Introduction As information technology (informationtechnology, IT) and operational technologies (operation technology, OT) of continuous integration, the need for a unified network architecture becomes urgent. Intelligent manufacturing, industrial networking, the development of big data is making this integration has become more urgent. The IT and OT for the different needs of communication has also led to a long period of time, the integration of these two areas there has been a big obstacle: Internet data and information in the field of the need for greater bandwidth, but for the industry, the real-time and certainty are key issues. These data usually can not be transmitted in the same network. Therefore, to find a unified solution has become an inevitable demand of industrial integration. Time-sensitive network (timesensitive network, TSN) is the international industry is actively promoting the new industrial communication technology. Time-sensitive network allows periodic and non-periodic data transmission in the same network, so that the standard Ethernet transmission has the advantage of certainty, and through vendor-independent standardization process, has become the focus of a wide range of key technologies. Currently, IEEE, IEC and other organizations have developed in the industrial application based on the underlying network TSN interoperability standards and specifications . Development and the needs of real-time communication technology 1 1.1 bus early age in the 1970s, with the programmable logic controller (programmable logic controller, PLC) produced, distributed control required for the bus also was born. So far, bus technology has been developed for nearly 50 years, Pioneer has developed a variety of each bus, which has a different index in multiple media, signal level, parity, physical interface, baud rate and so on. In the 1990s, as competition intensifies, companies in IEC fight for dominance, resulting in a “battle bus.” IEC thus creating as many as 18 bus standard, created major obstacles to access. 1.2 Real-Time Ethernet stage of the 21st century, with the decline in the cost of standard Ethernet, the bus began to enter the era of Ethernet-based real-time network. In 2001, B & R introduced the Ethernet POWERLINK industrial applications; in 2003, based on the Profibus, Siemens has developed a PROFINET, Rockwell, ABB developed the Ethernet-based DeviceNet application layer protocol / IP, Beckhoff developed EtherCAT, Rexroth developed based on the SERCOS SERCOSIII. These networks are powered by a standard Ethernet media, namely the physical layer and the data link layer uniform standard, while still maintaining the original application layer application layer, designed to protect the user’s investment in software assets. After the integration of the needs of 1.3 in 2014, the era of network intelligence, along with the proposed industrial 4.0, industrial networking, intelligent manufacturing becomes increasingly urgent demand for connectivity needs had changed. Typically do not support traditional Ethernet Switch Network (taking into account the delay, usually transparent transmission of the HUB), which is a polling mechanism (e.g., Profinet, POWERLINK, Ethernet / IP) or frame bundling techniques (e.g. EtherCAT, SERCOSIII) such that the standard Ethernet and real time Ethernet data can not be transmitted in the same network. However, for edge computing, industrial networking, intelligent global manufacturing optimization, the manufacturing site control non-real time data needed and required real-time data management and production optimization layer to be centralized through a unified network, a unified data analysis and data processing platform, and can be executed onward transport controller; and some global optimization work does not need to control levels, but to directly to the side or edge of the cloud. This allows the same network needs become urgent. Further, for the end user in terms of manufacturing , produced by the device and systems are often from different business system configuration, and must have a unified network protocol specification. Therefore, vendor-independent bus has become more necessary in the smart era. Industrial development of communication technology shown in Figure 1 .
Industrial development of communication technologies FIG 1
Development of a brief 2 TSN TSN is not in itself a new technology. IEEE released the IEEE 1588  precision clock synchronization protocol in 2002. In 2005, IEEE 802.1 IEEE 802.1AVB set up a working group to begin developing / video transmission protocol set Ethernet-based audio architecture for addressing dataEthernet in real-time, low-latency traffic shaping and standards, while ensuring compatibility with Ethernet. AVB attracted the attention of automotive, industrial and technical organization of enterprises. TSN set up a working group which led to the development of clock synchronization, traffic scheduling, network configuration criteria set series. In this process, by the organization AVnu, IIC, OPC UA Foundation actively promote common standards TSN technology. Industrial enterprises (including B & R, TTTech, SEW, Schneider, etc.) to proceed to the industrial areas of time-critical tasks to develop shaper, set up a shaper of the Working Group, and in September 2016 held the first shaper work in Vienna group meetings. Then, there are more companies and organizations (including the German Industry 4.0 organization LNI, the US Internet industry organization IIC, China edge computing industry alliance ECC, industrial AII Internet Industry Alliance, etc.) to join TSN technology research, and build a number of tests bed. 2019, IEC and IEEE working group set up cooperation IEC 60802, and convened the first working group meeting in Japan in order to develop the industrial sector TSN underlying interoperability can be achieved. Meanwhile, the OPC UA Foundation has also set up (field level communication, FLC) Working Group, the TSN technology integration with OPC UA specification in order to provide high bandwidth for smart manufacturing, industrial areas of the Internet, low-latency, semantic interoperability industrial communication architecture. FIG 2 schematically illustrates a relevant IEEE standards organization TSN release process . Wherein, IEEE802.1Qat  is an early method of network configuration, the IEEE 802.1Qcc  is its enhanced version, published in 2018.
