A high-reliability low-latency data transmission method

Abstract

The application discloses a high-reliability low-latency data transmission method, relates to the technical fields of digital information transmission and time-sensitive networks, and comprises the following steps: binding a time-checking era identifier and a planned absolute gate instance serial number for a data frame; determining a first available gate instance by combining an entry hardware timestamp, a gate window phase and a time required for completing transmission of each bridging device, and calculating an integer residual of the first available gate instance and the planned absolute gate instance serial number; determining a gate instance breakpoint at a first change of adjacent residuals, and marking a current data frame by a data plane at the breakpoint; and triggering a preset conditional gate slot for a corresponding deterministic data flow only downstream of the breakpoint, and enabling the current data frame to be transmitted once from an original gate window or the conditional gate slot after completing a conflict check. The application solves the problem that a time-sensitive network is difficult to timely locate a first incorrect window position and continue a current first abnormal frame during time-checking era switching.

Description

Technical Field

[0001] This invention relates to the fields of digital information transmission and time-sensitive networking, and in particular to a highly reliable, low-latency data transmission method. Background Technology

[0002] Time-Sensitive Networking (TSN) is used to transmit data with strict arrival time requirements between industrial control devices. Unlike ordinary Ethernet's best-effort transmission, TSN assigns a periodic gating window to each deterministic data stream. Bridging devices forward data frames only when the corresponding gating window is open. When all bridging devices operate using the same time base and according to the same gating schedule, data frames can enter the gating window on time at each hop, thus maintaining stable end-to-end latency. Patent CN110495146A discloses a scheme for verifying the scheduling and actual transmission results of time-sensitive networks, used to check whether data frames arrive within the allowed latency.

[0003] Time synchronization involves a brief transition during master clock link recovery, active master clock switching, or clock servo step synchronization. Different bridging devices may not complete gating time mapping updates at the same physical moment. Upstream bridging devices may be operating according to the new era, while downstream bridging devices may still be operating according to the old era, or even in reverse order. In this case, data frames may still be sent as planned upstream, but may have missed the planned gating window by the time they arrive downstream. Patent CN116368777A discloses recording time information at terminal stations and bridging devices in time-sensitive networks to detect early arrivals, late arrivals, and transmission delay anomalies. This type of diagnostic can detect late arrivals, but reporting the anomaly information and then adjusting the scheduling by the controller still requires analysis, solution, and dissemination processes.

[0004] Even a small time synchronization offset can cause a wait of a complete gating cycle. Taking a deterministic data stream with a gating cycle of 1ms as an example, the planned gating window of a bridging device is open between 120μs and 128μs in the current cycle. When a data frame arrives at 204μs, it is only 76μs later than the window closing time and is still within the same 1ms cycle; this data frame can no longer use the gating window for this cycle and must wait for the next cycle to open, incurring an additional wait of approximately 916μs. Therefore, simply determining which cycle the timestamp belongs to is insufficient to determine whether the current gating window is still usable, and merely comparing the phase within the cycle can easily lead to the loss of information about which gate instance the data frame should actually enter.

[0005] Long-term widening of the gating window across the entire path will continuously consume the transmission time of other deterministic data streams, and setting dedicated backup windows for each stream will also increase normal bandwidth overhead. Frame duplication will generate additional packets during normal operation, and retransmission after failure requires waiting for feedback. Existing solutions struggle to simultaneously determine which pair of adjacent bridging devices the error window begins from, and to immediately provide a conflict-free partial transmission opportunity with other deterministic data streams while the first abnormal frame is still at the breakpoint bridge. Therefore, a data transmission method is needed that can identify which gate instance can still be used for the current frame, locate the first gate instance breakpoint, and promptly resume the current frame downstream of the breakpoint. Summary of the Invention

[0006] The technical problem to be solved by the present invention is to provide a highly reliable and low-latency data transmission method to solve the problem that time-sensitive networks have difficulty in timely locating the first misaligned window position and continuing the current first abnormal frame during the time synchronization period.

[0007] To achieve the above objectives, the present invention adopts the following technical solution: A highly reliable and low-latency data transmission method includes: configuring a hop-by-hop gated window and a shared connection resource pool for deterministic data streams in a time-sensitive network, and binding a time synchronization era identifier and the planned absolute gate instance sequence number of each bridging device to the data frame; Each bridging device selects the gate control reference time, cycle duration, and gate control window phase according to the time synchronization era identifier, obtains the entry hardware timestamp of the current data frame, and determines the remaining open duration of the current gate instance; When the remaining open time is not less than the sum of the processing time, the transmission time and the protection margin, the current door instance is determined as the first available door instance; otherwise, the next door instance is determined as the first available door instance. The integer difference between the first available gate instance number and the corresponding planned absolute gate instance number is determined as the available gate instance residual; The location where the residual of an adjacent available gate instance first changes along the path is determined as the gate instance breakpoint, and the data surface at the gate instance breakpoint is used as the current data frame to write the breakpoint marker. Downstream of the breakpoint of the gate instance, the conditional gating slot in the shared continuation resource pool is triggered only for the deterministic data stream carrying the breakpoint marker and the corresponding time synchronization era identifier, so that the current data frame is sent once from the original gating window or the conditional gating slot; The conditional gating slot pre-completes conflict verification. The interval from the completion of residual detection to the slot start point is no greater than the continuation waiting limit. The sum of the detection duration, the interval, the gating trigger effective duration, and the transmission duration is no greater than the transition increment limit, and the transition increment limit is no greater than the continuation waiting limit.

[0008] Optionally, binding the time synchronization era identifier and the planned absolute gate instance sequence number includes: generating a time synchronization era identifier for each gated time mapping change; recording the gated reference time, cycle duration, gated window phase, and complete time conversion relationship with the previous time synchronization era for each time synchronization era; the source end writes the source end gate instance sequence number corresponding to the in-stream frame sequence number into the data frame; each bridging device adds the source end gate instance sequence number to the pre-configured hop-by-hop gate instance advance amount to obtain the corresponding planned absolute gate instance sequence number.

[0009] Optionally, determining the residual of the available gate instance further includes: saving the reference time difference, sub-period phase difference, frequency ratio, and switching time between the old and new eras to form an affine time conversion relationship; when the era used by the bridging device is inconsistent with the era carried by the data frame, the affine time conversion relationship is used to convert the entry hardware timestamp to the era carried by the data frame, and then the first available gate instance is determined; the integer sequence number of the first available gate instance is retained; and the shortest signed ring distance between the converted timestamp phase and the planned gating window phase is determined as the phase residual.

