Method for high-performance data transmission in a data network with partial real-time requirements and system for carrying out the method

Description

[0001] Industrial applications, such as distributed IO systems in industrial automation, depend on the availability and reliability of deterministic data transport. This means that the data to be transmitted is delivered to the recipient as reliably as possible in real time, with low latency. Data transmission in these networks is characterized, among other things, by the fact that smaller, predictable data volumes are transmitted at regular intervals in a pre-planned manner, for example, measured values ​​from sensors or control commands to actuators in the industrial plant. The data volumes are usually summarized in a burst, also called a stream, i.e., a set of multiple data frames.

[0002] In the past, hardware specifically developed for this purpose and for use in industrial environments was used to build industrial networks, preferably in BUS topology, meaning all participants are connected via a common message path. Such a BUS system is exemplified in the Figure 1 shown, whereby various participants 11, 12, 13 are connected via the transmission link NB, containing the network elements B0 ... B3, which can communicate with each other.

[0003] However, the BUS system is not used exclusively; an alternative topology is, for example, the Ring NR, as in the Figure 2 shown, here the network elements B0 ... B3 are connected to each other in a ring structure, which allows two transmission directions when transmitting data packets between the connected subscribers 11, 12, 13.

[0004] Currently, standardized Ethernet technology is used in almost all modern networks for data transport.

[0005] Manufacturers of industrial automation systems are therefore developing solutions based on Ethernet, with dedicated hardware extensions to meet the requirements of the applications and guarantee the desired real-time behavior of the network.

[0006] One of these systems is PROFINET, short for Process Field Network. PROFINET uses TCP / IP and IT standards, is real-time Ethernet capable, and enables the integration of fieldbus systems. PROFINET itself defines two real-time protocols: RT (Real Time) and IRT (Isochronous Real-Time) to transmit time-critical data for PROFINET IO applications with different real-time requirements.

[0007] Profinet RT runs on standard Ethernet hardware and handles Profinet IO applications with cycle lines of up to 10 ms. At the data plane level, it uses the priorities of standard Ethernet switching technology to transmit real-time data with a higher priority than normal data that does not need to be transmitted in real time. The advantage of Profinet RT is that it does not require time synchronization of the network elements, thus requiring minimal engineering effort.

[0008] However, the exclusive use of Profinet RT offers only a limited real-time behavior, because even the transmission of a high-priority data packet has a transmission delay of up to one maximum data frame per hop from one network element to the next.

[0009] For IO applications with fast control loops such as motion control in a drive system, Profinet IRT offers a high degree of determinism and can provide cycle times below 1 ms, down to 31.25 µs.

[0010] Profinet IRT uses time synchronization with a jitter (i.e. a temporal clock jitter during the transmission of digital signals) of less than 1 µs, a transmission with time planning (time-scheduled, similar to TDMA, Time Division Multiple Access) and so-called "cut-through" switching, i.e. the switching station forwards the data frame to be transmitted before it has been completely received by the previous switching station.

[0011] As already explained, these two transmission methods require specific hardware that is suitable for use in an industrial Ethernet.

[0012] An improved method for Profinet IRT is already known from the paper "Automatic Packing Mechanism for Simplification of Scheduling in Profinet IRT" by Ralf Schlesinger et al., IEEE Transactions on Industrial Informatics, Vol. 12, No. 5, ISSN 1551-330, which features frame transmission synchronization. The subframes are packed into a carrier frame with a predefined transmission time window.

[0013] Cena G. et al. describes in "Performance analysis of Ethernet Powerlink networks for distributed control and automation systems", Computer Standards & Interfaces, Vol. 31, No. 3, ISSN 0920-5489 a solution for industrial communication based on the Ethernet Powerlink standard, and in particular the transmission of acyclic real-time data.

[0014] No-wait Packet Scheduling for IEEE Time-sensitive Networks (TSN), by Frank Dürr et al. in Real-Time Networks and Systems, ISBN 978-1-4503-4787-7, describes a transmission of data packets based on Ethernet, where the switching nodes are time-synchronized when forwarding the data packets.