FIG 2 TSN standards published procedure
. 3 TSN TSN basic concept consists of a series of technical standards. Which is divided into clock synchronization, the data stream scheduling policy (i.e. shaper) and the network and user configuration TSN three portions standards. 3.1 VLAN technology according to the network architecture, the network is usually divided into standard Ethernet, Ethernet deterministic. TSN achieve the data transmission capacity of a hybrid network to meet the requirements of the respective poll presence / bundling standard Ethernet frame technique peer distributed architecture, network employed certainty,And enables the network to play to their strengths functions. TSN based on IEEE 802.1Q  VLAN (virtual local area network, VLAN), and the priority criteria. IEEE802.1Q support quality of service (quality of service, QoS). QoS is a basic network technology to provide better services for network communication. It is a network security mechanism is used to solve the problem of network delay and congestion. The original Internet was not designed QoS mechanism. To meet the needs of different user quality of service applications require a network can be configured according to user needs and resource scheduling. IEEE 802.1Q standard is a network comprising a QoS mechanism, it is possible to provide network performance predictable, and efficient allocation of network bandwidth, so that rational use of resources. 3.1.1 TSN TSN is the standard IEEE 802.1Q VLAN. The criteria used to define the insert 4 bytes wherein standard Ethernet frames. Tag bits define the TSN  as shown in Figure 3.
FIG. 3 of the tag bits define the TSN
① identification tag protocol: identifying the type of network, which is representative of a network TSN, labeled 0X8100. ② priority code (prioritycode point, PCP) consists of three-bit code. ③ discard flag: data networks for low QoS requirements, may be discarded, to ensure QoS of the high priority data. ④VLANIdentifier (VID): identifying a VLAN network, the number of subnets 12 represents the support, i.e., 12 th of 40,962, VID = 0 for identifying a frame priority, 4 095 (FFF) as a reserved value. So, VID can represent up to 4094 subnets. This suggests that TSN is for large-scale data transmission and design. 3.1.2 priority TSN has defined the concept of a service level (class of service, CoS) of. For network TSN, corresponding to different priorities of service codes in FIG. 3 PCP. 3 PCP codes define 0 (minimum) to 7 (highest) priority eight, the transmission type corresponding basis, the greatest efforts, excellent effort demanding applications, delay and jitter of less than 100The video ms, delay and jitter is less than with 10 ms audio, the internal network control, network control. Which have different matching network scenario, the follow-up schedule, the design will take into account factors of the data stream scheduling configuration. Only refers to the standard IEEE 802.1Q TSN 3.2 ISO / OSI reference model data link layer of the second layer in a position TSN ISO / OSI model in the. TSN position seven layer architecture of  as shown in FIG.