[0010] Optionally, the data plane determines the door instance breakpoint by: the data frame carrying the available door instance residual and path location identifier written by the previous bridging device; the current bridging device calculating the current available door instance residual locally; when the current available door instance residual is not equal to the available door instance residual written by the previous bridging device, and the current data frame cannot be sent within the remaining open time of the planned gating window, the current bridging device determines its own entry point as the door instance breakpoint and writes the breakpoint marker; the first data frame to encounter the door instance breakpoint directly enters the arbitration of the shared connection resource pool.

[0011] Optionally, the pre-configured shared connection resource pool includes: selecting a conditional gating slot within the gating period of the protected port; incorporating the maximum frame transmission duration, maximum non-preemptive blocking duration, upper bound of intra-bridge forwarding jitter, upper bound of clock error, and protection interval into the conditional gating slot and its isolation interval; verifying the separation of the isolation interval from the occupied interval of the scheduled deterministic data stream; and writing the mapping between the breakpoint location, downstream path location, and flow identifier and the same set of hop-by-hop conditional gating slots into the data plane of each bridging device.

[0012] Optionally, triggering the conditional gating slot includes: the data plane at the breakpoint of the gate instance selects a frame to be continued from the candidate abnormal frames in order of deadline; if the deadlines are the same, the frames to be continued are selected in order of flow identifier; a continuation pool identifier is written for the frame to be continued; the bridging device downstream of the breakpoint of the gate instance enables the corresponding conditional gating slot according to the continuation pool identifier; when the same data frame is sent through the original gating window, the corresponding conditional gating slot is closed; when the data frame is sent through the conditional gating slot, the next original gating window is skipped; concurrent abnormal frames that have not obtained a conditional gating slot enter the fault isolation process.

[0013] Optionally, the difference between the cutoff time of the protected flow and the normal forwarding delay is determined as the upper limit of the connection waiting; the start time of each group of hop-by-hop condition gating slots is staggered by the sum of a frame transmission time and the in-bridge forwarding time, so that the data frames entering the shared connection resource pool are continuously forwarded downstream of the gate instance breakpoint after passing through the first condition gating slot; when any port cannot provide a conflict-checked condition gating slot within the upper limit of the transition increment, the corresponding flow enters the fault isolation process.

[0014] Optionally, a breakpoint probe is generated after the first abnormal frame enters the shared connection resource pool. The breakpoint probe carries a flow identifier, a time synchronization era identifier, a gate instance breakpoint, a connection pool identifier, and a source gate instance sequence number. When forwarding the breakpoint probe, each bridging device downstream of the gate instance breakpoint adds the source gate instance sequence number to the hop-by-hop gate instance advance amount of the current hop to obtain the planned absolute gate instance sequence number, and writes the first available gate instance sequence number and the planned absolute gate instance sequence number. The receiving end reads the first available gate instance sequence number and the planned absolute gate instance sequence number of each hop in path order, and arranges the integer difference between the two as a probe gate instance residual sequence.

[0015] Optionally, when all the jump residuals in the probe gate instance residual sequence and the available gate instance residuals of the continuous service frames are 0, a cancellation flag is sent downstream of the gate instance breakpoint to restore the corresponding conditional gate slot to the conditional closed state; when the probe gate instance residual sequence still contains non-zero residuals, the corresponding flow stops using the shared connection resource pool and starts the fault isolation process; after the first abnormal frame enters the shared connection resource pool, the cancellation or fault isolation is performed using the probe gate instance residual sequence and the continuous service frames.

[0016] Optionally, each bridging device stores old-generation mappings and new-generation mappings until the end of the corresponding transition period; when a data frame carries an old-generation identifier, the old-generation mapping is used to determine the first available gate instance number and the data frame is allowed to use the conditional gating slot corresponding to the old generation; when a data frame carries a new-generation identifier, the new-generation mapping is used to determine the first available gate instance number and the data frame is allowed to use the conditional gating slot corresponding to the new generation; after the transition period ends, the corresponding old-generation mapping is deleted when it is confirmed that there are no data frames in transit in the old generation.

[0017] The beneficial effects of this invention are as follows: This invention can directly determine which gate instance is still usable for the current data frame, avoiding the mistaken assumption that a frame still within the current cycle can still catch the current cycle's gating window. The data frame carries a time synchronization era identifier and a bridge-by-bridge plan absolute gate instance sequence number. After reading the entry hardware timestamp, the bridging device compares the remaining open time of the gating window with the time required for processing, transmission, and protection. If the remaining time is sufficient, the current gate instance remains usable; if the remaining time is insufficient, the next gate instance becomes the first usable gate instance. Taking the aforementioned 1ms cycle as an example, with a planned gate instance of 248 and a gating window phase of 120μs to 128μs, when the data frame arrives at phase 204μs, the first usable gate instance is directly taken as 249, and the usable gate instance residual changes from 0 to 1. This determination does not require waiting for the timestamp to cross the entire cycle boundary; it can identify anomalies when the frame has just missed the gating window.

[0018] This invention can pinpoint an end-to-end delay to the adjacent bridging device that first lost its planned gating window, ensuring that recovery actions only apply to the truly affected path segments. Data frames carry the available gate instance residuals written by the previous bridging device along the path. The current bridging device calculates its local residual and compares it with this residual. When the adjacent residuals change for the first time and the current frame cannot be transmitted within the planned gating window, the current bridging device designates its entry point as a gate instance breakpoint. Compared to simply reporting receiver timeouts or hop-by-hop reporting of general early / late deviations, this breakpoint indicates both where the problem originated and provides information that the current frame needs to use the next gate instance, eliminating the need for subsequent bridging devices to switch the entire path to standby scheduling.

[0019] This invention enables direct continuation of the first abnormal frame after a gate instance breakpoint is detected, reducing the response delays caused by centralized reporting, re-solving, and temporary table entries. The shared continuation resource pool is already written into the data plane of each bridging device during deployment. After the breakpoint is marked with a breakpoint flag and continuation pool identifier for the current data frame, downstream bridging devices can trigger the same set of conditional gating slots. Continuous service frames and breakpoint probes only serve for post-event confirmation, early cancellation, or fault isolation; they are not prerequisites for first frame continuation. Existing two-frame confirmations or controller table modifications may not be able to process the first frame in time. This invention first recovers the current frame and then confirms its status, separating the recovery and confirmation actions in time.

