SDN-based priority scheduling method for multi-source data transmission in aviation

Abstract

This invention relates to the field of aviation data transmission and network scheduling technology, and discloses a priority scheduling method for multi-source aviation data transmission based on SDN. The method includes: an SDN controller defining a field mask that excludes timestamp and sequence number fields, and distributing the field mask and scheduling rules to a network switch; the network switch applying the field mask to calculate hash values ​​only for predetermined key data fields within the data packet, and comparing them with a baseline hash value; triggering high-priority or low-priority transmission paths based on the comparison result. This invention achieves data semantic recognition at the transmission scheduling mechanism level, can penetrate changes in non-key fields, identify whether key parameters have changed, and allocates high-priority paths only for transmitting semantically novel data, solving the problem of high-priority channels being ineffectively occupied by low-value redundant data.

Description

Technical Field

[0001] This invention relates to a priority scheduling method for multi-source aviation data transmission based on SDN, belonging to the field of aviation data transmission and network scheduling technology. Background Technology

[0002] In current digital information transmission, especially in aviation multi-source data networks, systems generally rely on a source-identity-based Quality of Service (QoS) assignment method. Data streams from the Flight Data Recorder (FDR) are given a fixed high priority to utilize low-latency transmission channels, while other data, such as Flight Quality Monitoring (QAR) or cabin maintenance data, are allocated to lower-volume transmission channels. This static tag-based scheduling method is a conventional means in the field to ensure the priority of critical services. However, those skilled in the art have found in engineering practice that the operating logic of this static assignment method contains a long-standing fundamental contradiction: a mismatch between priority and the information value carried by the data packets. During most of the smooth cruise phase of a flight mission, the FDR data streams given the highest priority are essentially highly redundant, containing continuous flight attitude or altitude parameter packets with the same content. Existing transmission scheduling mechanisms cannot detect this redundancy in data content. The mechanism can only continuously call upon the most valuable low-latency channel resources based on its static priority to transmit these redundant data packets that do not contain new information.

[0003] This behavior not only results in the ineffective occupation of high-priority channels but also crowds out overall network bandwidth that could be used for other services, leading to low overall throughput and efficiency of the transmission system. More importantly, the inherent limitation of this mechanism is that it consumes valuable low-latency guarantees reserved for sudden and novel events on the continuous confirmation of stable and redundant states. When a truly high-value sudden change signal, FDR data, undergoes a drastic change and needs to be transmitted, its timeliness has already been degraded by a large amount of redundant data on the same channel. During the review, it was found that the existing technology has the following main limitations in this regard: 1. The scheduling mechanism lacks the ability to perceive data content; its scheduling decisions are based solely on the source label of the data packet and cannot distinguish between redundant and novel information; 2. High-priority channels are continuously occupied by high-priority, low-value redundant data streams, resulting in low efficiency in bandwidth resource allocation at the transmission layer; 3. The approach in this field of introducing complex algorithms into the control plane or application layer of software-defined networking (SDN) to identify data value is inconsistent with the data plane itself. The core transmission requirements of low latency and high speed are not met, and this problem cannot be solved without sacrificing performance. This approach, which relies on complex calculations in the control plane, has been reflected in existing technologies. For example, Chinese invention patent CN117880256B discloses a method for ensuring QoS of video streams in data center networks based on multi-controller SDN. This method attempts to prioritize video streams through an M / M / N+1 queuing model and uses an RNN-improved DDPG deep reinforcement learning algorithm to iteratively train and plan the optimal path. However, the decision logic of this method is entirely built on the control plane composed of the main controller and sub-controllers. Its essence is still based on the source attributes of the data stream, i.e., identifying it as a video stream and QoS requirements, to allocate network resources. It fails to sink down to the data plane to achieve real-time identification of the data packet content itself. Similarly, it cannot solve the core problem faced by this invention, namely, how to identify and filter out those high-priority, low-value redundant data packets to prevent them from ineffectively occupying high-speed channels.

[0004] Therefore, the technical problem to be solved by this invention is how to construct a method that can identify the redundancy and novelty of data content in real time in a low-overhead manner in the network data plane, and dynamically adjust its transmission priority accordingly, so as to release high-priority channel resources from redundant data and leave them for high-value novel data. Summary of the Invention

[0005] This invention provides a priority scheduling method for multi-source data transmission in aviation based on SDN. Its main purpose is to solve the problem in the prior art that the redundancy and novelty of data content cannot be identified in real time on the data plane, resulting in high-priority transmission channels being ineffectively occupied by high-priority, low-value redundant data.