In the position of FIG. 4 TSN seven layer architecture of
3.3 precision clock synchronization and delay calculation for communications, industrial control, etc., all the tasks are based on the time reference. Thus, accurate clock synchronization is the basis of the standard. TSN must first solve the network clock synchronization calculation and delay issues to ensure the entire network scheduling high consistency. 3.3.1 TSN clock synchronization mechanism by the standard IEEE 802.1AS  developed for industrial and upgraded IEEE 802.1AS-rev  configured. IEEE 802.1AS is based on IEEE 1588 V2 precision clock synchronization protocol development, called generalized gPTP- clock synchronization protocol. gPTP is a distributed master-slave structure, in all its gPTP network clock synchronization with the master clock. First established by the primary and secondary best master clock algorithm (best clock master algrothms, BCMA), it is called master clock (clock master, CM) and the clock (clock slave, CS). Each node runs a gPTP gPTP Engine. PTP IEEE1588 used is clock information distributed by the IP network L3 and L4 of the transport network layer by IPv4 or IPv6 multicast or unicast. GPTP is embedded in the MAC layer hardware, only works L2, data frames directly inserting the time information, and the data frame to the network as each node. gPTP application Rapid Spanning Tree Protocol (papid spanning tree protocol, RSTP). It is a node in a network path planning to generate an optimal path from the network configuration. TSN which is calculated by the bridging node and distributed in tabular form for eachEnd nodes is stored. When a TSN node wants to send data, it checks this table, calculates the shortest path, the shortest path across the network to the transmitting node needs to receive. IEEE 802.1AS clock structure  as shown in FIG. Clock structure
FIG. 5 of the IEEE802.1AS
FIG. 5, lower left 802.1AS most upstream endpoint information from the CM receives time. The time information includes the cumulative time from the GM to the CM upstream. For full-duplex path between an Ethernet LAN, and direct calculation of the local CS CM peer delay measurements and for correcting the reception time. After adjusting (correcting) the reception time, the clock should be synchronized with the local clock gPTP GM domains. SN also supports cross-communication network, each node has a routing table RSTP given. The core 802.1AS that the time stamp mechanism (Timestamping). PTP message at the access port includes 802.1AS function, will trigger the local real-time clock (real time clock, RTC) sampled according to the protocol, and their value is compared with the RTC from the CM information corresponding to the port; path delay using when measurement and compensation technique to match the value of the time clock RTC PTP domain. When AVB PTP synchronization mechanism covers the entire LAN, among the network node apparatus can be exchanged by periodically PTP messages, and adjust the clock to achieve precise frequency matching algorithm. Eventually, all the nodes are synchronized to the PTP same “clock” (Wall Clock) time, i.e. the time the master node. 7 the maximum hop network environment, can be theoretically guaranteed PTP clock synchronization error is less than 1 μs. IEEE 802.1AS-rev clock is a multi-master system, the main advantage is to support a new connection type (e.g., WiFi), to improve the ability to support redundant paths, enhance the perceived time of switching the main clock time network performance. When there is a GrandMaster down, it can be quickly switch to ensure that a new master clock, in order to achieve high availability system. For in-vehicle system, IEEE 802.1AS can be employed; for industrial high availability is considered, using the AS-Rev version. Method 3.3.2 TSN delay measurement network for the network clock, the main clock synchronization accuracyDepending on the residence time (residence time) and the link delay (link latency). In gPTP, the time synchronization procedure with the same manner as IEEEStd 1588-2008: the master clock transmission time synchronization information to all the temporal direct sensing system connected thereto. The sensing system time after receiving the information must be time synchronization by adding to the dissemination of information from the master clock time of transmission of the present node corrects the synchronization time information. If this time is a perception system is time-aware bridge, it must perceive time synchronization information (including additional delay forwarding process) forward after the correction system to the other time it is connected. Delay data transmission process  shown in FIG. These delays can be accurately calculated.
FIG. 6 during data transmission delay
In order to ensure the above-described process is working properly, the whole process has to be accurately known in two time intervals: ① forward delay (dwell time); ② time synchronization information transmission path delays between two system time-aware. Residence time in the time measuring bridge internal perceptual relatively simple; and a delay in the transmission path depends on several factors, including the media-related attributes and path length. For each type of local area network or a transmission path, there are different ways to measure the propagation time. However, these methods are based on the same principle: measuring received the message from a message transmitted by one device and the other device time period, and then sends another message in the opposite direction, and perform the same measurement. In this process, can be calculated Pdelay: (1) a ratio of r: (2) the network delay measurement principle [13-14] as shown in FIG. FIG delay measurement principle (a) 1 step (b) 2-step
FIG. 7 network
Fig.7Schematic diagram of network delay measurement can be seen from Figure 7, network latency measuring step 1 and step 2 two kinds. Because in this network may have a node can not provide accurate clock. For time-aware nodes, since the time information is sent as the data payload, so each node with the time information. For some non-time sensor networksNetwork, it needs to send a transmission to another node after transmitting a data frame time information. Accordingly, IEEE 802.1AS-Rev enhanced support for the step 1, so that real time is improved. 3.4 network transmission For TSN, its data scheduling mechanism is the key. TSN transmission of data  as shown in FIG. Network data received by the port, performs frame filtering, flow measurement, queuing frames. Transmission selection section, TSN scheduling mechanism will play a role. Working Group IEEE 802.1Q defines different shaper (Shaper) mechanism to implement the schedule. It is a transmission selection algorithm (transmission selection algorithm, TSA). Each algorithm corresponding to the mechanism for scheduling, suitable for different application scenarios. Data transmission process as shown in FIG network TSN.
