port, receives a second unified data packet sent by a next level switch in parallel through a downlink port and receives a global unified data packet forwarded by an upper level switch in parallel through an uplink port;

in each level switch, a uniform data packet generated after the uniform data objects are sequenced is sent to the switch of the upper level through an uplink port;

the uplink port is connected with a port of a switch in a previous layer, the downlink port is connected with a port of a switch in a next layer, the port in the same layer is connected with the switch in the same layer, and the user port is connected with user equipment.

Referring to fig. 5, the switch ports in the present invention are divided into a downlink port, an uplink port, and a port in the same hierarchy, and for the I-th switch, when sending a unified data packet to the I + 1-th switch, the switch sends the unified data packet in parallel through each downlink port, so that each switch in the I + 1-th switch receives the same data at the same time, and the downlink channel and the parallel concurrency are realized. The I +1 th layer switch simultaneously sends a unified data packet to the I layer switch through an uplink port of the I +1 th layer switch, the I +1 th layer switch sequences unified data objects in the parallel received unified data packet, and then the unified data objects are sent out through a downlink port, so that uplink channels and serial concurrence are realized.

In an embodiment of the present invention, the forming process of the single-clock synchronous concurrent network is as follows:

after the power is supplied in any topology network, each switch forming the any topology network obtains the arbitration result of the root switch and the hierarchy of the switch in the any topology network by sending the message of the MAC address of the root switch and the hierarchy of the switch in the any topology network after the preset root arbitration duration or within the root arbitration total duration, and a synchronous concurrent network is formed according to the principle that ports among switches in the same hierarchy are forbidden and uplink ports of the switches with the MAC addresses not being minimum in a plurality of switches in the current layer connected with the same upper layer switch are forbidden;

a root switch in the single clock synchronization concurrent network issues a clock synchronization signal;

and the switch of each layer executes global high-precision unified clock alignment so as to keep the generation time nodes of the same unified data object the same in the synchronous concurrent network and obtain the single clock synchronous concurrent network.

In an embodiment of the present invention, after a preset root arbitration duration or within a total root arbitration duration, each switch forming the arbitrary topology network obtains an arbitration result of the root switch and a hierarchy of the root switch in the arbitrary topology network by sending a message of a root switch MAC address and the hierarchy of the root switch in the arbitrary topology network, and forms a synchronous concurrent network according to a principle that ports between switches in the same hierarchy are disabled and an uplink port of a switch having a non-minimum MAC address in a plurality of switches in a current layer connected to the same upper layer switch is disabled, including:

the method comprises the following steps: when a system of any topology network is powered on, each switch sends an initial message to the connected switches and receives initial messages sent by other switches;

the initial message carries the MAC address of the switch, the information of the switch as a root switch and the distance between the switch and the root switch;

any network topology is managed hierarchically, each switch has 16 ports in a 10-layer network, and the number of the ports of the bottom layer network can reach 65536. Take switch numbers from 0 to 15, respectively, as an example. After the system is powered on, each switch sends an initial message to the switch directly connected with the switch for the first time, and informs the opposite side that the switch is the root switch.

Step two: comparing the MAC address of the root switch with the MAC address in the received initial message to arbitrate the root switch, obtaining the arbitration result of the root switch, and dividing the hierarchy of the root switch in any network topology according to the arbitration result; sending the arbitration result, the level of the arbitration result in any network topology and the MAC address of the root switch to the connected switches in a new message form;

refer to fig. 6. FIG. 6 is a tree topology after root arbitration according to the present invention. After each switch in any topology network receives the message of the opposite side, each switch is used for arbitrating whether the switch is the root switch or not, if not, the role of the switch is switched, and the message of the opposite side serving as the root switch is sent next time. In any network topology, after the physical connections of the switches are fixed, the maximum level of connections between the switches is known. In the case of maximum tier determination, the root switch arbitration process maximum time can be determined.