[0015] The principle of IRT is to protect IRT data transfer (the transmission of IRT data) from interference caused by other data transfers (RT and other non-RT data). This is achieved by using dedicated time windows (reserved bandwidths for the "Red Phase," "Green Phase," and "Yellow Phase" of Profinet IRT) and minimizing the delay of IRT data frames at the bridges (switching centers) through the use of cut-through switching described above.

[0016] The "Red Phase" is strictly reserved for PROFINET RT Class 3, which has the highest priority, meaning no other data packets may be transmitted during this time.

[0017] To meet these requirements, an offline engineering tool is needed to calculate the transmission time of each IRT data frame at the source (also called "injection time") and the IRT schedule at each bridge. To guarantee that all IRT frames are forwarded at each stage using cut-through switching without interfering with each other, the IRT scheduling tool must plan with the maximum synchronization error and calculate a sufficiently large time gap as a safety buffer for the respective injection time between any two data frames generated / sent by two different data sources and which are to be forwarded successively from the same egress port of a switch (bridge) along a transmission path. The forwarding process itself is highly dependent on the underlying hardware implementation.The schedule takes all hardware influences into account and is therefore only applicable in the known topology, with hardware devices considered in the planning. Such conflicts would lead to data loss because caching / queuing is typically not offered by the cut-through switch, and the data frame is discarded.

[0018] In the following, we will also refer to system design. This means that a predetermined number of data frames must be guaranteed to be transmitted in a network, i.e., without data loss and within a guaranteed transmission time, regardless of the underlying topology. For this purpose, the maximum number of RT data frames that can be guaranteed to be transmitted in the network without data loss is determined, given a known maximum number of hops in the network and maximum packet length of the RT frames (also called streams in TSN), as well as a known latency (delay time) at the switching centers.

[0019] Ethernet protocols are constantly being developed. The IEEE's AVB (Audio Video Bridging) Working Group has defined a set of features for the reliable transmission of audio and video data within a maximum latency.

[0020] A new type of traffic, called "reserved traffic," is introduced. Periodically occurring audio and video data is transmitted in so-called "streams." Network configuration information is hidden from the user by the Stream Reservation Protocol (SRP, IEEE 802.1Qat). It provides an access control mechanism implemented in combination with resource reservation, thus guaranteeing end-to-end latency for this periodically occurring data traffic (stream).

[0021] Furthermore, TSN has introduced the TAS (Time-Aware-Shaper, IEEE 802.1 QBR) mechanism as a new forwarding behavior to achieve the lowest possible latency.

[0022] Forwarding is solved by a schedule similar to IRT and enables the development of real-time systems with the lowest latency for each data frame.

[0023] In summary, almost all manufacturers of industrial real-time systems currently use hardware specifically developed for this purpose to achieve high-performance systems. However, these systems exhibit limited flexibility when it comes to changing application scenarios. Alternatively, standardized hardware is used, which exhibits poorer performance in more flexible application scenarios.

[0024] As already described above, PROFINET defines RT (Real-Time) and IRT (Isochromos Real-Time) for Profinet IO applications with different real-time requirements.

[0025] Profinet RT runs on standardized hardware and is not dependent on time synchronization mechanisms. It requires little upfront engineering effort. Because the concept is based on data packet prioritization, it offers only limited real-time capabilities.

[0026] Profinet IRT requires special hardware to provide a high degree of determinism based on upfront offline engineering and planning. The scheduling concept is less flexible and must provide safety buffers for stability reasons.

[0027] AVB introduces a highly flexible system that offers a simple protocol for end stations to meet the flexibility needs of industrial applications.

[0028] The performance or capability of an AVB network is sufficient to meet the requirements of audio and video applications, but the AVB systems are not powerful enough to meet the latency requirements of industrial applications.