The data transmission process in FIG. 8 TSN
3.5 stream reservation protocol can be seen from Figure 8, the presence of the filtering database of the network, monitoring the status transfer port, the queue manager. These are used for network resource allocation and scheduling problems. While the stream reservation protocol (stream reservation protocol, SRP) IEEE 802.1Qat mechanism is used for one pair of TSN standard configuration. It becomes IEEE802.1Qat CONTROL ENGINEERING China Copyright In 2010, standardization SRP, is incorporated in the standard IEEE 802.1Q-2011. SRP conceptual OSI model defines the flow of the second layer. SRP work is to establish AVB domain registration flow path, the development of AVB forwarding rules, calculate the worst case delay for AVB stream bandwidth allocation. SRP is to allow the speakers network (Talker) with the appropriate network resources to send data to the listener (Listener), and dissemination of this information in the network. And a bridge between the terminal nodes of the same data is recorded a maintenance requirement for one or more speaker listener registered flow path bandwidth resources. SRP is implemented on top of the original IEEE 802.1Qak-MRP multi-registration agreement. SRP standard provides a new multi-protocol Registration Protocol (multiple multicast registration protocol, MMRP) to manage the service stream bandwidth associated attributes, MSRP, MVRP, MMRP signal processing network provides the entire SRP protocol. About SRP mechanism, you can refer to the SRP AVnu documents . 4 flow control standards for the TSN, the standard management of the data stream consists of a series major ways. Universal networks typically follow a strict priority, rather TSN was this lack of certainty transmission mechanism introduces a new network scheduling, shaping method, and put forward a number of different shaper (Shaper) according to different application scenarios demand . This is also the core of the entire scheduling mechanism of TSN. 4.1 credit-based shaper mechanism defined in IEEE 802.1Qav time-sensitive stream forwarding and queuing (forwarding and queuing for time-sensitive streams, FQTSS) the sensitivity of the data forwarding mechanism, and has become the standard IEEE 802.1Q. As a major enhancement to the traditional Ethernet standard line forwarding mechanism, which initially developed primarily for limiting the A / V information buffer. Enhanced burst multimedia data stream will cause a large buffer congestion and packet loss. Packet loss will have to re-contract CONTROL ENGINEERING China Copyright , so that the service experience decline. It uses a credit-based shaper (credit-based shaper, CBS), in response to the data burst and gather information to limit the outbreak. CBS work queue timing  as shown in FIG.
FIG. 9 CBS timing chart showing a work queue
the CBS queue into Class A (Tight delay bound) and Class B (Loose delay bound). If there is no data transmission queue is set to zero credit, credit may be transmitted when the queue A non-negative. If there is data transmission that the credit will be decreased in accordance with SendSlope, while the other queue is IdelSlope speed increases, idleSlope is the actual bandwidth (bit / s), and SendSlope is the port transfer rate supported by the MAC service. Each queue control CBS data stream does not exceed the maximum bandwidth limit (75% maximum bandwidth) configuration. CBS and SRP integration can provide a delay 250 μs / bridge. Overall, IEEE802.1Qav Ethernet jump guaranteed 7 (Hop) worst 2 ms Class A and 50 ms Class B delay. Of course, this delay for industrial applications is unacceptable. In order to get better QoS, IEEE 802.1TSN TG further developed a time-aware Qbv shaper, Qbu preemptive MAC mechanisms. 4.2 temporal perceptual mechanisms of time perception shaper shaper (Time Awareness Shaper, TAS) in order to lower time granularity, more demanding industrial control applications and scheduling mechanism design is currently being used in the field of industrial automation business. TAS is defined by the IEEE 802.1Qbv, is set in advance based on a periodic gate control list mechanism to dynamically provide on / off control for the egress queue. Qbv defines a time window, is a time-triggered network (Time-trigged). In this mechanism the window is determined in advance. The control gate is periodically scans the list, press a different pre-defined order of transmission queues open port. Hardware, software exports have eight queues, each with unique transmission selection algorithm. Transmission control list (gate control list, GCL) controlled by a door. It is more doors open queue control entity determines the software. TAS working principle is shown in Fig.