[0020] This invention also reduces the cost of long-term network resource occupation for improving reliability. Conditional gating slots are triggered only downstream of the gate instance breakpoint for deterministic data streams carrying corresponding tags, and remain conditionally closed under normal circumstances. During deployment, maximum frame transmission time, non-preemptive blocking, intra-bridge forwarding jitter, clock errors, and guard intervals are already factored into the isolation interval, and conflict verification with existing deterministic data streams is completed. When multiple abnormal frames contend for the same resource, the data plane determines a unique frame to be continued based on the deadline and flow identifier, and the remaining frames enter the fault isolation process. After the same data frame is sent through the original gating window, the corresponding conditional gating slot is closed; after being sent through the conditional gating slot, the next original gating window is skipped, thus eliminating the need for full replication or retransmission to achieve reliability.

[0021] This invention provides a verifiable upper limit for latency, making the low-latency effect independent of ambiguous window width descriptions. For a 512-byte data frame and a 1Gbit / s link, the transmission time is 4.096μs; when the residual detection does not exceed 2μs, the interval from the completion of detection to the start of the conditional gating slot does not exceed 90μs, and the gating triggering and protection margin do not exceed 3μs, the transition increment is 99.096μs. In the aforementioned example, a conventional solution would require a wait of approximately 916μs; this invention can limit the additional latency caused by the timing transient to approximately 0.1ms. The aforementioned latency budget, together with the pre-completed conflict check, defines the connection conditions; flows that cannot obtain a conflict-free conditional gating slot along the entire path enter the fault isolation process, avoiding the trade-off of latency guarantees that would disrupt other deterministic data flows for single-frame connection. Attached Figure Description

[0022] Figure 1 This is a schematic diagram of time-sensitive network hop-by-hop transmission and gate instance breakpoints provided in an embodiment of the present invention; Figure 2 This is a schematic diagram illustrating the mechanism for determining the first usable gate instance provided in an embodiment of the present invention. Figure 3 This is a schematic diagram of the residual changes and breakpoint location of adjacent available gate instances provided in an embodiment of the present invention; Figure 4 This is a schematic diagram illustrating the affine time conversion between the old and new time-keeping eras provided in an embodiment of the present invention. Figure 5 This is a schematic diagram illustrating the triggering of downstream shared continuation resources for a breakpoint, provided in an embodiment of the present invention. Figure 6 This is a schematic diagram of the state transition of a single data frame transmission provided in an embodiment of the present invention; Figure 7 This is a schematic diagram of a highly reliable, low-latency data transmission process provided in an embodiment of the present invention; Figure 8This is a schematic diagram of breakpoint probe verification and connection cancellation provided in an embodiment of the present invention.

[0023] Figure labeling: 10 - Source end; 11 - First bridging device; 12 - Second bridging device; 13 - Third bridging device; 14 - Receiver end; 15 - Deterministic data stream; 16 - Gate instance breakpoint; 20 - Planned gating window; 21 - Entry hardware timestamp; 22 - First available gate instance; 23 - Available gate instance residual; 35 - Breakpoint marker; 40 - Affine time conversion relationship; 41 - Old era timeline; 42 - New era timeline; 50 - Gate instance breakpoint data surface; 51-First downstream bridging device; 52-Second downstream bridging device; 53-Third downstream bridging device; 54-Conditional gating slot; 55-Shared connection resource pool; 60-Candidate abnormal frame; 61-Frame to be connected; 62-Conditional gating slot occupancy status; 63-Send completed status; 64-Skip status; 65-Clear status; 70-Breakpoint probe; 71-First available gate instance sequence number; 72-Probe gate instance residual sequence; 73-Cancellation flag; 74-Fault isolation process. Detailed Implementation

[0024] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions in the embodiments of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this invention, and not all of them. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0025] Figure 1 This diagram illustrates hop-by-hop transmission and gate instance breakpoints in a time-sensitive network according to an embodiment of the present invention. A deterministic data stream 15 originates from the source 10, sequentially passes through the first bridging device 11, the second bridging device 12, and the third bridging device 13, and finally reaches the receiving end 14. Each bridging device has an egress port supporting time-aware shaping, and the port is opened according to a pre-configured gating period and gating window. The gating period can range from 125 microseconds to 10 milliseconds, typically 1000 microseconds. A period that is too short increases the number of gating entries and the switching burden; a period that is too long increases the waiting time caused by missing a gating window. A typical value of 1000 microseconds is suitable for millisecond-level control flow in gigabit industrial Ethernet.

[0026] Source device 10 writes a stream identifier, intra-stream frame number, synchronization time identifier, and source gate instance number to each data frame. The synchronization time identifier is used to distinguish the time caliber before and after a gating time mapping change. The gate instance number uses a monotonically increasing integer to represent the absolute order of the periodic window and does not reset to zero in each period. Source device 10 also initializes the previous position field to the source position and initializes the previous residual to 0. The first bridging device 11, the second bridging device 12, and the third bridging device 13 respectively store the gate instance advance amount of the current hop relative to source device 10, and add the source gate instance number to the corresponding advance amount to obtain the absolute gate instance number of the current hop plan. Each bridging device atomically writes the locally calculated available gate instance residual 23 and the current hop path position back to the data frame at the exit for the next bridging device to read. This approach preserves the information of which gating window the data frame should originally enter, and also enables the first bridging device 11 to compare the reference residual of source device 10 as the residual of the previous hop. When the path location is missing, duplicated, or does not conform to the pre-configured order, the data frame enters the fault isolation process 74.

[0027] In one embodiment, gated time mapping changes are triggered by an active master clock switch, clock servo step correction, or gated reference reloading. The configuration plane generates a new time synchronization era identifier for each change and records the gated reference time, gating period, gating window phase, and complete time conversion relationship between the preceding and following eras for each era. Old records are retained during the transition period. The transition period can be 2 to 20 gating periods, typically 5 gating periods; this range covers the maximum transit time in the link and the device update difference, preventing the arrival of old era data frames when there are no usable conversion bases.