Step three: each exchanger circularly compares the MAC address of the exchanger with the MAC address in the received new message to arbitrate the root exchanger and obtain the arbitration result of the root exchanger; determining a target switch for sending the MAC address of the root switch according to the arbitration result; dividing the hierarchy of the target switch in any network topology according to the hierarchy of the target switch in any network topology; and sending the arbitration result, the level of the root switch in any network topology and the MAC address of the root switch to the connected switches in a new message form until the arbitration result of the root switch in the message of the root switch reaches global unification or the preset total root arbitration time length after the preset root arbitration time length is reached.

Illustratively, the maximum physical connection distance in an arbitrary topology network is 4. Supposing that the switch 0 is a root switch, for the first time, the switch 0 sends information of the switch 0 which is the root switch to the

switches

2, 8, 10, 12 and 14 which are directly connected with the switch 0, and meanwhile, the switch 0 also receives messages sent by the

switches

2, 8, 10, 12 and 14, wherein each switch in the messages represents that the switch is the root switch; the switch 0 compares the MAC addresses of the switch and each switch and determines that the switch is a root switch; for the first time, switch 2, switch 8, switch 10, switch 12, and switch 14 also compare their respective MAC addresses to the received MAC addresses, taking switch 2 as an example. The switch 2 receives the messages of the switch 14, the switch 0, the switch 1, the switch 8 and the switch 9 for the first time, the switch 2 compares the MAC address of the switch 2 with the MAC address in the message, finds that the switch 0 is not the root switch but the switch 0 is the root switch, and sends the message of which the switch 0 is the root switch to the switch 14, the switch 0, the switch 1, the switch 8 and the switch 9 which are connected with the switch for the second time; taking switch 1 as an example for the second time, switch 1 arbitrates itself as the root switch and sends out for the first time, receives switch 0 as the root switch for the second time, discovers switch 0 as the root switch after arbitrating by itself, and sends to switch 9, switch 5, switch 13, switch 2 and switch 8 that are connected to itself for the third time. And thirdly, after the switch 5 receives the message, the switch 0 is found to be the root switch through arbitration, and the message is sent out for the fourth time. Taking the switch 15 as an example, the switch 15 determines the switch 4 as the root switch through the first three times, arbitrates the root switch as 0 for the fourth time, and arbitrates the root switch 0 as the root switch for the fourth time. Root arbitration can be achieved in four passes.

The preset arbitration time length is positively correlated with the maximum possible layer number of any network topology, and the root arbitration total time length is set according to the maximum possible network layer number of any network topology. The preset arbitration duration is the product of the duration of performing root switch arbitration once and the maximum possible layer number of any network topology, and the preset total root arbitration duration is 40 microseconds.

Referring to fig. 6, in fig. 6, after the initial message transmission, it is assumed that switch 0 is the root switch. The physical connection distance between the switch 2, the switch 8, the switch 10, the switch 12, and the switch 14 and the root switch is 1, the hierarchy of the switches directly connected to the root switch is 1, and then each switch only needs to determine a target switch in an arbitration result, and the hierarchy of the switch can be obtained according to the hierarchy of the target switch. Thus, after the root switch arbitrates, the hierarchy of any network topology is converted into a tree structure, and the root switch is a root node of the tree network topology. The physical connection distance between the root switch and the root switch is 0, the hierarchy of the root switch in the network topology is 0, the hierarchy of the switches in any network topology is determined in sequence, and therefore the root switch is arbitrated by the root switch, the role of the root switch is converted according to the arbitration result, and other switches are informed to achieve microsecond-level root arbitration decision.

It will be appreciated that data transmitted at the switch meets the constraint of a shortest communication frame length, which is at least 64 bytes in length, and that the ethernet protocol specifies that the shortest communication frame length also has a delay of 64 bytes. Under the background that the transmission of a single byte needs to consume 8 nanoseconds, the time for arbitrating the root switch once is not less than (64+64) multiplied by 8-1024 nanoseconds, so that the result arbitrated by the root switch can be ensured. Of course, since the shortest communication frame length is only the specification of the existing ethernet protocol, the 1024 ns limit is also only to enable the root arbitration process to better adapt and match the existing ethernet.