[0006] To achieve the above objectives, the present invention provides an SDN-based priority scheduling method for aviation multi-source data transmission, comprising the following steps:

[0007] An SDN controller predefines a field mask for identifying predetermined key data fields within the data packets of the target data stream, based on the data format characteristics of the target data stream. The predetermined key data fields exclude timestamp and sequence number fields.

[0008] The SDN controller sends the field mask, the first scheduling rule for high-priority transmission paths, and the second scheduling rule for low-priority transmission paths to a network switch.

[0009] The network switch applies a field mask in the data plane to extract only predetermined key data fields from a base packet of the target data stream, and calculates a base hash value only for the extracted predetermined key data fields, and caches the base hash value;

[0010] When a network switch receives a new data packet from a target data stream, it applies a field mask to extract only predetermined key data fields from the new data packet and calculates a new hash value only for the extracted predetermined key data fields.

[0011] The new hash value is compared with the base hash value. If the new hash value is inconsistent with the base hash value, the first scheduling rule is triggered, the new data packet is scheduled to a high-priority transmission path, and the base hash value is updated to the new hash value. If the new hash value is consistent with the base hash value, the second scheduling rule is triggered, the new data packet is scheduled to a low-priority transmission path.

[0012] Preferably, the target data stream is the data stream from the aircraft flight data recorder; the high-priority transmission path is the low-latency transmission path; and the low-priority transmission path is the batch compression transmission path.

[0013] Preferably, the field mask is a list of byte offsets that identify the predetermined key data fields within the data packets of the target data stream; the network switch performs the calculation of the base hash value and the new hash value in the data plane, and performs the steps of comparing the new hash value with the base hash value.

[0014] Preferably, if the new hash value is inconsistent with the base hash value, the first scheduling rule is triggered, including: setting a metadata tag for the new data packet, and the network switch matching the new data packet with the first scheduling rule based on the metadata tag.

[0015] Preferably, the method further includes: performing admission control on the new data packet before scheduling it to a high-priority transmission path; if the admission control determines that the new data packet is allowed, scheduling it to a high-priority transmission path; if the admission control determines that the new data packet is not allowed, scheduling it to a predefined second-highest-priority transmission path.

[0016] Preferably, admission control is implemented based on a token bucket mechanism; the SDN controller configures the token bucket mechanism for high-priority transmission paths, and the token bucket mechanism defines an average admission rate and a burst admission size; if admission control determines that it is allowed, it corresponds to the existence of available tokens in the token bucket mechanism, and a new data packet consumes one available token; if admission control determines that it is not allowed, it corresponds to the existence of no available tokens in the token bucket mechanism.

[0017] Preferably, the method further includes: maintaining a redundancy counter associated with the target data stream; clearing the redundancy counter to zero when performing the step of updating the baseline hash value to the new hash value; incrementing the redundancy counter by one before triggering the second scheduling rule and determining whether it has reached a preset threshold N; if the redundancy counter has not reached the preset threshold N, then triggering the second scheduling rule; if the redundancy counter has reached the preset threshold N, then forcibly triggering the first scheduling rule, scheduling the new data packet to the high-priority transmission path, updating the baseline hash value to the new hash value, and simultaneously clearing the redundancy counter to zero.

[0018] Preferably, the preset threshold N is calculated as follows: ,in, A preset heartbeat time period used to confirm the activity of the target data stream; The average time interval between data packets sent in the target data stream.

[0019] Preferably, the method further includes: at the network switch, monitoring the frequency of triggering the second scheduling rule within a preset time period to obtain a redundancy frequency index; and reporting the redundancy frequency index to the SDN controller.

[0020] Preferably, the method further includes: the SDN controller determining the stability of the target data flow based on the redundancy frequency index; when the SDN controller determines that the stability of the target data flow has decreased, the SDN controller performs predefined flow control operations on at least non-target data flows in the network.

[0021] Compared with the prior art, the beneficial effects of the present invention are:

[0022] 1. This invention achieves semantic recognition of data by having the SDN controller predefine predetermined key data fields within data packets and by having the hash calculation of the data plane (including the base hash value and the new hash value) be performed only on these selected key fields. This mechanism enables hash comparison to penetrate the continuous changes in non-key fields such as timestamps or sequence numbers in data packets and identify whether key flight parameters have actually changed. As a result, this method can truly and exclusively use high-priority transmission paths to transmit semantically novel data packets, while correctly diverting data packets with only timestamp changes but semantic redundancy to low-priority transmission paths such as the P4 queue. This solves the inherent problem of mismatch between priority and information value in existing static scheduling methods.