[0028] Figure 7 This is a schematic diagram of a high-reliability, low-latency data transmission process provided in an embodiment of the present invention. The bridging device first reads the time synchronization era identifier and the planned absolute gate instance sequence number carried in the data frame. Next, the bridging device obtains the entry hardware timestamp 21, converts the time according to the era carried in the data frame, and determines the first available gate instance 22. Subsequently, the bridging device calculates the available gate instance residual 23 and compares it with the residual written in the previous hop. If the residual changes for the first time and the current frame cannot be sent within the planned gating window 20, a breakpoint marker 35 is written, and shared connection resources are arbitrated; if the judgment is not valid, subsequent data frames are processed according to the original gating plan. Downstream of the breakpoint is only the matching data frame triggering the conditional gating slot 54, causing the frame to be sent once through the original gating window or conditional gating slot 54. After the first frame completes the connection, probes and continuous service frames are used to determine whether to cancel the connection or enter the fault isolation process 74.

[0029] Figure 2This is a schematic diagram illustrating the mechanism for determining the first usable gate instance provided in this embodiment of the invention. The bridging device calculates the current absolute gate instance and window state to which the entry hardware timestamp 21 falls, based on the gating reference time, gating period, and the phase of the planned gating window 20. The bridging device also obtains the processing time and transmission time from the port capability table. The processing time can be from 0.5 microseconds to 10 microseconds, including the time required for table lookup, arbitration, and queuing. The transmission time is calculated based on the number of bytes in the data frame and the link rate. A protection margin is used to cover hardware timestamp errors, gating trigger jitter, and quantization errors in the implementation; it can be from 0.5 microseconds to 10 microseconds, typically 3 microseconds. A protection margin that is too small will cause the frame tail to cross the gate closing boundary, while a margin that is too large will prematurely abandon a still usable window.

[0030] The gating window 20 is designed with an open boundary of O and a closed boundary of C, and an entry hardware timestamp 21 of t. The bridging device uses the larger value between t and O as the starting point for calculating the remaining open time. The remaining open time is obtained by subtracting this starting point from C. If the calculation starting point is not earlier than C, the remaining open time is clamped to 0, thus avoiding unsigned counter underflow. The bridging device compares this remaining open time with the sum of processing time, transmission time, and protection margin. If the conditions are met, the current absolute gate instance is determined as the first available gate instance 22; otherwise, the next absolute gate instance is removed. This conservative approach still reserves full time for processing when a frame arrives early, and does not mistakenly treat the waiting before opening the window as a transmittable window. When the link supports frame preemption, the preemption verification and recovery overhead is included in the processing time; when the link does not support frame preemption, the maximum non-preemptive blocking time is also included in the processing time, and no fourth term is added to the three comparison terms. The maximum non-preemptive blocking duration can be taken from 0 microseconds to 125 microseconds, and its upper limit is determined by the port's maximum non-preemptive frame and link rate.

[0031] exist Figure 2 In the example shown, the gating period is 1000 microseconds, the planned absolute gate instance number is 248, and the planned gating window 20 is open from phase 120 to 128 microseconds of the period. The data frame arrives at phase 204 microseconds. Although the ingress hardware timestamp 21 still belongs to this 1000-microsecond period, the planned gating window 20 has been closed for 76 microseconds, and the current gate instance cannot complete the transmission. Therefore, the bridging device determines the next gate instance 249 as the first available gate instance 22. Subtracting the planned absolute gate instance number 248 from the first available gate instance number 249 yields an available gate instance residual 23 of 1. This determination directly addresses whether the window can complete the transmission, avoiding the mistake of considering the current window as still usable even if it is still within the same period.

[0032] Figure 4This is a schematic diagram of affine time conversion between old and new time synchronization eras provided in an embodiment of the present invention. The affine time conversion relationship 40 includes at least the reference time difference, sub-period phase difference, frequency ratio, and switching time. Let the time on the old era time axis 41 be t, the switching time be tc, the frequency ratio be a, and the reference and phase synthesis offset be b. For the time within the applicable segment, the corresponding time on the new era time axis 42 is a multiplied by the difference between t and tc, and then tc and b are added. a can be from 0.9999 to 1.0001, and b can be from -1000 microseconds to 1000 microseconds. The specific boundaries are determined by the maximum frequency deviation and step range allowed by the time synchronization system.

[0033] During the conversion, the era carried by the data frame is used as the target caliber. When converting from the old era timeline 41 to the new era timeline 42, the aforementioned forward relationship is used; when converting from the new era to the old era, tc and b are subtracted first, then divided by a, and finally tc is added, forming the inverse transformation of the forward relationship. The bridging device uses the affine time conversion relationship 40 in the same direction to simultaneously process the open and closed boundaries of the entry hardware timestamp 21 and the local activity window, and then compares it with the planned gating window 20 parameters saved by the era carried by the data frame. If only the entry hardware timestamp 21 is converted while the window boundary of another era is continued to be used, the conversion error will be mixed into the remaining duration of the window. When the cross-era offset includes a complete cycle, the converted time is subtracted from the target era reference time, and the absolute gate instance number is obtained by rounding down by the gating period, and the remainder is used as the phase within the cycle. The switching point itself is classified into the new era according to the inclusion end rules issued by the configuration plane.

[0034] In some embodiments, the device also calculates a phase residual. The phase residual is the shortest signed ring distance between the converted timestamp phase and the phase of the planned gating window 20, and its range is limited to half the negative gating period to half the positive gating period. The phase residual is used to diagnose the offset direction and remaining margin, while the available gate instance residual 23 is represented by an integer representing the number of absolute gate instances that need to be crossed. The two residuals have different functions; the integer residual is used for breakpoint location to avoid false breakpoints caused by the phase jumping from positive to negative values ​​at period boundaries.

[0035] Figure 3This is a schematic diagram illustrating the residual change and breakpoint location of adjacent available gate instances provided in an embodiment of the present invention. The data frame carries residual 0 when leaving the first bridging device 11 and also carries residual 0 when leaving the second bridging device 12. The third bridging device 13 calculates residual 1, and the receiving end 14 subsequently maintains residual 1. The first change in residual occurs between the second bridging device 12 and the third bridging device 13, thus the entry point of the third bridging device 13 is determined as the gate instance breakpoint 16. The gate instance breakpoint data plane 50 then writes a breakpoint marker 35, which includes at least the breakpoint location, flow identifier, synchronization era identifier, and connection request bit. The path location identifier is pre-allocated by the configuration plane according to the forwarding order of the deterministic data flow and is used to uniquely represent the current bridging device's position in the path.