[0029] TSN is a highly deterministic system that provides a mechanism for configuring time-based transmission within switches / bridges. This provides the provider of a TDMA-based system with the calculation method and scheduling concept for the required configuration of these systems outside of IEEE standardization activities.

[0030] The object of the invention is to provide an improved method and a device for high-performance data transmission in an industrial data network with at least partially real-time requirements, which can be implemented with standard Ethernet switching elements for the communication network.

[0031] This object is achieved by a method for transmitting real-time data having the features according to patent claim 1.

[0032] The problem is also solved by a system having the features according to patent claim 9.

[0033] The device, in particular the switching node, is suitable and designed for the preferential, high-performance transmission of a quantity of data packets on the way from a data sender to a data receiver, which are received by the switching node via an input port and transmitted via an output port to a second switching node, wherein a time slot of the transmission bandwidth is exclusively reserved in each case and the start of the transmission time slot is time-synchronized at all transmitting switching nodes in the network, wherein a check is carried out before a new reservation to determine whether this is still permitted.

[0034] Further advantageous embodiments of the invention are described in the subclaims.

[0035] The proposed method and apparatus describe a real-time network schedule for industrial control and monitoring applications, using standard Ethernet switching elements for the communication network, particularly based on the new mechanisms defined by the IEEE 802.1 TSN Task Group. The schedule combines time-clocked data transmission with the stream reservation concept, thus providing determinism with guaranteed maximum latency and runtime access controls.

[0036] The determination of the above-mentioned system design can be done in the four steps described below: A worst-case topology is assumed (e.g., a bus system with maximum distance between transmitters and receivers). The TAS window length for the cycle is specified, e.g., 50% of the time is allocated for the scheduled real-time transmission and 50% for other data traffic. The remaining burst length is calculated as the difference between the window length and the latency component. The burst length calculated in this way then determines the number of RT frames that can be transmitted within this time.

[0037] As long as the assumed boundary conditions are met, participants with RT data frames can be added and removed at any point in the network without jeopardizing the guaranteed transmission. A reservation protocol checks the boundary conditions (i.e., maximum packet length and maximum number of RT data frames) and, if necessary, rejects further registration.

[0038] The approach introduces a new resource reservation mechanism and simplifies the switching of data packets over the industrial Ethernet by using a so-called "daisy-chain" topology (i.e. the switching nodes are arranged like a "chain", in bus or ring topology, see Figure 1 or 2 ), which is widely used in industrial control systems.

[0039] In other topologies, the maximum number of stations must be limited. Star topologies and meshed networks (i.e., fully interconnected topologies) can theoretically be used, as long as a maximum "hop count"—the number of stations from sender to receiver—can be defined, depending on the currently used topology. Such a transport network can be modeled as consisting of exchanges / bridges and end stations, with the bridges interconnected to form a transport network via trunk ports, as described above.

[0040] The switching centers and bridges are connected to the end stations (the sources and sinks of the generated data) via so-called "edge ports." An example setup is shown in Figure 1 shown.

[0041] The described scheme is topology-independent and requires less configuration information compared to pure TDMA (or similar) systems. The configuration is based on a special model for the conventional reservation mechanism / stream reservation for RT real-time data transfer in a preconfigured RT phase. DATA PLANE

[0042] As a first step toward achieving guaranteed latency in the network, scheduled data traffic is transmitted analogously in IEEE 802.1 Qbv-2016. Similar to Profinet IRT with its red and green phases, the cycle time is divided into two phases. It begins with a protected time window reserved exclusively for the transmission of real-time data (RT phase). This is followed by an unprotected window for other data transfers.

[0043] Unlike Profinet IRT, which requires explicit configuration of the yellow phase, the data transfer scheduled according to the IEEE standard provides an implicit "guard band" function in the gate operations. To align the scheduled time phases of all switches / bridges to common start and end times, which form the basis for scheduled data transfer, based, for example, on IEEE 802.1AS-2011 or another time-synchronized protocol such as IEEE 1588, all network elements (bridges and end stations) must be time-synchronized.

