A Scheduling Method for Time-Triggered Flows in Time-Sensitive Networks

Abstract

The present invention discloses a scheduling method for time-triggered flow in a time-sensitive network, which belongs to the field of communication. Two highest priority transmission queues are used to transmit TT flows, and the two transmission queues have different priorities. The TT flow in the highest priority transmission queue is an HT flow, and the flow in the second highest priority transmission queue is an LT flow. It is assumed that the best path in the TT flow path set is the original path, and the remaining paths are redundant paths. Considering the schedulability of the LT flow, combined with the worst response time analysis, it is considered whether the LT flow is converted into the HT flow, and finally a feasible scheduling scheme is found. When the bandwidth capacity is sufficient, the priority of the LT flow that exceeds the deadline can be increased, and the response time of the HT flow can be reduced while ensuring that the LT flow meets the schedulability.

Description

Technical Field

[0001] The present invention relates to the field of communications, and in particular to a scheduling method for time-triggered flows in a time-sensitive network. Background Art

[0002] In modern industrial and communication networks, real-time and reliability have become key factors. Although traditional Ethernet has been widely used in many fields, it still has significant deficiencies in handling time-sensitive data transmission. To meet this challenge, time-sensitive networking (TSN) came into being. TSN is a communication network supported by a series of IEEE 802.1 protocols. It aims to meet the stringent requirements of data communication in industrial automation, automotive electronics, audio and video synchronization, aerospace and other fields by providing deterministic low latency and high reliability of high quality of service (QoS). By introducing technologies such as clock synchronization, traffic shapers and frame duplication elimination, TSN can support mixed mission-critical and non-mission-critical traffic in the same network, thereby improving the overall performance and reliability of the network.

[0003] During the TSN message transmission process, due to interference in the environment (such as lightning, electromagnetic interference, etc.), data transmission errors or transient link failures may occur, and these effects are difficult to eliminate. In order to ensure the reliability of TSN network transmission, the TSN working group joined the IEEE802.1CB frame replication and elimination protocol to meet the reliability of the time-triggered flow (TT flow) in TSN by increasing the number of redundant flows and redundant paths. Although adding redundant messages can bring higher reliability to the network, it also brings negative effects, such as occupying more bandwidth, causing TT flows to have higher jitter (the change in transmission delay) and end-to-end response time, and even causing TT flows to exceed the deadline, ultimately resulting in a decrease in the schedulability of the entire network. In particular, when the scheduling table is poorly configured, resulting in improper allocation of TT flow transmission time slots or there are too many redundant TT flows on the link that need to be transmitted, the delay interference between TT flows is also more serious, and the schedulability of TT flows will also be worse. At the same time, considering that the probability of transmission error is extremely low, and the scheduling table of TT flow in a TSN is configured offline; when there is a lot of redundant TT flow in the network, the delay jitter suffered by the TT flow is relatively large compared to the extremely low probability of failure, which is a waste of bandwidth utilization in actual use. However, in order to ensure network reliability, if the link redundancy strategy is adopted, the trade-off between reliability and schedulability has always been unsolvable. Therefore, it is necessary to improve the schedulability of TT flow in the network system as much as possible while ensuring reliability. In terms of link redundancy, there is currently no relevant research that considers the delay impact of redundant TT flow on TT flow. Summary of the invention

[0004] In order to solve the above technical problems, the present invention provides a scheduling method for time-triggered flows in a time-sensitive network with simple algorithm and high reliability.

[0005] The technical solution of the present invention to solve the above technical problem is: a scheduling method for time-triggered flow in a time-sensitive network, comprising the following steps:

[0006] Step S1: construct a topology map of the entire time-sensitive network according to the actual time-sensitive network structure;

[0007] Step S2: Obtain the parameters of the time-triggered flow to be sent by each node in the topology graph, that is, the TT flow;

[0008] Step S3: Based on the IEEE 802.1CB frame duplication and elimination protocol, the redundant frames and the original frames are transmitted simultaneously through different paths to determine the number of redundant flows required for each TT flow from the sending address to the destination address;

[0009] Step S4: A path set that meets the requirement of the number of redundant flows in step S3 is configured for each TT flow through the shortest path algorithm. The shortest path in the path set is the best path, that is, the original path, and the rest are redundant paths.

[0010] Step S5: When each TT flow passes through the first switch, the frame is duplicated, and each duplicated frame is transmitted through a different path in the path set. The receiving end detects whether the frame transmission is wrong according to the frame check code FCS, and eliminates the wrong and duplicate frames.

[0011] Step S6: setting a preliminary scheduling scheme, making the time-triggered flow transmitted on the best path the HT flow, and the time-triggered flow transmitted on the redundant path the LT flow, making the HT flow the highest priority and the LT flow the second priority;

[0012] Step S7: Determine whether the bandwidth occupied by the TT flow exceeds 70% of the total bandwidth. If so, do not perform schedulability analysis on the LT flow, and directly use the preliminary scheduling plan as the final plan for scheduling. If not, proceed to step S8.

[0013] Step S8: Analyze the response time of each LT flow in an iterative manner to determine whether the LT flow on the redundant path can complete scheduling. If so, continue to maintain the second priority for transmission. If not, upgrade it to the first priority and turn it into an HT flow. Perform schedulability analysis on all LT flows again until all LT flows meet schedulability requirements and obtain the final scheduling solution.

[0014] Step S9: Use the M / G / 1 queuing theory model to analyze the starvation probability of the TT flow and verify the reliability of the final scheduling solution.