[0036] In one embodiment, determining the gate instance breakpoint 16 may further require that the current data frame cannot be sent within the remaining open duration of the planned gating window 20. This additional condition excludes situations where adjacent devices experience integer residual changes due to rounding or configuration differences, but the current frame can still be safely sent within the original gating window, thereby avoiding triggering downstream connections at such locations. Even if the gate instance breakpoint 16 is determined solely by the first change in the residual of adjacent available gate instances, the current data frame is still sent once from the original gating window or conditional gating slot 54, and will not be sent repeatedly. The bridging device that determines the gate instance breakpoint 16 accordingly directly sends the current frame to the shared connection resource pool 55 for arbitration, without waiting for a second abnormal frame or for the controller to reissue the gating table. Continuous service frames are used to confirm whether the abnormality persists and do not constitute a prerequisite for processing the first abnormal frame.

[0037] Figure 5 This is a schematic diagram illustrating the triggering of shared connection resources downstream of the breakpoint provided in this embodiment of the invention. The shared connection resource pool 55 is not a window temporarily created at runtime, but rather a set of condition-gated slots 54 selected from the idle intervals of each protected port during the deployment phase. The configuration plane establishes an isolation interval for each candidate slot. The isolation interval covers the maximum frame transmission duration, the maximum non-preemptive blocking duration, and the protection interval, while also taking into account the upper bound of intra-bridge forwarding jitter and the upper bound of clock error. The upper bound of intra-bridge forwarding jitter can be taken from 0.5 microseconds to 20 microseconds, and the upper bound of clock error can be taken from 0.1 microseconds to 5 microseconds. The configuration plane verifies that the isolation interval is separated from the interval occupied by the already scheduled deterministic data flow 15 before adding the condition slot to the resource pool.

[0038] The gate instance breakpoint data plane 50 looks up the resource mapping based on the breakpoint location, downstream path location, and flow identifier, and writes the connection pool identifier to the current frame. The first downstream bridging device 51, the second downstream bridging device 52, and the third downstream bridging device 53 only enable the corresponding conditional gating slot 54 when the breakpoint marker 35, the synchronization era identifier, and the connection pool identifier all match simultaneously. The gate state of other flows remains unchanged, and the upstream of the breakpoint does not switch scheduling. This limits the use of temporary resources to the actually affected downstream segments.

[0039] The start times of a set of hop-by-hop condition-gated slots 54 are staggered according to the sum of a frame transmission duration and the intra-bridge forwarding duration, ensuring that the frame to be connected, 61, sent from the first downstream bridging device 51, can continuously reach the available slots of the second downstream bridging device 52 and the third downstream bridging device 53. The configuration plane performs a full path check before distribution. If any port cannot provide a conflict-free condition-gated slot 54 within the transition increment limit, the mapping group is marked as unavailable, and the corresponding flow directly enters the fault isolation process 74; it will not initiate only half-connection during runtime. If a device misses the minimum lead time before the first downstream condition-gated slot 54 is triggered, and subsequent candidate slots exceed the connection waiting limit, the entire group is not started and enters the fault isolation process 74. If a mid-connection condition-gated slot 54 fails after connection has started, the failed device locally stops continuing the connection for the current frame and propagates a failure flag carrying the connection pool identifier to subsequent devices; subsequent devices close the corresponding condition-gated slot 54 upon receiving the failure flag. This process does not trace completed upstream transmissions.

[0040] Figure 6 This is a schematic diagram illustrating the state transition of a single data frame transmission according to an embodiment of the present invention. The gate instance breakpoint data plane 50 first registers the detected frames as candidate abnormal frames 60. When multiple candidate abnormal frames 60 compete for the same resource, the frame to be continued 61 is selected according to the earliest to latest deadline; if the deadlines are the same, selection is based on the value of the flow identifier or a pre-configured order. Unselected frames do not occupy the conditional gating slot 54 and directly enter the fault isolation process 74. This deterministic order ensures that different bridging devices obtain consistent results for the same batch of competing frames.

[0041] After frame 61 is selected, it enters the conditional gating slot occupancy state 62. Data frames only carry connection eligibility and connection pool identifiers; a single transmission token is a hop-by-hop local state established by each bridging device using the egress port, flow identifier, intra-flow frame sequence number, and synchronization era identifier as keys, and is not consumed across devices. The original gating window and conditional gating slot 54 point to the same local queue item and contend for the local token through a comparison and swap operation. If the original gating window completes transmission first, the local port atomically consumes the token, sets the transmission completion state 63, and closes the unused conditional gating slot 54. If the conditional gating slot 54 completes transmission first, it also consumes the local token, sets the transmission completion state 63, and sets the skip state 64, preventing the next original gating window corresponding to this frame from dequeuing. When the downstream device receives this frame, it re-establishes the hop token using the same key, thus enabling an independent mutual exclusion selection.

[0042] The state 63, indicating completion of transmission, remains in this hop until a completion receipt is received or the state retention time expires, after which it transitions to the clear state 65. The state retention time can be 1 to 3 gating cycles. If the link layer reports a transmission failure and the token has not yet been consumed, it can continue waiting for the condition gating slot 54 as long as the connection waiting limit has not been exceeded; if the token has been consumed or the time limit has expired, a second transmission will not be performed, and the fault isolation process 74 will be initiated. This state machine simultaneously covers the boundary cases of original window open-back, condition slot open-first, and loss of transmission receipt.

[0043] The upper limit for the connection waiting time of a protected flow is obtained by subtracting the normal forwarding delay from its deadline, and can be set from 50 microseconds to 2 milliseconds, typically 100 microseconds. A connection waiting time upper limit that is too small will reduce the number of available condition slots, while a limit that is too large may result in the transmission being completed even though the service deadline has passed. The upper limit for the transition increment can be set from 80 microseconds to 500 microseconds, typically 100 microseconds, and should be set in the specific configuration to not exceed the connection waiting time upper limit of the corresponding protected flow. Each bridging device should be deployed using end-to-end worst-case analysis to determine the specific values, ensuring that the sum of the detection duration, the interval from detection completion to the slot start point, the gating trigger activation duration, and the transmission duration does not exceed this upper limit.