[0023] 2. This invention further constructs a complete dynamic transmission path assignment logic covering the entire lifecycle of the data stream. This logic uses hash value comparison as the basic identification mechanism and configures corresponding collaborative processing mechanisms for two extreme working conditions that will inevitably occur during its operation: When a large number of novel data hash values ​​are inconsistent in a short period of time and may overwhelm high-priority paths, an admission control mechanism is activated. Instead of scheduling all novel data packets to high-priority paths, the overflow portion is diverted to a predefined second-highest priority transmission path, thereby ensuring the absolute availability of high-priority paths; When redundant data hash values ​​are consistently generated, such as when a sensor malfunctions, a redundancy counter starts counting, and when it reaches a preset threshold, it forcibly schedules this redundant data packet to a high-priority path, thereby providing a low-overhead proof of liveness at the transport layer. The combination of these two collaborative mechanisms enables this method to have adaptive effectiveness in the face of two opposite and non-obvious transport layer challenges: legitimate congestion and silent failure.

[0024] 3. This invention transforms the action of redundant packet scheduling performed by the data plane into a quantifiable and novel source of network situational awareness information. By monitoring the frequency of this redundancy event at the network switch, the system obtains a real-time indicator characterizing the stability of the data flow. This redundancy frequency indicator is reported to the SDN controller, enabling the control plane to obtain predictive instability alarm capabilities derived from the data content level for the first time. Based on this indicator, when the stability of multiple critical data flows decreases simultaneously, the controller can proactively perform flow control on other non-target data flows such as non-critical services in the network before actual congestion occurs on high-priority paths. This elevates the scheduling mechanism from a passive response to a predictive management level. Attached Figure Description

[0025] Figure 1 This is a diagram illustrating the system architecture and data plane scheduling interaction of the present invention.

[0026] Figure 2 This is a diagram illustrating the mechanism of the four core functions of dynamic scheduling in this invention.

[0027] Figure 3 This is a flowchart illustrating the identification and scheduling of semantically novel data packets according to the present invention. Detailed Implementation

[0028] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be described in detail below. Obviously, the described embodiments are only some embodiments of this invention, not all embodiments. 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.

[0029] This invention discloses an SDN-based priority scheduling method for multi-source aviation data transmission. Its operational architecture includes an SDN controller acting as the control plane and one or more network switches acting as the data plane. The SDN controller defines data flow identification policies and scheduling rules, and distributes these to the network switches. The network switches then execute the rules in the data plane, identifying and assigning forwarding paths to real-time data flows. The execution of this method begins in the control plane, where the SDN controller predefines policies for a target data flow. In aviation data transmission scenarios, a target data flow can originate from aircraft flight data recordings. The data packets contain both time-varying timestamp and sequence number fields, as well as predetermined key data fields characterizing the flight status, such as flight altitude, speed, and attitude. To achieve semantic recognition of the data content, the primary task of the SDN controller is to define a field mask. This field mask identifies the predetermined key data fields within the data packet and explicitly excludes non-key fields such as timestamps and sequence numbers. In a specific engineering implementation, this field mask can be defined as a list of byte offsets. Taking a 64-byte data packet as an example, if bytes 0 to 7 contain the timestamp and sequence number, bytes 8 to 1... If 1 byte represents the height and 12 to 15 bytes represent the speed, then the field mask generated by the controller is the set of byte offsets [8, 9, 10, 11, 12, 13, 14, 15] within the data packet. To ensure that the definition of the field mask accurately reflects the format characteristics of a specific data stream, the following calibration procedure can be performed during the pre-configuration phase of system deployment: On the data plane of the network switch or a dedicated analysis node, capture data packets of the target data stream, obtaining sample sets for both stable states (such as the cruise phase) and dynamic states (such as the maneuver phase), capturing 10,000 consecutive data packets for each state; secondly, for stable states... The system performs entropy calculations on the byte-by-byte offset changes in the stationary state sample set. Byte offsets with entropy values ​​higher than a preset high-order threshold, such as the 95th percentile of the entropy distribution, are initially classified as non-critical data fields. Next, by comparing the stationary state sample set with the dynamic state sample set, byte offsets that only change in the dynamic state sample set are identified and initially classified as predetermined critical data fields. Finally, the system verifies and corrects the two initial classification results based on the known data format definition document of the target data stream, forming the final list of byte offsets used to identify predetermined critical data fields, which is then distributed by the SDN controller.