FIG TAS 10 operating principle
TAS mechanism in order to ensure the data transmission before the network is idle, before starting transmission of the entire set need a guard bandwidth (Guardbound) . Guardband occupy the maximum Ethernet frame transmission lengths to ensure that the worst case – even with the front of a standard Ethernet frame is being transmitted, will not let GCL in the network is occupied before restarting a cycle. 4.3 preemptive mechanism in the MAC TAS mechanism, there will be two problems: ① protection bandwidth consumption of a certain sampling time; ② low risk priority inversion. Thus, TSN 802.1Qbu and IEEE 802.3 Working Group to develop a IEEE 802.3br , to MAC preemption mechanism. Preemptive MAC based transport mechanism  as shown in FIG. Which uses a frame of 802.3TG preemption mechanism, a given divided to two MAC service interface, they referred to may be preempted MAC (pAMC-Preemptable MAC) and fast MAC (eMAC-express MAC). pMAC eMAC may be preempted, after entering the data stack eMAC waiting data transfer is completed before transmission.
FIG transport mechanism 11 based on the MAC preemptive
preemption, guard bandwidth can be reduced to a minimum low priority frame segment. However, in the worst case, the low priority segments can be completed before a next high priority. Of course, this transmission only seize the connecting layer interface – i.e. for preemptive MAC, switches require dedicated hardware support layer MAC chip. 4.4 Periodic queuing and forwarding mechanisms due to the CBS mechanism can be achieved only soft real-time class, path topology leads to an increase continued delay. The worst-case delay associated with buffering requirements topology, hops, switch. Therefore, TSN working group to promote periodic queuing and forwarding (cyclic quening forwarding, CQF) mechanism (also known as peristalsis shaper). As a method of synchronizing a enqueue and dequeue, such operation allows the CQF LAN frame transmission synchronization bridge in one cycle, to achieve zero congestion and packet loss bounded delay, and independently of the network topology exists. IEEE 802.1Qch standard defines CQF be used in conjunction with standard IEEE 802.1Qci each other. IEEE 802.1Qci-t show, according to the time to reach it, speed, bandwidth, each input queue of bridge nodes for filtering and regulation, for protecting the excessive use of bandwidth, burst size and transmission errors or malicious endpoints. CQF mechanism IEEE 802.1Qch adopted strategy followed a “period a week to go step” for data transmission gives certainty. CQF can be combined with the frame grab IEEE 802.1Qbu used to reduce the size of the frame to complete the smallest frame segment transmissionLost cycle time. For CQF work, all frames must be kept within its assigned cycle. Thus, the time period to be considered, so that the period of time intervening bridges the first and last transmissions are aligned to ensure the required waiting time border. For this purpose, the CQF combined policies Qci inlet and IEEE 802.1Qbv shaper, ensure that all frames held in a certain delay range, and to ensure that its transmitted within the allocated time. 4.5 asynchronous traffic shaping mechanism CQF TAS and provide data for the low latency, the time-dependent network is highly synergistic and enhanced in a forced packet transmission cycle. But its efficient use of bandwidth is not high. Thus, TSN Working Group IEEE 802.1Qcr asynchronous traffic shaping (asynchronous traffic shaper, ATS) mechanism. ATS design is based on the degree of urgency of the scheduler. By smoothing each TSN for re-shaping the flow jumps to obtain streaming mode , the implementation of each stream queue, and such emergency priority data streams may be transmitted in priority. ATS runs asynchronously form, bridges and terminal nodes without time synchronization. ATS more efficient use of bandwidth, a load can be run at the mixing time of high-speed connection applications, such as cyclic and acyclic data flow. 5 TSN network configuration standard IEEE 802.1Qcc for TSN, after the clock synchronization, scheduling policy, it is necessary to consider the issue of network configuration. In the AVB, SRP is a distributed network configuration mechanism. In the more stringent industrial applications, we need to be more efficient and user-friendly configuration. IEEE 802.1Qcc is the generally accepted standard configuration. Patterns for TSN centralized network configuration  12.
