Wide area deterministic network cross-rate forwarding method based on boundary regulation and control
By introducing a boundary control mechanism into a wide-area deterministic network, the problems of periodic residue and increased jitter in the cyclic forwarding mechanism are solved, and the end-to-end latency and jitter are reduced. It is suitable for multi-rate scenarios and compatible with existing control plane strategies.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- NANJING UNIV OF POSTS & TELECOMM
- Filing Date
- 2026-01-30
- Publication Date
- 2026-04-28
AI Technical Summary
In existing wide-area deterministic networks, the circular forwarding mechanism lacks data plane "window capacity constraints" and "cross-rate boundary suppression", which leads to problems such as cycle residue, long delay tail, and increased jitter, especially in the case of multi-source confluence, cross-rate boundary, frequency drift and mixed traffic coexistence.
By introducing boundary control mechanisms into the data plane, including the Boundary Limit Extension Strategy (BDG) and the Boundary Slice Credit Strategy (BSG), the serviceable capacity constraints are met within each port window, avoiding long window overload residue and short window micro-bursts at cross-rate boundaries, thereby reducing end-to-end latency and jitter.
It effectively reduces end-to-end latency and jitter, improves network performance stability, is suitable for multi-rate scenarios, is compatible with existing control plane policies, has low transformation costs, supports multi-rate and non-integer rate ratios, and takes into account the coexistence of mixed traffic.
Smart Images
Figure CN121940360A_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of network communication and deterministic network technology, specifically relating to a cross-rate forwarding method for wide-area deterministic networks based on boundary control. Background Technology
[0002] With the development of applications such as the Industrial Internet, collaborative manufacturing, remote control, and highly reliable real-time communication, networks no longer only focus on the average metrics of best-effort (BE) services, but also need to provide predictable latency, jitter, and packet loss behavior for critical services. Time-Sensitive Networking (TSN) achieves fine-grained time slot scheduling within a local area network through time synchronization and gated queues, but its dependence on strict phase synchronization and link latency stability makes it difficult to scale in cross-domain and long-distance link conditions. To provide deterministic service capabilities in wide-area bearer networks, the Internet Engineering Task Force (IETF) proposed the Deterministic Networking (DetNet) technical framework, which aims to achieve end-to-end deterministic guarantees for time-sensitive services based on existing bearer technologies such as Internet Protocol (IP), Multiprotocol Label Switching (MPLS), and Segment Routing (SR). Therefore, various forwarding mechanisms based on periods / epochs / circular queues have emerged, such as Cyclic Queuing and Forwarding (CQF) mechanism, Cycle-Specified Queuing and Forwarding (CSQF) and its multi-period extension schemes.
[0003] The common idea behind these mechanisms is to discretize the port service time into periods (also known as "windows") and enable time-sensitive services to perform "reception in the same period, transmission in the next period" or its variations between nodes at a fixed rhythm, thereby introducing a controllable queue waiting at each hop and forming an end-to-end analyzable delay. To adapt to the reality of "frequency synchronization only, phase inconsistency" in wide-area environments, some schemes avoid the requirement for strict phase alignment by carrying a period identifier in the header, having nodes maintain a period mapping table, or having the control plane calculate the period sequence.
[0004] However, existing solutions mostly focus on the "mapping" level: that is, answering the data packet at the 1st level. Which period, which queue, from which port, and when should a hop enter? In engineering implementation, a more fundamental but often overlooked problem is: even if the mapping is correct, if the amount of data allowed to enter the queue in a certain period exceeds the capacity that the link can send in that period, the queue cannot be drained before the end of the period, inevitably resulting in cross-period residue. After the residue accumulates over multiple hops, it will cause the end-to-end latency tail to become significantly longer, jitter to increase, and even undermine the deterministic assumption of periodic forwarding. This problem is particularly prominent in the following typical situations: (1) Multiple source confluence: Multiple ingress flows are close in phase in certain cycles, and the aggregated arrival volume exceeds the single cycle capacity in a short period of time; (2) Cross rate boundary: Multiple short cycle data of high-speed links are aggregated into a long cycle of low-speed links, which is prone to long cycle overload and cycle end residue; conversely, a long cycle release volume may be "dumped" into the first short cycle, causing sub-cycle level micro-bursts; (3) Mapping error and frequency drift: When the frequency is synchronized but the phase is inconsistent, early / late arrival, timing error and processing jitter will change the arrival distribution within the cycle, causing instantaneous overload in some cycles; (4) Coexistence of mixed traffic: When time-sensitive services and best-effort services share port resources, if there is a lack of clear cycle capacity budget and execution constraints, performance fluctuations may occur due to unclear preemption or reclamation.