[0030] After defining the field mask, the SDN controller distributes the field mask, the first scheduling rule for high-priority transmission paths, and the second scheduling rule for low-priority transmission paths as a unified rule set to the network switch responsible for processing the target data flow. High-priority transmission paths correspond to low-latency transmission paths in the network, ensuring data timeliness, while low-priority transmission paths can correspond to batch compression transmission paths, improving bandwidth utilization efficiency. The network switch receives and loads these rules in the data plane, entering an initialization state. When the first data packet of the target data flow, i.e., the baseline data packet, arrives at the switch, the switch's data plane processing logic is activated. Using the field mask distributed by the SDN controller, it extracts only the data packet containing the mask from the baseline data packet. The specified key data field is used to continue the previous example, that is, only the data from the 8th to the 15th byte is extracted. Then, the switch performs a hash operation only on this part of the extracted key data field. An algorithm such as CRC32 can be used to calculate a base hash value, and the base hash value is cached in the switch's local high-speed memory. At this point, the scheduling base of the data plane is established. When the base hash value and the new hash value are calculated in the data plane of the network switch, the choice of hash algorithm needs to weigh the computational overhead and the probability of collision. Taking the CRC32 algorithm as an example, it is chosen because the algorithm usually has efficient hardware circuit implementation or native instruction support in modern programmable switching chips, so its calculation latency is negligible and meets the line-speed processing requirements of the data plane.

[0031] Meanwhile, for a specific input set consisting of predetermined key data fields such as flight altitude and speed, the hash collision probability generated by the CRC32 algorithm is assessed to be lower than the tolerance threshold of this scheduling method for occasional semantic misjudgments, i.e., misjudging novel data as redundant data. In other embodiments, if the switch data plane hardware has the corresponding capability, hardware-optimized versions of other algorithms such as MurmurHash or XXHash can also be used to obtain a lower collision probability under specific data distributions. When subsequent new data packets of the target data stream arrive at the network switch, the real-time scheduling mechanism starts running. The switch repeatedly performs the initial extraction and calculation actions at high speed in the data plane, that is, using the same field mask to extract only the same set of predetermined key data fields from the new data packets and using the same hash algorithm to calculate a new hash value. This calculation process and subsequent comparison steps are all performed in the data plane of the network switch to ensure line-speed processing capability. The scheduling decision involves comparing the new hash value with the cached baseline hash value. If the new hash value matches the baseline hash value, it indicates that the critical flight parameters of the new data packet have not changed, meaning it is semantically redundant data. In this case, the switch triggers the second scheduling rule, scheduling the new data packet to a low-priority transmission path. If the new hash value does not match the baseline hash value, it indicates that the critical flight parameters have changed, meaning it is semantically novel data. In this case, the switch triggers the first scheduling rule, scheduling the new data packet to a high-priority transmission path. The switch must immediately perform an update operation, replacing the cached baseline hash value with the newly calculated hash value, making it the baseline for the next comparison. In one implementation, the switch triggers the first scheduling rule by setting a specific metadata tag for the new data packet. The switch's internal forwarding logic is pre-configured by the controller to match the data packet carrying this metadata tag with the first scheduling rule, thereby completing the assignment of the high-priority path.

[0032] To further improve the system's operational stability under specific conditions, this method may also include two collaborative processing mechanisms: First, admission control for excessively fast novel data flow. When changes in flight status lead to the continuous generation of a large number of semantically novel data packets, to prevent congestion on high-priority transmission paths, this method adds an admission control step before scheduling new data packets to high-priority paths. This admission control can be implemented based on a token bucket mechanism, where the SDN controller configures an average admission rate and a burst admission size for high-priority paths. When a novel data packet (with inconsistent hash values) arrives, if there is an available token in the token bucket, it is deemed allowed, one token is consumed, and the data packet enters the high-priority path. Priority path; if there are no available tokens in the token bucket, it is determined that the request is not allowed. In this case, the new data packet is not discarded, but is diverted to a predefined second-highest priority transmission path to ensure the availability of the high-priority path. Secondly, for active detection of silent failure, in order to distinguish between benign redundancy of normal sensor operation and malignant redundancy of sensor jamming, this method maintains a redundancy counter for the target data stream at the switch. When a novel data packet (with inconsistent hash value) is scheduled to a high-priority path, the redundancy counter is cleared to zero. When a redundant data packet (with consistent hash value) arrives, the switch increments the redundancy counter by one before triggering the second scheduling rule and determines whether it has reached a preset threshold. If the threshold is not reached If the threshold is reached, the second scheduling rule will be triggered to enter the low-priority path; If the system forcibly triggers the first scheduling rule, it will schedule this redundant data packet to a high-priority transmission path, update the base hash value, and reset the redundancy counter to zero. This mechanism provides a low-overhead transport layer heartbeat signal, with a preset threshold. The determination can be based on a heartbeat time period that ensures the activity of the data stream. Average transmission time interval of data packets The calculation method is as follows: In the formula, This is a preset heartbeat time period used to confirm the activity of the target data stream. The average time interval between data packets in the target data stream; in a... Taking a heartbeat period of seconds as an example, if the average data packet sending interval is... milliseconds, then Finally, this method transforms the scheduling behavior of the data plane into a network situational awareness information source. While performing scheduling, the network switch is also responsible for monitoring the frequency of triggering the second scheduling rule within a preset time period, thereby obtaining a redundancy frequency index. This index is periodically reported to the SDN controller, which can determine the stability of the target data flow based on the redundancy frequency index. When the controller observes a decrease in the stability of the critical data flow, i.e., a decrease in the redundancy frequency index, it can perform proactive flow control operations on other non-target data flows in the network, such as non-critical business data flows, before actual congestion occurs on high-priority paths.