[0044] The end stations must feed the cyclic real-time data streams into the network at the beginning of each cycle via the so-called "edge" input ports and transmit all data frames to be transmitted as burst traffic. In each real-time-capable bridge, the received real-time data packets are buffered and placed in the correct output queue associated with the protected transmission window, and then transmitted in store-and-forward mode. CONFIGURATION

[0045] In contrast to PROFINET IRT, which requires complete offline planning and engineering in advance, this design adopts a configuration model that combines both offline configuration and runtime access control. One advantage of the proposed approach is the reduced engineering effort and the offered support for dynamic configurations of real-time data streams compared to other time-based real-time solutions such as PROFINET IRT.

[0046] The main configuration tasks that must be performed offline in advance are primarily in the area of ​​transport classes (QoS - Quality of Service), including transport class parameters and setting up scheduled data transfer.

[0047] A network-wide, consistent priority value must be defined as a QoS identifier and used by all real-time streams. The mapping of this priority value to a transport class (outgoing queue for real-time transmission) on each individual bridge is part of the bridge-local configuration. To set up the cyclic schedules for time-triggered data transmission, a control list must be configured per port gate on each bridge, using the managed objects specified in IEEE 802.1 Qbr. All bridges in the network must operate with the same schedule.

[0048] To realize automatic network configuration at the stream level at runtime, the principle of the Stream Reservation Protocol (SRP), IEEE 802.1 Q 2014, can be used in an advantageous embodiment. It was developed as a plug-and-play stream configuration mechanism for AVB systems.

[0049] The basic principle is to apply the protocol to the network to perform access control, latency control, bandwidth and resource reservation for each data stream that needs to be transmitted on the network. WORST-CASE LATENCY

[0050] A key aspect of the stream reservation mechanism is the ability to calculate the worst-case latency, which depends on the maximum frame size and the maximum number of hops in this design.

[0051] The calculation is based on the synchronized ingestion of RT data frames by the end stations at the beginning of the RT phase and a known maximum network diameter, which defines the worst-case latency of a single frame in the network. This enables simplified resource planning within the network.

[0052] RT frames are sent across the network in a scheduled transmission window reserved for real-time data transmission, preventing other traffic from interfering with the transmission. Pre-known inject times offer the opportunity for simplification—all bridges are configured with the same starting point and the same transmission window size.

[0053] A further simplification is based on the previously known maximum PDU (Protocol Data Unit, data frame) size of the real-time transmission, the introduction of a maximum size of the RT class simplifies the calculation of the worst-case latency and enables a new simplified model for the SRP (Stream Reservation Protocol).

[0054] Guaranteed maximum latency can be achieved if all real-time streams of the current cycle complete their data transmission before the estimated transmission window of the current transmission cycle expires, resulting in a bound maximum latency (processing time, also called "make span" for communication between industrial control devices).

[0055] The principle behind calculating worst-case latency is similar to the concept used in Stream Reservation for AVB systems with CBSA (Credit-Based Shaper) forwarding on each subnetwork link, where latency is calculated on a subnetwork link basis—not end-to-end in this approach. Therefore, the described system enables lower latencies compared to the AVB network. An AVB network can be interpreted as a single AVB bridge, and the Stream Reservation reservation scheme can be used.

[0056] The described system combines a simplified scheduling concept with dynamic stream reservation and access control. The advantage is reduced engineering effort and supports dynamic configuration of real-time streams. The new scheme is topology-independent and requires very little pre-calculation of the time schedule, resulting in simplified configuration information. The schedule configuration depends on the maximum hop count and transmission cycles in the network and is identical for each individual bridge.

[0057] All bridges receive the same configuration for the planned traffic during the design and configuration phase.

[0058] All end stations transmit their data in bursts at the beginning of the transmission phase; no frame-granular coordination between different end stations is necessary.