It is worth mentioning that: in the background of the existing ethernet, the time duration for each switch to perform root arbitration once is about 2 microseconds, and the arbitration time duration of the present invention is the product of the time duration for single root switch arbitration and the maximum possible layer number of any network topology, when the maximum possible layer number of any network topology is determined. In the existing network architecture, the maximum possible number of layers is 10. Therefore, the invention selects the preset arbitration total time length to be 40 microseconds, wherein not less than 20 microseconds in 40 microseconds is the root arbitration time, and the rest time is the time required by single clock alignment.

The invention can set the preset arbitration total time length to be 40 microseconds, so that the switch can obtain a globally unified root arbitration result within 40 microseconds, and certainly, the maximum possible level is determined in any network topology, so that the switch can be determined to arbitrate the globally unified root arbitration result after the corresponding time length.

In this context, the present invention preferably selects a duration of 2 microseconds. The length of time that each switch cycles through root arbitration is no less than 2 microseconds.

Step four: each exchanger obtains the arbitration result of the root exchanger in global unification in any network topology and the hierarchy of the root exchanger in any network topology;

step five: and according to the principle that ports among the switches in the same layer are forbidden and the uplink ports of the switches with the MAC addresses not the minimum in a plurality of switches in the current layer connected with the same upper layer switch are forbidden, forming the synchronous concurrent network.

It is worth mentioning that: each switch, when there is a switch connected with a plurality of upper level switches, determining the switch with the minimum MAC address in the MAC addresses of the plurality of upper level switches; disabling the uplink ports except the switch connected with the minimum MAC address, and disabling the ports of the same layer, so that the root switch to the last layer switch are converted into a tree network topology from any network topology;

the tree network topology comprises a root node and a plurality of branch nodes in a tree structure, the root switch is the root node of the tree network topology, each branch node forms the tree structure according to the hierarchy of the branch node, each branch node only has one father node, and each father node comprises at least one child node.

Referring to fig. 6, after each switch is divided into layers, since the connection mode is a full connection mode, after the switches in the same layer are determined at the root node, the ports in the same layer of the switches in the same layer are disabled in order to improve transmission efficiency and avoid broadcast storm when transmitting data. Meanwhile, there may be a case where more than one superior switch, that is, a child node has multiple parent nodes. In order to avoid the situation, the switch in the same layer detects whether a plurality of father nodes exist in the switch, if so, the uplink ports of other father nodes except the father node with the minimum MAC address are forbidden, and therefore, only one father node exists in one child node.

In one embodiment, the data circulation method of the single-clock synchronous concurrent network further includes:

each switch, when detecting that a new switch is accessed, sends a message carrying the hierarchy of the switch, the arbitration result of the root switch and the MAC address of the root switch in any network topology to the accessed new switch;

and the accessed new switch divides the hierarchy of the new switch in any network topology according to the hierarchy carried by the received message.

It is worth mentioning that: after the arbitration is finished, if a redundant link needs to be added, the topological network is expanded. RSTP requires re-execution of the root arbitration, root port traversal process. After the hierarchy of each switch is divided, when the invention detects that a new switch is accessed, the invention sends a message to the new switch, so that the new switch can know the hierarchy of itself only according to the hierarchy carried by the message, thereby confirming the hierarchy of itself in the network topology. Therefore, when a new node is accessed, the decision of the root node does not need to be made again. Therefore, the invention can realize rapid real-time free expansion networking.

In one embodiment, each switch performs round-robin arbitration on the root switch to compare its MAC address with the MAC address size in the received message, and obtaining the arbitration result of the root switch includes:

each of the switches is connected to a respective one of the switches,

comparing the MAC address of the user with the MAC address in the initial message to determine whether the MAC address of the user is the minimum;

if the MAC address of the self is minimum, determining that the self is a root switch, and continuously transmitting the initial message to the connected switches;

if the MAC address of the terminal is not the minimum, stopping sending the initial message; the switch with the minimum arbitration MAC address is the root switch, and the arbitration result of the root switch is obtained;

and determining the distance between the self and the root switch in physical connection as the self hierarchy in any network topology.