[0033] Example 1: This example describes a scenario where an aircraft is in a stable cruise phase of 2 hours. The onboard data network simultaneously carries two types of data streams: a target data stream from the flight data recorder (FDR) and a non-target data stream from flight quality monitoring. Following conventional static priority assignment, the FDR data stream is assigned high priority and uses a high-priority transmission path, while the flight quality monitoring data stream is assigned low priority and uses a low-priority transmission path for batch compression transmission. During this cruise phase, the aircraft's altitude, attitude, speed, and other critical flight parameters remain constant for an extended period. This results in over 99% of the data packets in the FDR data stream having identical content in their predetermined key data fields, but their timestamp and sequence number fields differ in each packet. Similarly, in the baseline scenario using conventional static priority scheduling, the network switch cannot recognize the semantic redundancy of the flight data logger packet content. The switch processes the entire packet, and due to the continuous changes in the timestamp field, each flight data logger packet is identified as novel data, resulting in all flight data logger packets being continuously scheduled to high-priority transmission paths. As a result, high-priority transmission paths are constantly occupied by these semantically redundant data streams, crowding out the total network bandwidth. Consequently, when a 5MB flight quality monitoring packet containing minor vibration anomalies needs to be transmitted, the low-priority transmission path where it resides becomes congested because bandwidth resources are ineffectively occupied by high-priority transmission paths, and the transmission task remains in a waiting state for a long time.

[0034] When deploying the scheduling method of this invention, the SDN controller has pre-issued a field mask, a first scheduling rule, and a second scheduling rule for the target data stream of the flight data recorder. The field mask is defined to cover only predetermined key data fields within the flight data recorder data packets, such as flight altitude, speed, and attitude parameters, while explicitly excluding timestamp and sequence number fields. When the cruise phase begins, the first flight data recorder data packet arrives at the network switch. The switch uses the field mask to extract the key fields, calculates and caches its base hash value, and schedules it to a high-priority transmission path according to the first scheduling rule. When the second flight data recorder data packet arrives, its timestamp has been updated, but the key fields remain unchanged. The switch uses the same field mask to extract the key fields, and the calculated new hash value matches the cached base hash value. At this point, the switch triggers the second scheduling rule, sending this flight data recorder data packet to the next high-priority transmission path. Flight data recorder packets are scheduled to low-priority transmission paths. Under this mechanism, all subsequent semantically redundant flight data recorder packets during the cruise phase are automatically diverted to low-priority transmission paths. Only when flight parameters actually change will new packets be scheduled to high-priority transmission paths due to inconsistent new hash values. This dynamic scheduling based on semantic novelty judgment reduces the occupancy rate of high-priority transmission paths from nearly 100% to nearly 0%, effectively releasing the high-priority transmission paths. The network bandwidth that was previously ineffectively occupied by high-priority transmission paths can now be used for transmission on low-priority transmission paths. The 5MB flight quality monitoring packets that were previously congested and waiting on low-priority transmission paths can now enter the batch compression transmission queue together with the semantically redundant flight data recorder packets that have been diverted to low-priority transmission paths, thereby completing the transmission quickly and improving the overall bandwidth utilization.