[0059] For each schedule window, the impact of jitter in the time synchronization and the injection mechanisms only needs to be considered once.

[0060] As a result, the performance of the described system can be better than that of PN-IRT in networks with higher connection speeds.

[0061] The effort required for configuration is reduced to a minimum and enables the dynamic addition of end stations and communication relationships.

[0062] The described system can be used to build strong transport networks that achieve high performance with existing TSN mechanisms by using a simple network configuration and avoiding overload situations, as well as guaranteeing deterministic behavior of the network.

[0063] For explanation, further figures are attached, showing Figure 3 the transmission of streams in a protected window, Figure 4 the transmission of streams with additional transmission gaps between the data packets, Figure 5 Switching delay during data transmission 802.1, Figure 6 Switching delay during data transmission with storing delay, Figure 7 with sequential switching, Figure 8 with parallel switching and Figure 9 an example with a ring topology and a max. hop count <= 4.

[0064] The Figures 3 and 4show an example transmission of data frames F1, F2, ... Fm on its stations from the transmitters 1 ...m to the receiver L (also called listener). The transmission takes place via several hops B0, B1, ... Bn in the network and is slightly delayed in each case (forwarding delay D1, D2, ... Dn). During transmission, further time gaps G1, G2 can arise between two data frames, which are also included in the total transmission time MS of the data stream. The total transmission time MS is shorter than the transmission time window RW reserved for the transmission of the data stream. A time delay can be seen between the first input of the data packets at the first exchange B0 and the first input of the data packets at the last bridge Bn, also referred to as forwarding delay FDxn.

[0065] The total transmission time MS is made up of this forwarding delay FDxn + the burst length BL.

[0066] Figure 3 shows an embodiment with the forwarding of packets in the store and forward method, ie the individual frame is first transmitted completely and only forwarded to the next station after complete reception. In Figure 4 As an alternative, a cut-through method is shown, in which the forwarding of the data packets is initiated before they have been fully transmitted. Here, the forwarding delay D1, ... Dn is significantly shorter. Interframe gaps G11, ... Gn2 are provided between the individual frames to ensure collision-free transmission. Thus, the burst length BL is presumably longer due to the existing gaps, but the forwarding delay is shorter than in the first example.

[0067] The maximum possible number of streams can be generally determined by the system design (topology with line and transmission in the bundle), see Figure 4 .

[0068] Operational, Figure 9 When a new reservation is made, it is checked whether it is still permitted based on the "Max. Burst Length" from the system design. In the example shown, the maximum number of hops is 4.

[0069] The real network can have fewer hops and a different topology. The transmitters and receivers can be connected anywhere in the network.

[0070] The system design guarantees that the specified maximum number of streams (see "Max. Burst Length") can always be transmitted over the network, even in the worst-case topology with the worst-case arrangement of listeners and talkers (as a "burst of frames"). Additional streams can also be added later. In the example shown, 6 streams are always supported; no further checks are performed based on the current network topology and the paths used by the stream.

[0071] During operation, the resources still available are checked; this value was already determined during the network design, taking the worst-case topology into account.

[0072] The Figures 5 to 8 Now show different embodiments of what happens within a switching center Bx, on the one hand at the input port Rx (burst in) and at the output port Tx (burst out).

[0073] Each data frame F1, F2 consists of a data header, also called a preamble (PRE), and the data portion MPDU_1, MPDU_2. For example, there is a gap (IFG) between the data packets, also called an inter-frame gap. Figure 5 shows an example where, when data is received, the first frame F1 is received completely (storing delay), but then, during the reception of a second frame F2, a corresponding forwarding of the data burst, at least from the first frame F1', is initiated via the output port.

[0074] This means that the switching delay is less than the sum of (IFG + PRE + MPDU) * bittime. Switching delay is only the time in the IFG transmission gap. Data packets MPDU_1 and MPDU_2 are transmitted in one go.