FIG 12 TSN centralized mode network configuration diagram
For SRP IEEE802.1Qat mechanism is provided, which is a network demand and resource allocation mechanism a distributed manner. New registration or withdraw registration, and request any changes will result in network latency and overload, reduce the transmission efficiency of the network. Therefore, TSN Working Group also provides centralized registration and flow Priceline IEEE 802.1Qcc support, called SRP enhanced mode. In this mode, the system throughReducing the size of the frequency reservation message (relaxation timer), so that when the triggered updates the link status and reservation changes. Further, IEEE802.1Qcc provides a set of tools for managing and controlling the overall network, enhanced by the UNI SRP, by a centralized network configuration (centralized network configuration, CNC) supplemented node. UNI L2 layer provides a generic service method. CNC UNI interact to provide the resource reservation operation, scheduling, and other types of remote management protocol, such as NETCONF or RESTCONF; Meanwhile, IEEE 802.1Qcc with IETF YANG / NETCONF data modeling language compatible. For fully centralized network, an optional CUC node with a communication standard API CNC, for finding a terminal node, the function searches the terminal node functions and user requirements, and terminal nodes TSN optimized configuration. Interaction with its higher level stream reservation protocol (e.g. RSVP) is seamless, similar to a conventional AVB using SRP mechanisms. IEEE 802.1Qcc still supports the original SRP fully distributed configuration mode, allowing between centralized management systems and distributed systems coexist. In addition, IEEE 802.1Qcc support called a hybrid configuration mode to provide mobility services for the legacy device. The configuration management system and the IEEE 802.1Qca reservation path control, and TSN shaper in combination, can achieve zero transmission end congestion loss. For the entire network, there must be efficient, easy to use network configuration, to obtain a terminal node, the bridge node resources, bandwidth, data load, the target address, clock and other information of each node to the central node and pooled into unified scheduling for optimal transmission efficiency. 6 TSN TSN prospect of very broad application prospects, at present, focusing on the following aspects. 6.1 the automotive sector in the automotive industry, with the development of advanced driver assistance systems (advanced driver assistance system, ADAS), the urgent need for greater network bandwidth and response capabilities to replace the traditional CAN bus. IEEE 802.1AVB is the automobile industry standards group initiated and being implemented. CurrentlyAudi, Mercedes-Benz, Volkswagen, have begun TSN Ethernet-based application testing and validation work. 2019, Samsung initiated by the Automotive Industry Development Alliance invested $ 90 million to TTTech, joint development of automotive electronic systems based on Ethernet. 6.2 Industrial Internet of Things industry things will mean a wider range of data connectivity needs, through machine learning, digital twin technology to better play the role of data, to provide support for the entire production line optimization. These data (including machine vision, AR / VR data) would require a higher bandwidth. Therefore, from the ICT sector CISCO, Huawei and other companies will focus on target over TSN network architecture to achieve this interconnection needs by OPC UA. OPC UA data played the role of norms and standards, and the TSN gives it real-time transmission capability. This architecture can be efficient from the sensor is connected to the cloud, in many scenarios may be omitted fall directly traditional industrial controller architecture layer, forming a new distributed computing architecture. Currently 6.3 industrial control, industrial areas, including B & R, Mitsubishi, Siemens, Schneider, Rockwell and other mainstream manufacturers have introduced products based on TSN. B & R switch introduce new TSN, PLC, Mitsubishi servo drive is used TSN art. Future, TSN will become the mainstream of industrial control field bus. TSN significance for the industry in terms of not just real-time, but rather by TSN realized from the control connection to the entire plant. TSN is the IEEE standard, more “neutral”, which has been widely supported. Future, TSN will become the common choice for industrial communications. 7 Conclusions This paper is intended to be introduced by the current panorama of international industry forefront of time-sensitive network, to explore the need for its cutting-edge development, technology, and to explore its application scenarios by analyzing the conduct of its key technology , technology path. Hope that through reading this cutting-edge technology, the domestic intelligence in manufacturing, industrial networking, big data applications in the field of R & D personnel to keep abreast of these industry trends, hot pursuit of international forefront of the development is conducive to adapt to our own respective products and technologies advance .