[0035] Example 2: This example is used to quantitatively verify the effectiveness of the scheduling method disclosed in this invention in terms of bandwidth resource allocation and transmission timeliness in a controlled network test environment. The test platform consists of an SDN controller, a network switch with data plane masking application and hash calculation capabilities, and a data stream generator. The platform simulates the aviation multi-source data transmission scenario described in Example 1. The network switch manages three transmission paths with different service qualities: a 10Mbps high-priority transmission path (P0), a 20Mbps second-highest-priority transmission path (P1), and a 50Mbps low-priority transmission path (P4). Two data streams were set up in the experiment: a target data stream, simulating a flight data recorder, with a constant rate of 12Mbps; and a non-target data stream, simulating flight quality monitoring, which is a 100MByte burst file. For comparison, a control group and the present invention were set up in the experiment. The sample group and the control group used static priority scheduling, that is, all target data streams of 12Mbps were scheduled to the P0 path and non-target data streams were scheduled to the P4 path; the sample group of this invention used the dynamic scheduling method as described in the specific implementation. The SDN controller issued a field mask, and the network switch performed semantic recognition based on hash comparison, scheduling semantically novel data to P0 and semantically redundant data to P4. At the same time, a token bucket admission control with an average admission rate of 8Mbps was set for the P0 path. Novel data exceeding the rate was scheduled to the P1 path. The experiment was divided into two phases: Phase 1, simulated cruise, lasting 60 seconds, the target data stream was set to 95% semantic redundancy, and non-target data streams started transmitting at the 5th second; Phase 2, simulated maneuver, lasting 10 seconds, the target data stream was set to 100% semantic novelty. During the experiment, the key performance indicators of each path were monitored, and the results are shown in Table 1.

[0036] Table 1: Comparison of Experimental Data

[0037]

[0038] Analysis of the data in Table 1 shows that in Phase 1, the control group sent all 12Mbps target data streams (including 11.4Mbps of redundant data) to the 10Mbps P0 path, resulting in the P0 path being continuously 100% congested and causing a critical data packet loss rate of 16.5%. The present invention's sample, through semantic recognition, only sent 0.6Mbps of novel data to P0, reducing its occupancy rate to 6.1% and the critical data packet loss rate to 0%. The 11.4Mbps of redundant data was diverted to the P4 path, sharing the bulk transmission channel with the non-target data stream. The transmission time of the data stream was 20.8 seconds, but the availability of the P0 path was guaranteed. In Phase 2, when all 12Mbps target data streams were novel data, the P0 path of the control group was continuously overloaded, resulting in the loss of 16.8% of critical data during transmission. The admission control mechanism of the sample group of this invention took effect, splitting the 12Mbps novel data stream into approximately 8.0Mbps (entering P0) and approximately 4.0Mbps (entering P1). The occupancy rates of both paths were within a reasonable range, the packet loss rate of critical data was 0%, and all novel data was transmitted completely through high-priority or second-highest-priority paths.

[0039] Example 3: This example is used to compare and verify the core role of field masks in achieving semantic recognition. The experiment uses the same test platform, transmission path (P0, P1, P4), data stream (12Mbps target data stream and 100MByte non-target data stream), and test phase (Phase 1 cruise, Phase 2 maneuver) as Example 2. A maskless comparison group is set up. The only difference between this group and the sample group of the present invention in Example 2 is that the SDN controller does not issue field masks. When the network switch performs hash calculation, it calculates based on the complete content of the data packet, including the changed timestamp field, but retains the same 8Mbps token bucket admission control mechanism as the sample group of the present invention in Example 2. In Phase 1 (cruise, 95% semantic redundancy), although most of the predetermined key data fields of the target data stream have not changed, since the timestamp field of each data packet is different, the network switch in the maskless comparison group determines that each new hash value is inconsistent with the base hash value after calculating the complete data packet.

[0040] The system identifies all 12Mbps target data streams, including 11.4Mbps of redundant data, as semantically novel data and sends them to the admission control module. This module splits the traffic, scheduling approximately 8.0Mbps to the P0 path and approximately 4.0Mbps to the P1 path. The P0 path occupancy rate reaches approximately 80.3%, and the P1 path occupancy rate reaches approximately 19.9%. In Phase 2 (maneuver, 100% semantic novelty), the system performs the same actions as in Phase 1, allocating the 12Mbps novel data stream to the P0 and P1 paths. Experimental data shows that, in the absence of the key technical feature of field masking, the mechanism of hashing the complete data packet cannot penetrate changes in non-critical fields and cannot identify semantically redundant data in the cruise phase. Its resource consumption in Phase 1 (cruise) is exactly the same as in Phase 2 (maneuver), resulting in the high-priority (P0) and second-highest-priority (P1) paths being ineffectively occupied by redundant data in the cruise phase.