[0044] For example, the data frame length is 512 bytes, the link rate is 1 gigabit per second, and the corresponding transmission time is 4.096 microseconds. Residual detection takes no more than 2 microseconds, the interval from detection completion to the start of conditional gating slot 54 does not exceed 90 microseconds, and the total gating trigger activation time and separately included protection margin do not exceed 3 microseconds. The transition increment, conservatively added together, is 99.096 microseconds, less than the typical transition increment upper limit of 100 microseconds. Compared to... Figure 2 In the example, the waiting time for the next scheduled gating window is approximately 916 microseconds, and this condition slot limits the additional wait time for a single time synchronization transient to approximately 0.1 milliseconds.

[0045] Figure 8This is a schematic diagram illustrating breakpoint probe verification and connection cancellation provided in an embodiment of the present invention. After the first abnormal frame has achieved a transmission completion status 63, the gate instance breakpoint data plane 50 generates a breakpoint probe 70. The breakpoint probe 70 carries a flow identifier, a synchronization era identifier, a gate instance breakpoint 16, a connection pool identifier, and a source-end gate instance sequence number, and also carries a fixed-length sequence area indexed by path position. When forwarding the breakpoint probe 70, each bridging device adds the source-end gate instance sequence number to the hop-by-hop gate instance advance amount of the current hop to obtain the planned absolute gate instance sequence number of the current hop, and writes the local first available gate instance sequence number 71 and the planned absolute gate instance sequence number into the corresponding slot according to the path position of the current hop, instead of overwriting the slot of the previous device. The sequence capacity is configured according to the maximum number of hops in the path, and can be 4 to 64 items; when the capacity is exceeded, the path position is repeated, or the position is out of bounds, an abnormal flag is written. The receiving end 14 reads the complete sequence area, subtracts the corresponding planned absolute gate instance sequence number from the first available gate instance sequence number 71 of each hop, and forms the probe gate instance residual sequence 72.

[0046] When the residuals of each jump in the probe gate instance residual sequence 72 and the available gate instance residuals 23 of consecutive service frames are all 0, the receiver 14 or the controller sends a cancellation flag 73 downstream of the gate instance breakpoint 16. After receiving the cancellation flag 73, the first downstream bridging device 51, the second downstream bridging device 52, and the third downstream bridging device 53 restore the corresponding conditional gate slot 54 to the conditional closed state and clear the corresponding connection pool identifier. The recovery determination can require the residuals to be 0 for 2 to 16 consecutive frames, typically 4 frames. Too few frames are easily affected by single jitter, while too many will prolong the temporary resource occupation time.

[0047] If the probe gate instance residual sequence 72 still contains non-zero residuals, or if consecutive service frames are not recovered within the preset verification time, the corresponding flow stops using the shared connection resource pool 55 and enters the fault isolation process 74. The probe verification time can be 1 to 16 gate cycles. If the breakpoint probe 70 is lost, the sequence has missing items, the path position is duplicated, the device does not respond within the verification time, or the abnormal flag is set, the cancellation flag 73 is not sent, and the fault isolation process 74 is entered after the verification time expires. The fault isolation process 74 can redirect the affected flow to a configured redundant path, degrade it to nondeterministic transmission, or report an unrecoverable fault to the upper-layer controller. The breakpoint probe 70 is generated after the first frame is reconnected, so the verification process does not delay the partial transmission opportunity of the first abnormal frame.

[0048] During the transition period, each bridging device simultaneously stores the mapping corresponding to the old era timeline 41 and the mapping corresponding to the new era timeline 42. When a data frame carries an old era identifier, the device uses the old mapping to determine the first available gate instance number 71 and only allows the use of the conditional gating slot 54 corresponding to the old era; when a data frame carries a new era identifier, the new era mapping and the corresponding conditional slot are used. The old mapping is deleted after confirming that there are no frames in transit in the old era. Confirmation of no frames in transit can be achieved by either the source end continuously sending new era frames and exceeding the maximum path dwell time, or by the receiver 14 reporting that the last frame of the old era has arrived.

[0049] In another embodiment, multiple protected flows share a connection resource pool 55 on the same aggregation port. The configuration plane first constructs a common isolation interval based on the maximum frame size and blocking upper limit of all flows, and then records the allowed slots for each flow. During runtime, arbitration is still based on the deadline and flow identifier, and each slot is granted only one waiting frame 61 within a gating cycle. If a high-priority flow continues to occupy the slot, causing a low-priority flow to be unable to be selected within the connection waiting limit, the low-priority flow enters the fault isolation process 74, without crowding out the already verified normal deterministic scheduling interval.

[0050] In one possible implementation, the configuration plane generates a connection resource verification table on a port-by-port basis. Each entry records the port identifier and gating period, as well as the start and end phases and allowed flow set of the conditional gating slot 54, and the maximum frame length and guard interval. The configuration plane extends the occupied interval in the normal gating table by the maximum non-preemptive blocking duration, and then extends the candidate conditional slots by the upper bound of the bridge forwarding jitter and the upper bound of the clock error. The normal window, candidate slot groups within the same resource pool, and slot groups from the old and new eras participate in the interval verification. When the extended intervals intersect, the candidate slots are eliminated; if it is necessary to reuse the same physical slot and static separation is not possible, each slot group is written into the same mutual exclusion arbitration domain, and the unique selected frame is determined according to the fixed order of the deadline, flow identifier, and time synchronization era identifier. Only after complete separation or the arbitration domain rules are determined is the mapping from the breakpoint position to the hop-by-hop slot group generated.

[0051] The cutoff time for the flow is provided by the service configuration, and the normal forwarding delay can be obtained by accumulating hop by hop through the planned gating table. The normal forwarding delay includes the queuing time at the source end (10), the planned waiting time of each bridging device, in-bridge processing, link transmission time, and the processing margin at the receiver (14). The connection waiting upper limit is strictly equal to the cutoff time minus this normal forwarding delay, used to constrain the complete additional duration from the start of anomaly detection to the completion of connection transmission; the transition increment upper limit is no greater than the connection waiting upper limit. If the result is less than 0, it indicates that the flow itself has no delay margin available for connection, and the configuration plane will not establish a shared connection resource mapping for it.