[0075] Figure 6 shows an example where the switching delay time is exactly equal to the sum of (IFG + PRE + MPDU) * bittime. Here, the transmission of each individual data packet is buffered separately.

[0076] The transmission gaps in the representation of Figure 6 This can occur when a bridge is not powerful enough and therefore creates gaps in the transmission – meaning it cannot transmit frames continuously. This occurs during transmission, is undesirable, and must be taken into account when designing the system.

[0077] In the Figure 7 In the case shown, the switching delay time is greater than the sum of (IFG + PRE + MPDU) * bittime.

[0078] Here, we have a transmission gap in the output port. In this case, switching occurs sequentially, meaning the first data frame MPDU_1 is processed first, then the second, and so on.

[0079] Analogously, Figure 8 a case with parallel switching is shown, the dashed area shows that switching for the second data packet had already started before the switching of the first data packet was completed.

Claims

1. Method for the prioritized, high-performance, regular transmission of a set of data packets (F1, F2, ... Fm) in an industrial network (NB, NR) consisting of switching nodes (B0, B1, B2, B3, ... Bn), wherein the set of data packets (F1, F2, ... Fm), on the way from a data transmitter (1, ... m) to a data receiver (L), is transmitted from a first switching node (B0, B1, B2, B3) to a second switching node (B1, B2, B3, Bn), and a transmission time window (RW) of the transmission bandwidth is exclusively reserved for transmitting the set of data packets (F1, F2, ... Fm) from the first switching node to the second switching node for all transmitting switching nodes (B0, B1, B2, B3) in the network (NB, NR), and the start of the transmission time window (RW) is time-synchronized for all transmitting switching nodes (B0, B1, B2, B3) in the network (NB, NR), and prior to the start of the transmission, a system design of the industrial network (NB, NR) must be known and planned, wherein the transmission time window (RW) is reserved taking into account the number of switching nodes (B0, B1, B2, B3) to be passed through that results from the network topology, with the result that the transmission of a burst corresponding to the set of data packets (F1, F2, ... Fm) and with an associated burst length is always guaranteed within the transmission time window (RW), wherein the transmission time window (RW) reserved for transmitting the set of data packets (F1, F2, ... Fm) is greater than a total transmission time (MS) of the data packets (F1, F2, ... Fm) that is composed of the burst length (BL) and a forwarding delay (FDxn), wherein the forwarding delay (FDxn) is the time delay between the first injection of the burst at the first switching node (B0) and the first injection of the burst at the last switching node (Bn) to be passed through.

2. Method according to Patent Claim 1, characterized in that the set of data packets (F1, F2, ... Fm) to be transmitted is provided with a QoS flag.

3. Method according to one of the preceding patent claims, characterized in that the time synchronization of the transmission time window (RW) in the network is carried out according to the TSN IEEE 802.1 AS-2011 standard.

4. Method according to one of the preceding patent claims, characterized in that the transmission time window (RW) is reserved in accordance with the Profinet IRT IEEE 802.1 Qbv standard.

5. Method according to one of the preceding patent claims, characterized in that a store & forward method is used to transmit the set of data packets (F1, F2, ... Fm) in one of the switching nodes (B0, B1, B2, B3), in particular the first of the data packets (F1) to be transmitted is received completely by the switching node (B0, B1, B2, B3) before it is forwarded to the subsequent switching node (B0, B1, B2, B3).

6. Method according to one of Patent Claims 1 to 5, characterized in that a cut-through method is used to transmit the set of data packets (F1, F2, ... Fm) in one of the switching nodes (B0, B1, B2, B3), in particular the first of the data packets (F1) to be transmitted has not yet been completely received by the switching node (B0, B1, B2, B3) before it is forwarded to the subsequent switching node (B0, B1, B2, B3).

7. Method according to one of the preceding patent claims, characterized in that at least one transmission gap (G1, ...) is planned in the set of data packets (F1, F2, ... Fm) for the transmission between the data packets in order to avoid collisions.