[0005] Therefore, a mechanism is needed that can be implemented by adding local constraints only to the data plane: it should be compatible with existing control plane mapping strategies and should also implement requirements such as "emptying every cycle, avoiding queue residue at the end of the window, and suppressing instantaneous admission peaks in short cycles" into executable rules, thereby effectively reducing end-to-end latency and jitter in wide-area multi-rate scenarios. Summary of the Invention
[0006] To address the issues of cycle remnants, long latency tails, and amplified jitter caused by the lack of "window capacity constraints" and "cross-rate boundary suppression" in cyclic forwarding mechanisms in wide-area multi-rate deterministic networks, this application provides a cross-rate forwarding method for wide-area deterministic networks based on boundary control. This method ensures that time-sensitive services meet the serviceable capacity constraints within each port window and avoids long-window overload remnants and short-window micro-bursts at cross-rate boundaries, thereby reducing end-to-end latency and jitter.
[0007] To achieve the above objectives, this application employs the following technical solution:
[0008] This application presents a cross-rate forwarding method for wide-area deterministic networks based on boundary control, which abstracts the deterministic network topology into a directed graph. ,node For switches or routers, edge For the port, the cross-rate forwarding data plane constraint method specifically includes the following steps:
[0009] Step 1: From the network switching node The receiving end acquires the time-sensitive data packets to be forwarded from the time-sensitive service flow after scheduling at the previous hop node. At the sending port of the network switching node, based on the sending rate of the sending end of the network switching node. Generate sending cycle And according to the sending cycle Divide the timeline into consecutive service windows. Trigger window boundary events at the boundary of each service window. According to the sending rate With transmission period Calculate the available service capacity within the service window corresponding to the network switching node. and will provide service capacity As a capacity constraint for subsequent data packet admission and transmission control;
[0010] Step 2: Read time-sensitive data packets Target transmission window obtained by control plane mapping Send queue with target Get the data packet length And read the number of bytes that have been admitted to the service window. and for limiting the target sending window Credit array of internal admission data ;
[0011] Step 3: Perform the current window admission judgment: If And when sending a window to the target When the credit constraint mechanism is enabled, the time-sensitive data packets The length satisfies Then determine the time-sensitive data packet. Meet the current window admission criteria and allow time-sensitive data packets to pass. Enqueue to target send queue And update the number of bytes admitted to the window. At the same time, the corresponding credit array will be deducted when the credit constraint mechanism is activated. When the time-sensitive data packet If the current window access conditions are not met, the current window access is deemed to have failed, and the time-sensitive data packet is... Packet objects marked for cross-rate boundary processing;
[0012] Step 4: When the service window ends, trigger the window boundary event. If no cross-window incident occurs, then If a window crossover occurs, the window crossover handling strategy will be triggered. This indicates the service window number used for actual transmission after correction by the cross-window processing strategy, driving the current service window's transmission queue. Send data in the current window in a first-in-first-out order, not exceeding the service capacity. The number of bytes is calculated, and the current service window's send queue is checked when the service window ends. Whether to empty the window, obtain the window emptying verification result. If the window emptying verification result is that cross-rate boundary processing has not been triggered, proceed directly to step 6 to perform outward transmission. If the window emptying verification result triggers cross-rate boundary processing, proceed to step 5. The cross-window processing strategy includes the Bounded Dispatch and Gating (BDG) strategy and the Burst Slicing and Gating (BSG) strategy.
[0013] Step 5: For cross-window processing objects marked in Step 3 or cross-window cases where the end-of-window check failed in Step 4 and were detected as crossing rate boundaries, select a boundary processing strategy based on the relationship between the sending rates of adjacent link ports: When the high-speed link is to the low-speed link, execute the boundary limit extension strategy BDG; when the low-speed link is to the high-speed link, execute the boundary slice credit strategy BSG. The data packets processed by the boundary limit extension strategy BDG and the boundary slice credit strategy BSG are redistributed to the sending queue of the subsequent service window.
[0014] Step 6: After completing Step 4 or Step 5, send the time-sensitive data packet processed by the Boundary Limit Extension Policy (BDG) and the Boundary Slice Credit Policy (BSG) to the next hop.
[0015] A further improvement of this application is that, in step 1, the service capacity... There are two ways to calculate it:
[0016] The first method, without considering the impact of capacity reservation or protocol overhead, is based on port speed. With transmission period Calculate serviceable capacity : ;
[0017] The second approach is to adopt the method when considering protection intervals or protocol overhead. Calculate serviceable capacity ,in <1, This represents the percentage of capacity reserved for protecting time slots or protocol overhead. Setting it to 0 indicates that no extra protection time is reserved during this transmission cycle, meaning that the entire cycle capacity can be used for data transmission. This is the sending port of the network switching node.