It is worth mentioning that: each root switch has two roles, namely, spontaneous and forwarding, in the process of root arbitration. If the self is the arbitration self as the root node, the message of sending the self as the root node information, namely the initial message sent when the system is just powered on, is continuously executed. If the forwarding is arbitration, the self is not the root node, the message of the self as the root node information is abandoned, and the message of the root node information is forwarded, so that the role conversion is realized.

In one embodiment, partitioning itself into a hierarchy in any network topology based on the hierarchy of target switches in any network topology comprises:

the method comprises the following steps: determining the physical connection distance between the target switch and the self;

step two: and determining the distance of the physical connection between the target switch and the sum of the levels of the target switch in any network topology as the level of the target switch in any network topology.

It is worth mentioning that: the physical connection distance between the switches reflects the physical connection relationship between the switches. And each switch adds one or subtracts one to determine the hierarchy of the switch according to the hierarchy of the target switch directly connected with the switch. The larger the physical connection distance from the root switch is, the lower the level of the tree network topology in which the root switch is located is.

In an embodiment of the present invention, the switches at each level perform global high-precision unified clock alignment to keep the generation time nodes of the same data the same in the synchronous concurrent network, and obtaining the single-clock synchronous concurrent network includes:

the method comprises the following steps: when receiving a clock synchronization signal sent by a directly connected upper-level switch, each level switch forwards the clock synchronization signal to a directly connected lower-level switch, simultaneously starts timing, and feeds back a response signal to the upper-level switch when timing is finished;

step two: when a next-level switch receives a response signal fed back by any directly-connected previous-level switch, calculating and recording actual measurement link delay between the next-level switch and the previous-level switch according to the time of receiving the response signal, the time of sending a clock synchronization signal to the previous-level switch and the time length of timing;

step three: the switch of each level analyzes the unified data packet sent by the switch of the previous level to obtain the sending time of the unified data packet, and determines the generation time node of the unified data packet according to the sending time and the recorded actually-measured link delay between the switch of the previous level and the switch of the previous level; and analyzing the unified data packet sent by the next-level switch to obtain sending time and actually-measured link delay, and determining generation time nodes of the unified data packet according to the sending time and the actually-measured link delay so as to keep the same generation time nodes of the same data in a single clock synchronous concurrent network.

In practical application, the clock synchronization signal is a frame of data; for example, the clock synchronization signal may be the shortest communication frame that is allowed to flow over the network.

It is worth mentioning that: the clocks of each node device are different, so that data timing errors can occur in the data transmission process. Or the data generation time determined by each device is different according to the time node. Therefore, the clock synchronization signal issued by the root switch (root node) can be unified by the time nodes of the whole network topology. That is, the data received at the same time, no matter which node device in the network topology is, the node device can know what time the data is generated, and the generation time of the data is that the root node is used as a reference to issue the clock synchronization signal uniformly.

In practical application, each switch in the synchronous concurrent network represents a node in a tree structure, and the manner of data transceiving between nodes (switches) based on measured link delay may include:

when the father node sends data to the child node, the father node sends the actually measured link delay and the data to be sent to the child node, so that the child node can acquire the real generation time or the real updating time of the data at the father node according to the carried actually measured link delay after receiving the data. When the child node sends data to the parent node, the parent node can directly determine the real generation time or update time of the data at the child node according to the time of receiving the data and the recorded measured link delay. It should be noted that the manner of performing data transceiving based on the measured link delay shown here is merely an example, and does not limit the embodiment of the present invention; any mode of acquiring the actually measured link delay based on the hardware actually measured mode and transmitting data between nodes based on the actually measured link delay provided by the embodiment of the invention belongs to the protection scope of the embodiment of the invention.