8. Method according to one of the preceding patent claims, characterized in that the transmission of the data packets (F1, F2, ... Fm) is suitable for real-time applications.

9. System comprising a plurality of switching nodes (B0, B1, B2, B3) configured for the prioritized, high-performance transmission of a set of data packets (F1, F2, ... Fm) in an industrial network (NB, NR) consisting of switching nodes (B0, B1, B2, B3, ... Bn), wherein the set of data packets (F1, F2, ... Fm), on the way from a data transmitter (1, ... m) to a data receiver (L), is received by the first switching node (B0, B1, B2, B3) via an input port (Rx) and is transmitted to a second switching node (B1, B2, B3, ... Bn) via an output port (Tx), and a transmission time window (RW) of the transmission bandwidth is exclusively reserved for transmitting the set of data packets (F1, F2, ... Fm) from the switching node to a second switching node, and with a time-synchronized start of the transmission time window (RW) for all transmitting switching nodes (B0, B1, B2, B3) in the network (NB, NR), wherein prior to the start of the transmission, a system design of the industrial network (NB, NR) must be known and planned, wherein the transmission time window (RW) is reserved taking into account the number of switching nodes (B0, B1, B2, B3) to be passed through that results from the network topology, with the result that the transmission of a burst corresponding to the set of data packets (F1, F2, ... Fm) and with an associated burst length is always guaranteed within the transmission time window (RW), wherein the transmission time window (RW) reserved for transmitting the set of data packets (F1, F2, ... Fm) is greater than a total transmission time (MS) of the data packets (F1, F2, ... Fm) that is composed of the burst length (BL) and a forwarding delay (FDxn), wherein the forwarding delay (FDxn) is the time delay between the first injection of the burst at the first switching node (B0) and the first injection of the burst at the last switching node (Bn) to be passed through.

10. System comprising a plurality of switching nodes (B0, B1, B2, B3) according to Patent Claim 9, wherein the set of data packets (F1, F2, ... Fm) to be transmitted is provided with a QoS flag.

11. System comprising a plurality of switching nodes (B0, B1, B2, B3) according to either of Patent Claims 9 and 10, characterized in that the time synchronization of the transmission time window (RW) in the network is carried out according to the TSN IEEE 802.1 AS-2011 standard.

12. System comprising a plurality of switching nodes (B0, B1, B2, B3) according to one of Patent Claims 9 to 11, characterized in that the transmission time window (RW) is reserved in accordance with the Profinet IRT IEEE 802.1 Qbv standard.

13. System comprising a plurality of switching nodes (B0, B1, B2, B3) according to one of Patent Claims 9 to 12, characterized in that a store & forward method is used to transmit the set of data packets (F1, F2, ... Fm) in the switching node (B0, B1, B2, B3), in particular the first of the data packets (F1) to be transmitted is received completely by the switching node (B0, B1, B2, B3) via the input port (Rx) before it is forwarded to the subsequent switching node (B0, B1, B2, B3) via the output port (Tx).

14. System comprising a plurality of switching nodes (B0, B1, B2, B3) according to one of Patent Claims 9 to 12, characterized in that a cut-through method is used to transmit the set of data packets (F1, F2, ... Fm) in the switching node (B0, B1, B2, B3), in particular the first of the data packets (F1) to be transmitted has not yet been completely received by the switching node (B0, B1, B2, B3) via the input port (Rx) before it is forwarded to the subsequent switching node (B0, B1, B2, B3) via the output port (Tx).

15. System comprising a plurality of switching nodes (B0, B1, B2, B3) according to one of Patent Claims 9 to 14, characterized in that at least one transmission gap (G1, ...) is planned in the set of data packets (F1, F2, ... Fm) for the transmission between the data packets (F1, F2, ... Fm).

16. System comprising a plurality of switching nodes (B0, B1, B2, B3) according to one of Patent Claims 9 to 15, characterized in that the transmission of the data packets (F1, F2, ... Fm) is suitable for real-time applications.