[0018] A further improvement in this application is that the boundary limit extension strategy (BDG) is deployed on the low-speed port at the boundary between the high-speed link and the low-speed link, when time-sensitive data packets... When the signal flows from the high-speed port to the low-speed port, the low-speed port uses a long cycle. High-speed ports use short cycles Maintain two state variables on the low-speed port: the cumulative number of time-sensitive data bytes arriving within the current low-speed long service window. The remaining bytes that were not allowed to proceed in the previous low-speed long service window due to capacity constraints and need to be processed in a deferred manner. The processing procedure is as follows:
[0019] Short window arrival accumulation: Whenever a batch of time-sensitive data packets arrives from a short window on a high-speed link. Corresponding number of bytes received Perform accumulation operation ;
[0020] Long window boundary release: When the slow-speed long window ends, a window boundary event is triggered. Calculate the amount of time-sensitive data bytes allowed to pass through within the current low-speed long service window. and will Bytes are allowed to enter the send queue corresponding to the current send window of the low-speed port. Then update the state variables. Clear .
[0021] A further improvement in this application is that the Boundary Slice Credit Strategy (BSG) is deployed on a high-speed port at the low-speed to high-speed boundary, when time-sensitive data packets... When the corresponding incoming byte volume enters the high-speed port from the low-speed port, a high-speed short window period is set at the boundary between the low-speed link and the high-speed link. Short window capacity limit The Boundary Slicing Credit Strategy (BSG) maintains two state variables: and credit array The processing procedure is as follows:
[0022] Long Window Arrival and Credit Allocation: The amount of time-sensitive data bytes allowed to be released within a low-speed, long service window. Then, execute Then After slicing, the following steps will be taken. Credit allocation in a short window: for ,distribute Or adopt equal allocation ,in Represented as the first The credit limit (in bytes) allocated for each short-cycle service window. And execute. ;like non-integer or For non-integer values, a proportional accumulator is used to discretize the credit allocation process to ensure that, across multiple low-speed, long-service window scales, any continuous... The cumulative credit allocation of each short-cycle service window and its corresponding The proportional deviation does not exceed a preset threshold, while ensuring that any short-window credit array No more than ;
[0023] Short window admission: In the sending window number At the start of a short-cycle service window, time-sensitive data packets Access requires meeting two conditions simultaneously: and ;
[0024] Implementation after access If there is still unused credit when the short window ends, you can choose to clear it or accumulate it to the next short window.
[0025] A further improvement of this application is that: when the span rate ratio When the value is a non-integer, for the credit array Discretize the allocation so that any continuous The cumulative allocation of each short window and The proportional error does not exceed the preset threshold, and each short window is allocated with Not exceeding .
[0026] A further improvement in this application is the increase in the number of passengers allowed through the long window. Slice allocation uses a credit mechanism: for subsequent... A short window maintains the credit array. At the start of each short-cycle service window, the amount of data allowed for admission must meet the following requirements: And deduct credit after admission.
[0027] A further improvement in this application is that: in step 4, the window end emptying check is sent to the queue of the current service window. Check if the queue occupancy is zero, and configure the minimum buffer depth for the port queue. ,in This is a safety factor used to absorb window alignment errors and link transmission jitter.
[0028] A further improvement of this application is that the cross-rate forwarding data plane constraint method is implemented through a cross-rate forwarding data plane constraint system, which is deployed in a network device that supports cyclic queuing and forwarding, including a window timing module, a window admission and emptying verification module, a boundary limit extension module (BDG), and a boundary slice credit module (BSG).
[0029] The window timing module includes a window number maintenance unit and a timed triggering unit, used to maintain the service window number. And generate window boundary events. ;
[0030] The window-based admission and queue emptying verification module includes an admission determination unit and a queue status detection unit, used to determine the available service capacity. Number of bytes admitted to the window Make admission and rejection decisions on arriving data packets, and adjust the sending queue at the end of the window. Perform send and empty checks;
[0031] The boundary limit extension module (BDG) includes an arrival accumulation unit and an extension maintenance unit, used to maintain the accumulated arrivals within the current long-cycle service window at the high-speed to low-speed boundary. And the remaining amount carried over from the previous long-term service window. and based on and The summation calculation of the low-speed port release volume ;
[0032] The boundary slice credit module (BSG) includes a credit array maintenance unit and a credit allocation unit, used to maintain a short window for credit array maintenance at the low-speed to high-speed boundary. Credit constraints are imposed on short-window admission and short-window sending.
[0033] The window timing module, the window admission and emptying verification module, the boundary limit extension module (BDG), and the boundary slice credit module (BSG) work together as a unit. When the window admission and emptying verification module rejects window admission or detects a cross-rate boundary, it triggers the corresponding boundary processing logic to achieve constraints on the cross-rate forwarding data plane.
[0034] The beneficial effects of this application are:
[0035] This application improves determinism and stability: This application introduces window admission and empty check in the data plane, transforming the "must empty every cycle" from an abstract assumption into an executable rule, avoiding the elongation of the latency tail and amplification of jitter caused by the accumulation of multiple hops at the end of the cycle, thereby improving end-to-end performance stability.
[0036] 2. Targeted suppression across rate boundaries: For aggregation overloading in the high-speed to low-speed direction, this application sets an upper limit on the long-cycle release volume through BDG and extends the excess accordingly, ensuring that the low-speed long window can be emptied at the end of the cycle; For dumping micro-bursts in the low-speed to high-speed direction, this application slices and credits the capacity according to the short-cycle capacity through BSG, significantly reducing the instantaneous injection peak at the sub-window level and alleviating downstream queue oscillations.