[0052] In some embodiments, the gate instance breakpoint data plane 50 uses a fixed-length register to store the previous hop residual, the single-transmission token of the current frame, and the state expiration time. The register key consists of a stream identifier, an intra-stream frame sequence number, and a synchronization era identifier, which can distinguish frames with the same intra-stream frame sequence number in different eras. After a frame leaves the current hop or enters the clear state 65, the register entry is reclaimed. If the register capacity is insufficient, a new candidate abnormal frame 60 does not overwrite the already occupied entry, but instead enters the fault isolation process 74, thereby maintaining the single-transmission semantics of the selected frame to be continued 61.

[0053] Gating triggering can be implemented using conditional enable bits in the port gating table. When the conditional enable bit is 0, conditional gating slot 54, although present in the timetable, will not be open to the normal queue; when the connection pool identifier matches and the single-transmission token is valid, the data plane sets the enable bit before the slot start point. The interval between the enable bit setting and the slot start point should not be less than the minimum advance specified by the hardware. The minimum advance can be between 0.5 microseconds and 5 microseconds. If the detection is completed later than the time corresponding to the minimum advance, the conditional slot will not be triggered again in this cycle, and the availability of the next candidate slot will be determined according to the connection waiting upper limit.

[0054] As an alternative implementation, breakpoint marker 35 and connection pool identifier can be transmitted only in the internal packet metadata of the bridging device, or they can be written into the extended field of the time-sensitive network data frame. When using the extended field, the intermediate device between the source 10 and the receiver 14 reads the field according to the pre-configured offset and deletes the extended field before the receiver 14 delivers the service load. When using internal metadata, adjacent devices transmit era and residual information through reserved sideband tags. Neither implementation changes the service load, and both use flow identifiers and intra-flow frame sequence numbers to ensure that the markers correspond to specific data frames.

[0055] In other embodiments, devices using both old-era and new-era mappings may coexist in the path. The time synchronization era identifier carried by the data frame determines the interpretation caliber, and the device cannot overwrite this identifier with its local current era. If the corresponding mapping is missing locally, the device does not guess the conversion result, but directly sends the frame to the fault isolation process 74 and records the missing era. When the configuration plane issues the new-era mapping, it first confirms that the devices in the path have retained the old-era mapping, and then switches the era identifier written by the source 10 to reduce such omissions.

[0056] In cases of brief time synchronization oscillations, the available gate instance residuals 23 of consecutive service frames may vary between 0 and 1. Cancellation of the connection requires that each jump residual in the probe gate instance residual sequence 72 and the available gate instance residuals 23 of the consecutive service frames simultaneously return to zero. If any condition is not met, the condition remains closed or the condition slot is used as already selected. If the change continues beyond the transition period, it is no longer considered a time synchronization transient and the process proceeds to fault isolation procedure 74. This distinguishes between one-time era switching and continuous clock failures, preventing the shared connection resource pool 55 from being occupied for an extended period.

[0057] Optionally, when redundant paths are configured, the fault isolation process 74 switches subsequent frames to the backup path that has completed gating verification; when no backup path is configured, the abnormal flow is moved into a restricted queue and reported to the controller. The restricted queue does not occupy the planning window of other deterministic data flows 15, and its rate can be 10% to 50% of the original service rate. The fault isolation process 74 only processes flows that are not selected or whose acknowledgments have not been recovered, does not reclaim sent data frames, and does not trigger application layer retransmissions.

[0058] The first bridging device 11, the second bridging device 12, and the third bridging device 13 can use the same data plane procedure and different configuration tables. The gating reference time, hop-by-hop advance amount, and condition slot phase for each device are written separately into the configuration plane. After deployment, the actual gate opening time can be verified using the hardware timestamp readback interface. If the actual deviation exceeds the upper limit of the clock error, the corresponding slot group is marked as unavailable. This verification is part of resource pool maintenance and does not affect data frames being transmitted according to the normal gating plan.

[0059] Figure 1 This indicates that breakpoint 16, representing a gate instance, can appear between any adjacent devices. Figure 3 Provide specific examples with residuals of 0, 0, 1, and 1. Figure 3 In the three-hop example, gate instance breakpoint 16 is located at the entrance of the third bridging device 13. Therefore, only the third bridging device 13 and its subsequent paths use the conditional gating slot 54, while the first bridging device 11 and the second bridging device 12 continue to use their original plans. In another scenario, if gate instance breakpoint 16 is located at the entrance of the first bridging device 11, the configuration face selects an entire set of slots starting from the first downstream bridging device 51. The location of gate instance breakpoint 16 serves as both a diagnostic result and an index for looking up local continuity mappings, eliminating the need to broadcast a new gating table for the entire path.

[0060] Receiver 14 can check for duplicates or missing frames based on the intra-stream frame sequence number. Since the transmission path shares a single-transmission token, the same sequence number normally appears only once. If a device restart causes state loss and duplicate sequence numbers appear, receiver 14 retains the first arriving frame, discards the later duplicate frame, and reports to the controller. This receiver protection only handles extreme restart anomalies and is not a normal requirement for achieving high-reliability, low-latency transmission in this invention.

[0061] The method provided by this invention can be implemented collaboratively by the data plane pipeline, time-aware shaper, gated table memory, and configuration plane control program of a time-sensitive network switch chip. The time synchronization flag, residual, and status bits can be set in internal metadata or extended headers; fields transmitted across devices are written by the source I / O or bridging device, and internal device fields are removed at the exit point. Residual comparison, deadline sorting, and status token updates are all integer comparisons or atomic updates of table entries, which can be completed in a common programmable switch chip pipeline.

[0062] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit them. Those skilled in the art should understand that modifications can be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features, and such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A highly reliable, low-latency data transmission method, characterized in that, include: In time-sensitive networks, configure hop-by-hop gating windows and shared connection resource pools for deterministic data streams, and bind time synchronization era identifiers and the planned absolute gate instance sequence numbers of each bridging device to data frames; Each bridging device selects the gate control reference time, cycle duration, and gate control window phase according to the time synchronization era identifier, obtains the entry hardware timestamp of the current data frame, and determines the remaining open duration of the current gate instance; When the remaining open time is not less than the sum of the processing time, the transmission time and the protection margin, the current door instance is determined as the first available door instance; otherwise, the next door instance is determined as the first available door instance. The integer difference between the first available gate instance number and the corresponding planned absolute gate instance number is determined as the available gate instance residual; The location where the residual of an adjacent available gate instance first changes along the path is determined as the gate instance breakpoint, and the data surface at the gate instance breakpoint is used as the current data frame to write the breakpoint marker. Downstream of the breakpoint of the gate instance, the conditional gating slot in the shared continuation resource pool is triggered only for the deterministic data stream carrying the breakpoint marker and the corresponding time synchronization era identifier, so that the current data frame is sent once from the original gating window or the conditional gating slot; The conditional gating slot pre-completes conflict verification. The interval from the completion of residual detection to the slot start point is no greater than the continuation waiting limit. The sum of the detection duration, the interval, the gating trigger effective duration, and the transmission duration is no greater than the transition increment limit, and the transition increment limit is no greater than the continuation waiting limit.