[0041] Example 4: This example combines Figures 1 to 3 This section describes a priority scheduling method for SDN-based multi-source data transmission in aviation, such as... Figure 1 As shown, its core is an aviation data priority scheduling process P1 that executes in the network switch data plane. This process receives new data packets from the target data stream E2 of the flight data recorder data source. The E1SDN controller reads the field mask and scheduling rules from the D1 policy library and sends them to P1. At the same time, it configures tokens in the D3 admission control token bucket. When the P1 process runs, it applies the policies issued by E1, reads and writes the D2 baseline hash value cache for hash value comparison, and reads, writes, or clears the count value of the D4 redundancy counter. When P1 identifies semantically novel data, it queries the D3 admission control token bucket and consumes the token. If admission is successful, the data is scheduled to the E3 high-priority transmission path data destination. If an overflow occurs, the data is diverted to the E4 second-highest-priority transmission path data destination. When P1 identifies semantically redundant data, it schedules it to the E5 low-priority transmission path data destination. At the same time, the P1 process is also responsible for reporting the redundancy frequency index to the E1SDN controller.

[0042] like Figure 2 As shown, the semantic recognition mechanism is implemented by defining field masks and calculating hashes only for key fields. The dynamic path assignment mechanism triggers the first scheduling rule for high-priority paths or the second scheduling rule for low-priority paths based on the comparison between the new hash value and the baseline hash value. To ensure system robustness, the cooperative processing mechanism introduces token bucket-based admission control and redundancy counter-based activity detection. In addition, the network situational awareness mechanism achieves predictive management by monitoring redundancy frequency indicators and reporting them to the controller. Figure 3As shown, when a new data packet sent by the data source arrives at the data plane of the network switch, the switch sequentially performs the following operations: application field mask, extraction of predetermined key data fields, calculation of new hash value, reading of the cached baseline hash value, and comparison of the new hash value with the baseline hash value. If the hash value is determined to be inconsistent, i.e., the data is semantically novel, the switch will set a metadata tag for the data packet, trigger the first scheduling rule, update the baseline hash value to the current new hash value, and schedule the data packet to a high-priority transmission path. Finally, the data packet is quickly delivered to the receiving end through the low-latency transmission of the high-priority path.

[0043] Example 5: This example illustrates an engineering step for determining key operating parameters in the method of the present invention. This step is based on objective system specifications and Service Level Agreement (SLA) requirements, rather than relying on experience. During the system initialization configuration phase, it is necessary to determine the average admission rate of the token bucket admission control. Sudden access size and redundant counter heartbeat threshold Regarding the specific values, the steps disclosed in this embodiment involve associating the settings of these parameters with the network switch hardware specifications and the service level agreement of the upper-layer monitoring system; for the token bucket parameters... and The technical objective of determining the P0 path is to prevent semantically novel data bursts from overflowing the physical queue of the high-priority transmission path (P0) while reserving bandwidth for other critical services in the P0 path, even when such bursts occur in the target data stream of the flight data recorder. The determination process involves: obtaining the hardware specifications of the P0 path on the network switch, determining that the physical buffer depth of the P0 queue is 256 KByte and its egress rate is 10 Mbps; subsequently, confirming that the bandwidth reserved for other critical services in the P0 path is 2 Mbps; and then calculating parameters accordingly. The value is the P0 egress rate minus the reserved bandwidth, i.e. Simultaneously calculate parameters The value is set to no more than a percentage of the P0 physical cache depth; here it is set to 50%, i.e. The resulting average admission rate of 8 Mbps and burst admission size of 128 KByte help prevent the P0 queue from overflowing during data bursts.

[0044] For redundancy counter threshold The technical objective of this determination is to provide data activity proof to the upper-level monitoring system within a predetermined time window by sending a forced heartbeat packet to the P0 path when the sensor experiences a silent failure (i.e., continuously sending redundant data). The determination steps are as follows: Obtain the data activity SLA index of the target data stream from the upper-level monitoring system, and determine that the monitoring system requires the silent failure state of the FDR data stream to be detected within 30 seconds; based on this, the heartbeat time period is set. 50% of the SLA cycle, that is ; Obtain the average data packet transmission time interval of the target data stream To obtain the FDR data stream The timeout is 100 milliseconds; finally, the threshold is calculated. Its calculation formula is Substitute the values ​​and you get In the formula, The preset heart rate time period. Set the average time interval for sending data packets; set a threshold. This allows the system to react to silent failures within a 15-second time window to meet SLA requirements. Furthermore, the SDN controller performs flow control operations based on redundancy frequency metrics. Its decision-making logic is as follows: The SDN controller periodically checks the redundancy frequency metrics from the data plane, e.g., every second. If both the redundancy frequency metrics of the FDR data stream and the EICAS data stream are below 90%, the controller determines that the system has entered an unstable state and issues a new rule limiting the total bandwidth limit of non-target data streams for cabin entertainment from 100Mbps to 20Mbps. If the above conditions are not met, the controller restores the bandwidth limit to 100Mbps. This logic provides an automated network resource reallocation mechanism to respond to changes in data source stability.