[0037] 3. This application reduces end-to-end latency and suppresses latency jitter: Under the premise of satisfying deterministic forwarding rhythm, this application eliminates cross-window residue and suppresses micro-bursts, making queue waiting more uniform and congestion more controllable, thereby reducing average end-to-end latency and converging the tail of latency distribution to reduce end-to-end latency jitter.
[0038] 4. This application is compatible with existing control plane strategies and has low modification costs: This application does not require the replacement of control plane routing and window mapping such as CSQF / MCCSQF. It only requires the addition of local counters, credit arrays and gating logic to the data plane, which is convenient for incremental deployment on existing DetNet / SR bearer devices.
[0039] 5. This application supports multi-rate and non-integer rate ratios: This application allows cross-rate ratios. The value can be an integer or a non-integer. The quota is discretized and the long-term ratio is maintained through a proportional accumulator, avoiding systematic deviations and periodic jitters caused by simple rounding. It is suitable for various rate combinations in actual networks.
[0040] 6. This application takes into account the coexistence of mixed traffic: This application adopts a clear periodic capacity budget for time-sensitive services, while retaining a mechanism for BE traffic to use the remaining bandwidth, avoiding "long-term idling for determinism", improving link utilization efficiency and reducing resource waste. Attached Figure Description
[0041] Figure 1 This is the flowchart for this application.
[0042] Figure 2 This is a schematic diagram of the module structure of the cross-rate forwarding data plane constraint system of this application.
[0043] Figure 3 This is a schematic diagram of the cross-rate processing flow when a window boundary event is triggered in this application.
[0044] Figure 4 This is a schematic diagram of the boundary limit extension strategy (BDG) of this application.
[0045] Figure 5 This is a schematic diagram of the application and the Boundary Slice Credit Strategy (BSG). Detailed Implementation
[0046] The embodiments of the present invention will be disclosed below with reference to the drawings. For clarity, many practical details will be described in the following description. However, it should be understood that these practical details are not intended to limit the invention. That is, in some embodiments of the invention, these practical details are not essential.
[0047] like Figure 2 As shown, this application is a wide-area deterministic network cross-rate forwarding method based on boundary control. The method is implemented through a cross-rate forwarding data plane constraint system, which is deployed in a network device that supports cyclic queuing and forwarding. The system includes a window timing module, a window admission and emptying verification module, a boundary limit extension module (BDG), and a boundary slice credit module (BSG).
[0048] The window timing module includes a window number maintenance unit and a timed triggering unit, used to maintain the service window number. And generate window boundary events. ;
[0049] The window-based admission and queue emptying verification module includes an admission determination unit and a queue status detection unit, used to determine the available service capacity. Number of bytes admitted to the window Make admission and rejection decisions on arriving data packets, and adjust the sending queue at the end of the window. Perform send and empty checks;
[0050] The boundary limit extension module (BDG) includes an arrival accumulation unit and an extension maintenance unit, used to maintain the accumulated arrivals within the current long-cycle service window at the high-speed to low-speed boundary. And the remaining amount carried over from the previous long-term service window. and based on and The summation calculation of the low-speed port release volume ;
[0051] The boundary slice credit module (BSG) includes a credit array maintenance unit and a credit allocation unit, used to maintain a short window for credit array maintenance at the low-speed to high-speed boundary. Credit constraints are imposed on short-window admission and short-window sending.
[0052] The window timing module, the window admission and emptying verification module, the boundary limit extension module (BDG), and the boundary slice credit module (BSG) work together as a unit. When the window admission and emptying verification module rejects window admission or detects a cross-rate boundary, it triggers the corresponding boundary processing logic to achieve constraints on the cross-rate forwarding data plane.
[0053] like Figure 1As shown, this application presents a wide-area deterministic network cross-rate forwarding method based on boundary control, used for network devices supporting cyclic queuing and forwarding. It allows for forwarding of input time-sensitive data packets without altering the path selection and target port or output interface mapping given by the control plane. To process the data packets, window capacity constraints and boundary slice forwarding are performed on the data plane, and the processed packets are sent to the next hop. The deterministic network topology is abstracted as a directed graph. ,node For switches or routers, edge For links or ports, the cross-rate forwarding data plane constraint method specifically includes the following steps:
[0054] Step 1: From the network switching node The receiving end acquires the time-sensitive data packets to be forwarded from the time-sensitive service flow after scheduling at the previous hop node. At the sending port of the network switching node, based on the sending rate of the sending end of the network switching node. The transmission period is generated by a local periodic timer. And according to the sending cycle Divide the timeline into consecutive service windows. , of which The time range for each service window is At the boundary of each service window, a window boundary event is triggered by a local periodic timer. This is used to indicate the end of the current service window and the start of the next service window; based on the sending rate. With transmission period Calculate the available service capacity within the service window corresponding to the network switching node. and will provide service capacity This serves as a capacity constraint for subsequent data packet admission and transmission control.