2. The highly reliable, low-latency data transmission method according to claim 1, characterized in that, Binding the synchronization era identifier and the plan absolute gate instance number includes: generating a synchronization era identifier for each gated time mapping change; Record the gating reference time, cycle duration, gating window phase, and complete time conversion relationship with the previous time synchronization era for each time synchronization era; The source end writes the source end gate instance number corresponding to the intra-stream frame sequence number into the data frame; Each bridging device adds the source-end gate instance number to the pre-configured hop-by-hop gate instance advance amount to obtain the corresponding planned absolute gate instance number.

3. The highly reliable, low-latency data transmission method according to claim 2, characterized in that, Determining the residuals of the available gate instances also includes: saving the reference time difference, sub-periodic phase difference, frequency ratio and switching time between the old era and the new era to form an affine time conversion relationship; When the era used by the bridging device is inconsistent with the era carried by the data frame, the affine time conversion relationship is used to convert the entry hardware timestamp to the era carried by the data frame, and then the first available gate instance is determined. Retain the integer index of the first available gate instance; The shortest signed ring distance between the converted timestamp phase and the planned gating window phase is determined as the phase residual.

4. The highly reliable, low-latency data transmission method according to claim 1, characterized in that, The data plane determines the breakpoint of the gate instance by: the data frame carrying the available gate instance residual and path location identifier written by the previous bridging device; The current bridging device calculates the residuals of currently available gate instances locally; When the current available gate instance residual is not equal to the available gate instance residual written by the previous bridging device, and the current data frame cannot be sent within the remaining open time of the planned gating window, the current bridging device will determine its own entry point as the gate instance breakpoint and write the breakpoint marker. The first data frame that encounters the breakpoint of the aforementioned gate instance directly enters the arbitration of the shared continuation resource pool.

5. The highly reliable, low-latency data transmission method according to claim 1, characterized in that, The pre-configured shared connection resource pool includes: selecting a conditional gating slot during the gating period of the protected port; The maximum frame transmission duration, the maximum non-preemptive blocking duration, the upper bound of intra-bridge forwarding jitter, the upper bound of clock error, and the protection interval are incorporated into the conditional gating slot and its isolation interval. Verify that the isolation interval is separated from the occupied interval of the scheduled deterministic data stream; Write the mapping between the breakpoint location, downstream path location, and flow identifier and the same set of hop-by-hop conditional gating slots into the data plane of each bridging device.

6. The highly reliable, low-latency data transmission method according to claim 5, characterized in that, Triggering the conditional gating slot includes: the data plane at the breakpoint of the gate instance selects the frame to be continued from the candidate abnormal frames in order of their deadlines; When the deadlines are the same, the frames to be continued are selected according to the order of the stream identifiers. Write the connection pool identifier into the frame to be connected; The bridging device downstream of the gate instance breakpoint activates the corresponding condition gate control slot according to the connection pool identifier. When the same data frame is sent through the original gated window, the corresponding conditional gated slot is closed, and when the data frame is sent through the conditional gated slot, the next original gated window is skipped. Concurrent abnormal frames that have not obtained the condition gating slot enter the fault isolation process.

7. The highly reliable, low-latency data transmission method according to claim 1, characterized in that, The difference between the cutoff time of the protected flow and the normal forwarding delay is determined as the upper limit of the connection waiting time; The start times of each group of hop-by-hop conditional gating slots are staggered by the sum of one frame transmission time and the bridge forwarding time, so that data frames entering the shared connection resource pool are continuously forwarded downstream of the gate instance breakpoint after passing through the first conditional gating slot. When any port is unable to provide a conflict-checked conditional gating slot within the upper limit of the transition increment, the corresponding flow enters the fault isolation process.

8. The highly reliable, low-latency data transmission method according to claim 1, characterized in that, A breakpoint probe is generated after the first abnormal frame enters the shared continuation resource pool; The breakpoint probe carries a flow identifier, a time synchronization identifier, a gate instance breakpoint, a continuation pool identifier, and a source-end gate instance sequence number. When forwarding the breakpoint probe, each bridging device downstream of the gate instance breakpoint adds the source gate instance sequence number to the hop-by-hop gate instance advance amount of the current hop to obtain the planned absolute gate instance sequence number, and writes the first available gate instance sequence number and the planned absolute gate instance sequence number. The receiving end reads the first available gate instance number and the planned absolute gate instance number of each hop in the path order, and arranges the integer difference between the two into a probe gate instance residual sequence.

9. The highly reliable, low-latency data transmission method according to claim 8, characterized in that, When all the jump residuals in the probe gate instance residual sequence and the available gate instance residuals of consecutive service frames are 0, a cancellation flag is sent downstream of the gate instance breakpoint to restore the corresponding condition gate slot to the condition closed state. When the residual sequence of the probe gate instance still contains non-zero residuals, stop the corresponding flow from using the shared continuation resource pool and start the fault isolation process; After the first abnormal frame enters the shared continuation resource pool, the revocation or fault isolation is performed using the probe gate instance residual sequence and the continuous service frames.

10. The highly reliable, low-latency data transmission method according to claim 1, characterized in that, Each bridging device stores the old-era mapping and the new-era mapping until the corresponding transition period ends; When a data frame carries an old-generation identifier, the old-generation mapping is used to determine the first available gate instance number and allow the data frame to use the conditional gating slot corresponding to the old-generation. When a data frame carries a new era identifier, the new era mapping is used to determine the first available gate instance number and allow the data frame to use the conditional gating slot corresponding to the new era; After the transition period ends, the corresponding old era mapping is deleted once it is confirmed that there are no data frames in transit in the old era.

Images