[0045] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention.

[0046] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A method for SDN-based transmission priority scheduling of aviation multi-source data, characterized in that, The method comprises the following steps: defining, by an SDN controller, a field mask for identifying a predetermined key data field in a data packet of a target data stream according to data format characteristics of the target data stream, the predetermined key data field excluding a timestamp and a sequence number field; the SDN controller issuing the field mask, a first scheduling rule for a high-priority transmission path, and a second scheduling rule for a low-priority transmission path to a network switch; the network switch applying the field mask to extract only the predetermined key data field from a reference data packet of the target data stream and calculating a reference hash value only for the extracted predetermined key data field, and caching the reference hash value; when the network switch receives a new data packet of the target data stream, applying the field mask to extract only the predetermined key data field from the new data packet and calculating a new hash value only for the extracted predetermined key data field; comparing the new hash value with the reference hash value, and if the new hash value is inconsistent with the reference hash value, triggering the first scheduling rule to schedule the new data packet to the high-priority transmission path and updating the reference hash value to the new hash value; if the new hash value is consistent with the reference hash value, triggering the second scheduling rule to schedule the new data packet to the low-priority transmission path. 2.The SDN-based aerial multi-source data transmission priority scheduling method of claim 1, wherein, The target data stream is a data stream from an aircraft flight data recorder; the high-priority transmission path is a low-latency transmission path; and the low-priority transmission path is a batch compression transmission path. 3.The SDN-based aerial multi-source data transmission priority scheduling method of claim 1, wherein, The field mask is a list of byte offsets identifying the predetermined key data field in the data packet of the target data stream; and the network switch performs the calculation of the reference hash value and the new hash value and the comparison of the new hash value with the reference hash value on the data plane. 4.The SDN-based aerial multi-source data transmission priority scheduling method of claim 1, wherein, If the new hash value is inconsistent with the reference hash value, triggering the first scheduling rule comprises setting a metadata flag for the new data packet, and the network switch matches the new data packet to the first scheduling rule based on the metadata flag.

5. The SDN-based aerial multi-source data transmission priority scheduling method according to claim 1, characterized in that, The method further comprises performing admission control on the new data packet before scheduling the new data packet to the high-priority transmission path; if the admission control determines to allow, scheduling the new data packet to the high-priority transmission path; and if the admission control determines to disallow, scheduling the new data packet to a predefined next-high-priority transmission path. 6.The SDN-based aerial multi-source data transmission priority scheduling method of claim 5, wherein, The admission control is implemented based on a token bucket mechanism; and the SDN controller configures the token bucket mechanism for the high-priority transmission path, the token bucket mechanism defining an average admission rate and a burst admission size; the admission control determines to allow corresponding to the presence of an available token in the token bucket mechanism, and the new data packet consumes an available token; and the admission control determines to disallow corresponding to the absence of an available token in the token bucket mechanism.

7. The SDN-based aerial multi-source data transmission priority scheduling method according to claim 1, characterized in that, The method further comprises: maintaining a redundancy counter associated with the target data flow; clearing the redundancy counter when the step of updating the reference hash value to the new hash value is performed; before triggering the second scheduling rule, adding one to the redundancy counter and determining whether the redundancy counter reaches a preset threshold N; if the redundancy counter does not reach the preset threshold N, triggering the second scheduling rule; if the redundancy counter reaches the preset threshold N, triggering the first scheduling rule forcibly, scheduling the new data packet to the high-priority transmission path, updating the reference hash value to the new hash value, and clearing the redundancy counter. 8.The SDN-based aerial multi-source data transmission priority scheduling method of claim 7, wherein, The preset threshold N is calculated by wherein, is a preset heartbeat time period for confirming the activity of the target data flow; is an average sending time interval of the data packets of the target data flow. 9.The SDN-based aerial multi-source data transmission priority scheduling method of claim 1, wherein, The method further comprises: at the network switch, monitoring a frequency of triggering the second scheduling rule within a preset time period to obtain a redundancy frequency indicator; and reporting the redundancy frequency indicator to the SDN controller. 10.The SDN-based aerial multi-source data transmission priority scheduling method of claim 9, wherein, The method further comprises: the SDN controller determining stability of the target data flow based on the redundancy frequency indicator; and when the SDN controller determines that the stability of the target data flow decreases, the SDN controller performing a predefined traffic control operation on at least a non-target data flow in the network.

Images