[0055] In step 1, the available service capacity There are two ways to calculate it:
[0056] The first method, without considering the dwell time, is based on the port speed. With transmission period Calculate serviceable capacity : ;
[0057] The second approach is to adopt the method when considering protection intervals or protocol overhead. Calculate serviceable capacity ,in <1, This represents the percentage of capacity reserved for protecting time slots or protocol overhead. This percentage characterizes the unavailable bandwidth reserved during the transmission cycle to ensure time determinism. Setting it to 0 indicates that the dead zone is ignored. This is the sending port of the network switching node.
[0058] Step 2: Read time-sensitive data packets Target transmission window obtained by control plane mapping Send queue with target Get the data packet length And read the number of bytes that have been admitted to the service window. and for limiting the target sending window Credit array of internal admission data ,in, This indicates the target transmission window number obtained from the control plane mapping. Indicates the target sending window Number of bytes already admitted
[0059] Step 3: Perform the current window admission judgment: If And when sending a window to the target When the credit constraint mechanism is enabled, the time-sensitive data packets The length satisfies Then determine the time-sensitive data packet. Meet the current window admission criteria and allow time-sensitive data packets to pass. Enqueue to target send queue And update the number of bytes admitted to the window. At the same time, the corresponding credit array is deducted when the credit constraint mechanism is activated. When the time-sensitive data packet If the current window admission conditions are not met, the current window admission is deemed to have failed, and the time-sensitive data packet is... Packet objects marked for cross-rate boundary processing;
[0060] Step 4: When the service window ends, trigger the window boundary event. If no cross-window incident occurs, then If a window crossover occurs, the window crossover handling strategy will be triggered. This indicates the service window number used for actual transmission after correction by the cross-window processing strategy, driving the current service window's transmission queue. Send data in the current window in a first-in-first-out order, not exceeding the service capacity. The number of bytes is calculated, and the current service window's send queue is checked when the service window ends. Whether to empty the window is determined by obtaining the window emptying verification result. If the window emptying verification result indicates that cross-rate boundary processing has not been triggered, proceed directly to step 6 to execute outward transmission. If the window emptying verification result indicates that cross-rate boundary processing has been triggered, proceed to step 5. The cross-window processing strategies include the Boundary Limit Extension Strategy (BDG) and the Boundary Slice Credit Strategy (BSG). The BDG and BSG modules use... The state variables are logically independent and are used for state maintenance within their respective modules.
[0061] In step 4, the end-of-window check is performed by sending the queue to the current service window. Check if the queue occupancy is zero, and configure the minimum buffer depth for the port queue. ,in This is a safety factor used to absorb window alignment errors and link transmission jitter.
[0062] Step 5: For cross-window processing objects marked in Step 3, or for cross-window cases where the window end-of-window check failed in Step 4 and were detected as crossing rate boundaries, select a boundary processing strategy based on the relationship between the transmission rates of adjacent link ports: When moving from a high-speed link to a low-speed link, execute the Boundary Limit Delay Strategy (BDG), applying the limit within each low-speed long window. Released, and the excess amount will be... The format is extended to the next long window; when the link is from a low-speed link to a high-speed link, the Boundary Slice Credit Strategy (BSG) is executed, and the long window allowance slice is allocated to subsequent links. Credit in a short window And constrain the number of admissions per short window to no more than Unused credits or unreleased arrivals are carried over to the next long window; data packets (or their corresponding release budgets) processed by the Boundary Limit Delay Policy (BDG) and the Boundary Slice Credit Policy (BSG) are reallocated to the sending queue of the subsequent service window.
[0063] Step 6: After completing Step 4 or Step 5, send the time-sensitive data packets processed by the Boundary Limit Extension Policy (BDG) and the Boundary Slice Credit Policy (BSG) from the corresponding port to the next hop.
[0064] like Figure 4 As shown, the boundary limit extension strategy (BDG) is deployed on the low-speed port at the boundary between the high-speed link and the low-speed link. When time-sensitive data packets... When the signal flows from the high-speed port to the low-speed port, the low-speed port uses a long cycle. High-speed ports use short cycles Multiple short-window arrivals will aggregate within a single long window. Without restrictions, this can lead to long-window overload and end-of-cycle remnants. Two state variables are maintained on the low-speed port: the cumulative number of time-sensitive data bytes arriving within the current low-speed long service window. The remaining bytes that were not allowed to proceed in the previous low-speed long service window due to capacity constraints and need to be processed in a deferred manner. Within each low-speed long window, according to the limit Allowing access will exceed the available service capacity. The amount The format is then carried over to the next long window, and the processing procedure is as follows:
[0065] Short window arrival accumulation: Whenever a batch of time-sensitive data packets arrives from a short window on a high-speed link. Corresponding number of bytes received Perform accumulation operation ;
[0066] Long window boundary release: When the slow-speed long window ends, a window boundary event is triggered. Calculate the amount of time-sensitive data bytes allowed to pass through within the current low-speed long service window. This is used to limit the size of data entering the low-speed port transmission window, and... Bytes are allowed to enter the send queue corresponding to the current send window of the low-speed port. Then update the state variables. Clear Explanation of nature: Due to No more than A low-speed port can send a maximum of [number] messages within a long window. Bytes, therefore a long window will not necessarily be empty at the end; Represents "deferred excess", the existence of which will bring additional waiting, but the waiting is limited to a finite number of long windows to avoid infinite accumulation. (5) Non-integer Handling: When and When the value is not an integer multiple or is not strictly uniform per short window, it can be calculated When introducing a proportional accumulator, for Implement slightly fluctuating quotas (e.g., alternating allocations between adjacent long windows). This is to eliminate long-term biases and avoid cyclical accumulation.