It is worth mentioning that: the root node issues clock synchronization signals to trigger branch nodes of each layer to sequentially forward the clock synchronization signals, timing is started while forwarding, and response signals are fed back to respective father nodes when timing is finished. In the process, for a child node (root node or branch node), the time when the child node sends a clock synchronization signal downwards and the time when a response signal fed back by the child node is received are known, and the time length of timing is also known, so that the actually measured link delay between the child node and the child node directly connected with the child node can be calculated according to the two times and the time length; therefore, the actually measured link delay can be taken into account when data is transmitted and received between the father node and the child node, and therefore the global high-precision unified clock is achieved in the whole network.

The link delay in the invention can be obtained by a hardware actual measurement mode, and the clock synchronization precision of the invention is only determined by the working frequency of a hardware physical chip. The time length of the timing is greater than or equal to 2 microseconds, namely, the interval of the clock synchronization signal at each time is 2 microseconds.

Illustratively, if a high-speed chip with the working frequency of 133MHz is used, the ultrahigh clock synchronization precision of about 8 nanoseconds can be achieved.

The calculation method of the actually measured link delay in the child node is as follows:

in the formula, t₂Time t for the child node to receive the response signal Ack sent by the child node₁For a moment, T, at which a child node forwards a clock synchronization signal Sync to the child node_calFor timing the duration of the time, t_delayIs the calculated measured link delay.

Referring to fig. 7, the I-layer switch forwards data from the minimum transmission unit (byte in gigabit network, 4 bytes in 10G below) received from a certain port, and all switches in the I +1 layer receive data synchronously with a deviation of no more than two clock cycles (16 ns): the board level delay plus line delay of data transmitted from layer I does not exceed 500ns (thousand optical networks) per "flow" through one layer of the network, and these delays are fixed after the system is in operation. The invention performs clock synchronization while forwarding data, and the longest clock synchronization time is about 20 ms.

In the step, by using the global high-precision unified clock, after the tree network topology level is determined, the time of each layer of unified clock is about 2 microseconds, and the arbitration time length principle of each time of root arbitration is the same here and is not repeated. The root arbitration process of the present invention can therefore converge at a subtle level. And each switch of the formed hierarchical network topology can autonomously select a link to switch data and can reestablish a new switching link at high speed in case of port or switch failure.

With reference to fig. 2 and 8, the entire synchronous concurrent network switch performs synchronous clock alignment to implement a synchronous clock domain, and the external system is an asynchronous clock domain. The data circulation method is carried out in an information processing layer, the edge processing layer is mainly an edge calculation process, and a main edge device completes the edge calculation process, such as a Camera communication device Camera Driver connected with a Camera, and other edge devices are not illustrated in a list.

The invention can add a communication management link between the external system and the synchronous concurrent network for buffering to isolate the influence of the externally input asynchronous clock on the synchronous clock domain. The unified data collected by an external system is subjected to primary buffer processing, then the data is extracted by an internal synchronous clock, and an externally input asynchronous clock domain is isolated, so that synchronous clock domains are realized among all nodes in the network.

According to the embodiment of the invention, after the root switch and the interface in any network topology are in failure, the rapid and real-time root arbitration and the reconstruction of the network link are realized, so that the synchronous concurrent network for data transmission and reading is reconstructed, the possibility of data loss is reduced, and the safety of data transmission is improved.

In one embodiment, after each switch obtains a global unified arbitration result of a root switch in any network topology and a hierarchy of the root switch in any network topology, the data circulation method of the synchronous concurrent network further includes:

in the description herein, references to the description of the term "one embodiment," "some embodiments," "an example," "a specific example," or "some examples," etc., mean that a particular feature, structure, material, or characteristic described in connection with the embodiment or example is included in at least one embodiment or example of the invention. In this specification, the schematic representations of the terms used above are not necessarily intended to refer to the same embodiment or example. Furthermore, the particular features, structures, materials, or characteristics described may be combined in any suitable manner in any one or more embodiments or examples. Furthermore, various embodiments or examples described in this specification can be combined and combined by those skilled in the art.

While the present application has been described in connection with various embodiments, other variations to the disclosed embodiments can be understood and effected by those skilled in the art in practicing the claimed application, from a review of the drawings, the disclosure, and the appended claims. In the claims, the word "comprising" does not exclude other elements or steps, and the word "a" or "an" does not exclude a plurality. A single processor or other unit may fulfill the functions of several items recited in the claims. The mere fact that certain measures are recited in mutually different dependent claims does not indicate that a combination of these measures cannot be used to advantage.

The foregoing is a more detailed description of the invention in connection with specific preferred embodiments and it is not intended that the invention be limited to these specific details. For those skilled in the art to which the invention pertains, several simple deductions or substitutions can be made without departing from the spirit of the invention, and all shall be considered as belonging to the protection scope of the invention.

Claims

1. A single clock synchronous concurrent network, comprising a plurality of switches, each external system being connected to the synchronous concurrent network through a switch,

each switch in the single-clock synchronous concurrent network is aligned through a single clock, so that the generation time nodes of the same data are kept the same in the single-clock synchronous concurrent network;

and each unified object in the global unified data packet forwarded by each switch is consistent at a generation time node of the synchronous concurrent network.

2. A data circulation method of a single clock synchronous concurrent network, which is characterized in that the single clock synchronous concurrent network as claimed in claim 1 is provided, and the data circulation method of the synchronous concurrent network comprises the following steps:

the other switches that are not root switches,

analyzing the second unified data packet, and determining that the next scheduling period comes when the received second unified data packet or the external system data packet carries a scheduling clock signal; determining the generation time of a unified data object in each external system data packet according to the clock of the second unified data packet, and determining the generation time of each unified data object in each second unified data packet according to the link delay calculated during the alignment of the single clock; and sequencing the unified data objects in the second unified data packet and the external system data packet according to the preset sequencing rule to generate a third unified data packet, and uploading the third unified data packet to the switch directly connected to the previous level.

3. The data circulation method of the single-clock synchronous concurrent network according to claim 2, wherein the preset ordering rule is:

sorting according to data priority;

4. The data circulation method of the single-clock synchronous concurrent network as claimed in claim 2, wherein the system equipment data packets are formed by sorting priorities of data objects, and the data object generated by each equipment carries a system platform identifier, a system identifier, a subsystem identifier, an equipment identifier and a model identifier from high to low;

the external system data packet comprises a plurality of uniform data objects which are divided into the following data objects according to the data priority from high to low: strictly real-time data objects related to clock signals and emergency handling events, real-time data objects of periodic status data, and non-real-time data objects of the non-emergency class.

5. The data circulation method of single-clock synchronous concurrent network as claimed in claim 3, wherein the global unified data packet forms a global data virtual space;

the global data virtual space comprises a plurality of virtual areas, and one virtual area corresponds to an external system data packet formed by sequencing unified data objects generated by the external system according to a preset sequencing rule in a scheduling period.

6. The method for data circulation in a single-clock synchronous concurrent network according to claim 5, wherein the receiving, by the root switch, the first uniform data packet sent by the next hierarchical direct switch in parallel comprises:

the root switch receives the switching data of each unified data object in the first unified data packet sent by each next-level direct connection switch at the specific position of the global data virtual space in parallel;

and other switches which are not root switches receive the global unified data packet of the current scheduling period, the external system data packet sent by the directly connected external system and the specific position switching data of each unified data object in the second unified data sent by the next-level switch in the global data virtual space in parallel.

7. The data circulation method of the single-clock synchronous concurrent network according to any one of claims 2 to 6, wherein each switch comprises a plurality of ports, and the plurality of ports are divided into an upstream port, a downstream port, a same-level port and a user port;

the uplink port is connected with a port of an upper-level switch, the downlink port is connected with a port of a lower-level switch, the same-level port is connected with the same-level switch, and the user port is connected with user equipment.