[0067] like Figure 5 As shown, the Boundary Slice Credit Strategy (BSG) is deployed on high-speed ports at the low-speed to high-speed boundary, when time-sensitive data packets... When the corresponding incoming byte volume enters the high-speed port from the low-speed port, a high-speed short window period is set at the boundary between the low-speed link and the high-speed link. Short window capacity limit If the throughput of a long window is concentrated in the first short window, the high-speed port uses a short cycle. The Boundary Slicing Credit Strategy (BSG) maintains two state variables: and credit array The processing procedure is as follows:
[0068] Long Window Arrival and Credit Allocation: The amount of time-sensitive data bytes allowed to be released within a low-speed, long service window. Then, execute Then After slicing, the following steps will be taken. Credit allocation in a short window: for ,distribute Or adopt equal allocation and execute ,like non-integer or For non-integer values, a proportional accumulator is used to ensure accurate distribution of the proportion across multiple slow, long windows, while maintaining the credit array of any short window. No more than ;
[0069] Short window admission: In the sending window number At the start of a short-cycle service window, time-sensitive data packets Access must meet two conditions simultaneously: and ;
[0070] Implementation after access If there is still credit left unused when the short window ends, you can choose to clear it or accumulate it to the next short window. The specific strategy can be determined based on the implementation complexity and jitter control requirements.
[0071] When the span rate ratio When the value is not an integer, use a proportional accumulator or Error compensation method for credit array Discretize the allocation so that any continuous The cumulative allocation of each short window and The proportional error does not exceed the preset threshold, and each short window is allocated with Not exceeding .
[0072] Passenger volume of long windows Slice allocation uses a credit mechanism: for subsequent... A short window maintains the credit array. At the start of each short-cycle service window, the amount of data allowed for admission must meet the following requirements: Credit is deducted after admission. Through the above credit slicing and admission constraint mechanism, the amount of data released over a long period is smoothly distributed within multiple short-period windows, suppressing instantaneous bursts at the sub-window level and reducing queuing jitter in downstream links.
[0073] This application does not replace the computation of path and window mapping in the control plane. The control plane (e.g., path selection based on SR / DetNet and CSQF-like periodic sequence allocation) provides the target window number and target queue for each hop. This application enqueues targets in the data plane according to the mapping result, but adds "window capacity constraints" and "boundary gating constraints". In other words, the mapping determines "which window / queue to put in", while this application determines "the maximum number that can be placed in this window".
[0074] Within each port window, time-sensitive services can occupy a maximum of [percentage missing]. The remaining bandwidth within the window can be sent by BE traffic, or further limited by BDG / BSG. When the load of time-sensitive services is low, BE traffic can automatically fill idle space to avoid resource idleness; when the load of time-sensitive services increases, C / Q and boundary gating ensure that time-sensitive services will not cause cross-window residue due to overloading, thereby improving overall predictability.
[0075] This application can be implemented on software switches, programmable data planes (such as those based on P4 / programmable ASIC / FPGA), or traditional network processors. The core requires state variables including a window counter, an admission accumulation count, and boundary port parameters. Arrays, whose implementation overhead depends on the number of ports, windows, and... The value of is related to the engineering requirements and can usually be controlled within an acceptable range.
[0076] In contrast, various circular forwarding mechanisms have been disclosed in this field, such as CQF and its three-queue implementation, CSQF and its extensions in the DetNet / SR environment, multi-cycle queue mapping and scheduling schemes (such as multi-cycle CSQF / Multi-CQF), and epoch-based Paternoster. These schemes mainly address "how to perform cycle mapping and forwarding in multi-hop / cross-domain environments," while this invention focuses on supplementing the data plane with executable mechanisms for "window capacity constraints" and "cross-rate boundary suppression," thus complementing the aforementioned schemes.