8. The data circulation method of the single-clock synchronous concurrent network according to any one of claims 2 to 6, wherein the single-clock synchronous concurrent network is formed by the following steps:

after the power is supplied to any topological network, each switch forming the any topological network obtains the arbitration result of the root switch and the hierarchy of the switch in the any topological network by sending the message of the MAC address of the root switch and the hierarchy of the switch in the any topological network after the preset root arbitration time or within the total root arbitration time; according to the principle that ports among the switches in the same level are forbidden and the uplink ports of the switches with the MAC addresses not the smallest in a plurality of switches in the current level connected with the same upper-level switch are forbidden, a synchronous concurrent network is formed;

the root exchanger in the synchronous concurrent network issues a clock synchronization signal;

and the switch of each level executes global high-precision unified clock alignment so as to keep the generation time nodes of the same unified data object the same in the synchronous concurrent network and obtain the single clock synchronous concurrent network.

9. The method as claimed in claim 8, wherein after a preset root arbitration duration or within a total root arbitration duration, each switch constituting the arbitrary topology network obtains an arbitration result of the root switch and a hierarchy level of the root switch in the arbitrary topology network by sending a message of a root switch MAC address and the hierarchy level of the root switch in the arbitrary topology network, and forms the synchronous concurrent network according to a principle that ports of switches in the same hierarchy are disabled and an uplink port of a switch with a non-minimum MAC address in a plurality of switches in a current layer connected to the same upper layer switch is disabled, the method comprising:

after each switch forming any topology network is powered on in a system of any topology network, sending an initial message to the connected switches and receiving initial messages sent by other switches;

each switch compares the MAC address of the switch with the MAC address in the received initial message to arbitrate the root switch, obtains the arbitration result of the root switch, and divides the hierarchy of the switch in any network topology according to the arbitration result; sending the arbitration result, the level of the arbitration result in any network topology and the MAC address of the root switch to the connected switches in a new message form;

each exchanger circularly compares the MAC address of the exchanger with the MAC address in the received new message to arbitrate the root exchanger and obtain the arbitration result of the root exchanger; determining a target switch for sending the MAC address of the root switch according to the arbitration result; dividing the hierarchy of the target switch in the arbitrary network topology according to the hierarchy of the target switch in the arbitrary network topology; sending the arbitration result, the level of the root switch in any network topology and the MAC address of the root switch to the connected switches in a new message form until the arbitration result of the root switch in the message of the root switch reaches global unification or the preset total root arbitration time length after the preset root arbitration time length is reached;

each switch obtains the arbitration result of the root switch which achieves global unification in the arbitrary network topology and the hierarchy of the root switch in the arbitrary network topology;

the preset arbitration time length is positively correlated with the possible maximum layer number of the arbitrary network topology, and the root arbitration total time length is set according to the maximum network layer number of the arbitrary network topology;

and according to the principle that ports among the switches in the same layer are forbidden and the uplink ports of the switches with the MAC addresses not the minimum in a plurality of switches in the current layer connected with the same upper layer switch are forbidden, forming the synchronous concurrent network.

10. The data circulation method of single-clock synchronous concurrent network according to claim 8, wherein the switches of each layer perform global high-precision unified clock alignment so that the generation time nodes of the same data are kept the same in the synchronous concurrent network, and obtaining the single-clock synchronous concurrent network comprises:

when receiving a clock synchronization signal sent by a directly connected upper-level switch, each level switch forwards the clock synchronization signal to a directly connected lower-level switch, simultaneously starts timing, and feeds back a response signal to the upper-level switch when timing is finished;

when a next-level switch receives a response signal fed back by any directly-connected previous-level switch, calculating and recording the actually-measured link delay between the next-level switch and the previous-level switch according to the time of receiving the response signal, the time of sending a clock synchronization signal to the previous-level switch and the time length of timing;

the switch of each level analyzes the unified data packet sent by the switch of the previous level to obtain the sending time of the unified data packet, and determines the generation time node of the unified data packet according to the sending time and the recorded actually-measured link delay between the switch of the previous level and the switch of the previous level; analyzing the unified data packet sent by the next-level switch to obtain sending time and actually-measured link delay; and determining the generation time node of the unified data packet according to the sending time and the actually-measured link delay, so that the generation time nodes of the same unified data object are kept the same in the single clock synchronous concurrent network.