[0077] This application defines an upper limit for the window capacity of each port with a sending cycle as the granularity. And when a data packet arrives, window-based admission is performed (the cumulative number of packets enqueued within the window does not exceed a certain limit). This application proposes two types of boundary handling for window aggregation or splitting mismatch caused by cross-rate boundaries: BDG and BSG. This application relies solely on frequency synchronization, is compatible with existing control plane schemes based on periodic / queue mapping, achieves zero window end-of-window residue, upper bounds for additional boundary waiting, and suppression of sub-window level micro-bursts, thereby reducing end-to-end latency and jitter and improving link utilization.
[0078] The above description is merely an embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principle of the present invention should be included within the scope of the claims of the present invention.
Claims
1. A method for cross-rate forwarding in wide-area deterministic networks based on boundary control, characterized in that: Abstracting the topology of wide-area deterministic networks into directed graphs. ,node For switches or routers, edge For links or ports, in network switching nodes At the sending port, input time-sensitive data packets To process the target data, window capacity constraints and cross-rate boundary adjustments are performed on the data plane. The wide-area deterministic network cross-rate forwarding method specifically includes the following steps: Step 1: From the network switching node The receiving end acquires the time-sensitive data packets to be forwarded from the time-sensitive service flow after scheduling at the previous hop node. At the sending port of the network switching node, based on the sending rate of the sending end of the network switching node. Generate sending cycle And according to the sending cycle Divide the timeline into consecutive service windows. Trigger window boundary events at the boundary of each service window. According to the sending rate With transmission period Calculate the available service capacity within the service window corresponding to the network switching node. and will provide service capacity As a capacity constraint for subsequent data packet admission and transmission control; Step 2: Read time-sensitive data packets Target transmission window obtained by control plane mapping Send queue with target Get the data packet length And read the number of bytes that have been admitted to the service window. and for limiting the target sending window Credit array of internal admission data ; Step 3: Perform the current window admission judgment: If And when sending a window to the target When the credit constraint mechanism is enabled, the time-sensitive data packets The length satisfies Then determine the time-sensitive data packet. Meet the current window admission criteria and allow time-sensitive data packets to pass. Enqueue to target send queue And update the number of bytes admitted to the window. At the same time, the corresponding credit array will be deducted when the credit constraint mechanism is activated. When the time-sensitive data packet If the current window access conditions are not met, the current window access is deemed to have failed, and the time-sensitive data packet is... Packet objects marked for cross-rate boundary processing; Step 4: When the service window ends, trigger the window boundary event. If no cross-window incident occurs, then ; If a window crossover occurs, the window crossover handling strategy will be triggered. This indicates the service window number used for actual transmission after correction by the cross-window processing strategy, driving the current service window's transmission queue. Send data in the current window in a first-in-first-out order, not exceeding the service capacity. The number of bytes is calculated, and the current service window's send queue is checked when the service window ends. Whether to empty the window, obtain the window emptying verification result. If the window emptying verification result is that cross-rate boundary processing has not been triggered, proceed directly to step 6 to execute outward transmission. If the window emptying verification result triggers cross-rate boundary processing, proceed to step 5. The cross-window processing strategy includes the Bounded Dispatch and Gating (BDG) strategy and the Burst Slicing and Gating (BSG) strategy. Step 5: For cross-window processing objects marked in Step 3, or for cross-window cases where the end-of-window check failed in Step 4 and were detected as crossing rate boundaries, select a boundary processing strategy based on the relationship between the transmission rates of adjacent link ports: When the transmission rate is from a high-speed link to a low-speed link, execute the Boundary Quota Extension Strategy (BDG); when the transmission rate is from a low-speed link to a high-speed link, execute the Boundary Slice Credit Strategy (BSG); and reallocate the data packets processed by the Boundary Quota Extension Strategy (BDG) and the Boundary Slice Credit Strategy (BSG) to the transmission queue of the subsequent service window. Step 6: After completing Step 4 or Step 5, send the time-sensitive data packet processed by the Boundary Limit Extension Policy (BDG) and the Boundary Slice Credit Policy (BSG) to the next hop.
2. The method for cross-rate forwarding in wide-area deterministic networks based on boundary control according to claim 1, characterized in that: In step 1, the available service capacity There are two ways to calculate it: The first method, without considering the impact of capacity reservation or protocol overhead, is based on port speed. With transmission period Calculate serviceable capacity : ; The second approach is to adopt the method when considering protection intervals or protocol overhead. Calculate serviceable capacity ,in <1, This represents the percentage of capacity reserved for protecting time slots or protocol overhead. Setting it to 0 indicates that the dead zone is ignored. This is the sending port of the network switching node.
3. The method for cross-rate forwarding in wide-area deterministic networks based on boundary control according to claim 1, characterized in that: The boundary limit extension strategy (BDG) is deployed on the low-speed port at the boundary between the high-speed link and the low-speed link, when time-sensitive data packets... When the signal flows from the high-speed port to the low-speed port, the low-speed port uses a long cycle. High-speed ports use short cycles Maintain two state variables on the low-speed port: the cumulative number of time-sensitive data bytes arriving within the current low-speed long service window. The remaining bytes that were not allowed to proceed in the previous low-speed long service window due to capacity constraints and need to be processed in a deferred manner. The processing procedure is as follows: Short window arrival accumulation: Whenever a batch of time-sensitive data packets arrives from a short window on a high-speed link. Corresponding number of bytes received Perform accumulation operation ; Long window boundary release: When the slow-speed long window ends, a window boundary event is triggered. Calculate the amount of time-sensitive data bytes allowed to pass through within the current low-speed long service window. and will Bytes are allowed to enter the send queue corresponding to the current send window of the low-speed port. Then update the state variables. Clear .
4. The method for cross-rate forwarding in wide-area deterministic networks based on boundary control according to claim 1, characterized in that: Boundary Slicing Credit Strategy (BSG) is deployed on high-speed ports at the low-speed to high-speed boundary, when time-sensitive data packets... When the corresponding incoming byte volume enters the high-speed port from the low-speed port, a high-speed short window period is set at the boundary between the low-speed link and the high-speed link. Short window capacity limit The Boundary Slicing Credit Strategy (BSG) maintains two state variables: and credit array ,in The table in the Boundary Slice Credit Strategy (BSG) represents the long-term cumulative release volume to be sliced. Service window The corresponding credit limit (in bytes). The processing procedure is as follows: Long Window Arrival and Credit Allocation: The amount of time-sensitive data bytes allowed to be released within a low-speed, long service window. Then, execute Then After slicing, the following steps will be taken. Credit allocation in a short window: for ,distribute Or adopt equal allocation ,in This represents the ratio of the long period of the low-speed port to the short period of the high-speed port. Indicates the current short-cycle service window The A follow-up short-cycle service window Indicates assignment to the first A short window of credit limit, and execution ,like non-integer or For non-integer values, a proportional accumulator is used to ensure accurate distribution of the proportion across multiple slow, long windows, while maintaining the credit array of any short window. No more than ; Short window admission: In the sending window number At the start of a short-cycle service window, time-sensitive data packets Access must meet two conditions simultaneously: and ; Implementation after access If there is still unused credit when the short window ends, you can choose to clear it or accumulate it to the next short window.
5. The method for cross-rate forwarding in wide-area deterministic networks based on boundary control according to claim 3 or 4, characterized in that: When the span rate ratio When the value is a non-integer, for the credit array Discretize the allocation so that any continuous The cumulative allocation of each short window and The proportional error does not exceed the preset threshold, and each short window is allocated with Not exceeding .
6. The method for cross-rate forwarding in wide-area deterministic networks based on boundary control according to claim 5, characterized in that: Passenger volume of long windows Slice allocation uses a credit mechanism: for subsequent... A short window maintains the credit array. At the start of each short-cycle service window, the amount of data allowed for admission must meet the following requirements: And deduct credit after admission.
7. The method for cross-rate forwarding in wide-area deterministic networks based on boundary control according to claim 1, characterized in that: In step 4, the end-of-window check is performed by sending the queue to the current service window. Check if the queue occupancy is zero, and configure the minimum buffer depth for the port queue. ,in This is a safety factor used to absorb window alignment errors and link transmission jitter.
8. The method for cross-rate forwarding in wide-area deterministic networks based on boundary control according to claim 1, characterized in that: The cross-rate forwarding data plane constraint method is implemented through a cross-rate forwarding data plane constraint system, which is deployed in network devices that support cyclic queuing and forwarding. The system includes a window timing module, a window admission and emptying verification module, a boundary limit extension module (BDG), and a boundary slice credit module (BSG). The window timing module includes a window number maintenance unit and a timed triggering unit, used to maintain the service window number. And generate window boundary events. ; The window-based admission and queue emptying verification module includes an admission determination unit and a queue status detection unit, used to determine the available service capacity. Number of bytes admitted to the window Make admission and rejection decisions on arriving data packets, and adjust the sending queue at the end of the window. Perform send and empty checks; The boundary limit extension module (BDG) includes an arrival accumulation unit and an extension maintenance unit, used to maintain the accumulated arrivals within the current long-cycle service window at the high-speed to low-speed boundary. And the remaining amount carried over from the previous long-term service window. and based on and The summation calculation of the low-speed port release volume ; The Boundary Slice Credit Module (BSG) includes a credit array maintenance unit and a credit allocation unit, used to maintain the credit array corresponding to short-cycle service windows at the low-speed to high-speed boundary. Credit constraints are imposed on short-window admission and short-window sending. The window timing module, the window admission and emptying verification module, the boundary limit extension module (BDG), and the boundary slice credit module (BSG) work together as a unit. When the window admission and emptying verification module rejects window admission or detects a cross-rate boundary, it triggers the corresponding boundary processing logic to achieve constraints on the cross-rate forwarding data plane.
9. A network device, comprising a processor, a memory, and a plurality of ports, characterized in that, The memory stores instructions that, when executed by the processor, cause the network device to perform the method described in any one of claims 1-7.
10. A computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the method of any one of claims 1-7.