[0015] In the above-mentioned scheduling method for time-triggered flow in the time-sensitive network, in the step S1, the time-sensitive network is TSN, and the TSN is composed of multiple nodes and links connecting the nodes, wherein the nodes are switches SW or terminal systems ES, and the switches and the terminal systems are connected through links with fixed transmission bandwidth;

[0016] The TSN topology is modeled as a directed graph G = (V, L), where V represents the node set, denoted as V = ES ∪ SW, and L represents the link set between nodes, denoted as ;use Represents a node, =[ u , v ] indicates a slave node u To Node v The link u , v ∈V,ρ∈L,[ u , v ]= [ v , u ], ESs communicate with each other through a path composed of multiple links or through multiple paths.

[0017] In the above-mentioned scheduling method for time-triggered flow in the time-sensitive network, in step S2, the parameters of the TT flow include period, deadline, sending address, and destination address.

[0018] In the above-mentioned scheduling method for time-triggered flows in a time-sensitive network, in step S4, the specific steps of forming a path set are:

[0019] Step S41: Select the original path: determine the original path of each TT flow according to the shortest response time or the shortest route. This is the default path for data transmission when redundancy is not considered.

[0020] Step S42: Select the safety requirement level: According to the application scenario, select the safety requirement PFH according to the safety level in the DO-178B standard. The levels range from A to E. Level A corresponds to mission-critical scenarios and requires the highest reliability, while level E corresponds to situations with no safety impact. By selecting different levels, the safety and reliability of different flows can be ensured;

[0021] Step S43: Select redundant paths: After determining the required security level, determine redundant paths, which must be disjoint with the original paths to ensure that the redundant paths do not share any critical links or nodes that may cause simultaneous failures; use the shortest path algorithm or the disjoint path algorithm to generate redundant paths, and finally integrate the path set consisting of the original path and the redundant path into the routing configuration of the time-sensitive network;

[0022] Step S44: Path set evaluation: The number of paths n in the path set must satisfy the following formula:

[0023] n≥log Fi PFH

[0024] Where Fi is the failure probability value of the path with the highest failure probability in the path set, and PFH is the probability of error per hour; the obtained path set is evaluated to ensure that the reliability requirements are met. If the reliability requirements are not met, the redundant path is reselected;

[0025] Step S45: Select the final path set: Generate a path set for each TT flow, which includes the original path and the redundant path, to ensure that the time-sensitive TSN can maintain the required performance and reliability even when a failure occurs. The original TT flow selects the original path in the path set for transmission, and the redundant TT flow selects the redundant path in the path set for transmission.

[0026] In the above-mentioned scheduling method for time-triggered flow in the time-sensitive network, in step S5, the frame duplication and elimination in IEEE 802.1CB enables the sending end to duplicate the time-triggered frame into two frames with the same content and send them to the receiving end at the same time. The receiving end checks whether there is an error based on the check code of the frame. If there is no error in the transmission of the frame, the frame is received, otherwise the frame is discarded; at the same time, the receiving end eliminates the frames with duplicate content through the recovery function IRF and the sequence recovery function SRF.

[0027] In the above-mentioned scheduling method for time-triggered flows in the time-sensitive network, in step S8, before performing schedulability analysis on the LT flow, the type of delay interference suffered by the transmission of the LT flow is analyzed, and the delay interference suffered by the transmission of the LT flow includes the delay interference of the highest priority HT flow and the delay interference of the LT flow with the same priority, wherein the analysis process of the delay interference of the highest priority HT flow is:

[0028] First, an offset is determined for each frame in the stream instance of all TT streams; suppose there is a stream instance belonging to stream s i Frames i j , where j is the index of the frame, and there is another frame s a b , and s a bhas a lower priority than stream s i Frames in

[0029] Then, analyze s a b The worst response time; when s a b With Flows i When a frame in the same time is activated, the worst delay interference may occur, so this time is called s a b The key candidate moment of i The activation time of any flow instance in the frame is a key candidate time. In addition, there may be other time points in the time-sensitive network. a b Higher priority flows, these are higher than s a b The activation time of the frame of the stream instance in the higher priority stream will be a key candidate time. The period of all TT streams constitutes a super period; there are multiple s in this super period. a b The key candidate moment is calculated as s a b The interference delay at each key candidate time, among which the largest interference delay is s a b The worst case interference delay encountered; i j For a b There are two types of interference. The first is that a b The frame activated before the key candidate moment of s, but due to the delay, a b Activate at the same time; the second is in s a b activation that occurs after the critical candidate moment;

[0030] Next, in the link The minimum common period of all flows on ,link HT frame m h It needs to be transmitted N times within the minimum common period. The kth transmission is the kth frame. The period is , considering m h Offset and protection band G, G=0 without holding and release mechanism, m h Phase set Through formula (1), we can get:

[0031] (1)

[0032] Assume that HT stream frame m c The activation time of m is the key candidate time, and m h As the interference frame of the LT flow being analyzed, equation (2) is expressed by m h Offset Subtract m c Phase Then m h The period modulo, we get m h With m c The phase difference between :

[0033] (2)

[0034] Among them, mod is the remainder function;

[0035] Since other frames do not apply jitter to the time-triggered frame, there is no need to consider jitter in equation (2);

[0036] Finally, in m c When the activation time is the key candidate time, the transmission duration under duration t is The worst-case delay interference caused by HT frames to LT frames , see formula (3):

[0037] (3).

[0038] The analysis process of the scheduling method for time-triggered flows in the above time-sensitive network affected by the delay interference of LT frames with the same priority level is as follows:

[0039] Based on the analysis of the phase difference, the phase difference between the interference frame and the analyzed frame is obtained. Then, based on the transmission time required for the interference frame, the interference frame m with the same priority is obtained by formula (4): h For the analyzed frame m i The delay impact :

[0040] (4)

[0041] In the formula is m i and m h The phase difference, and m h and m i are of the same priority, is frame m i The path set; when ,and When , formula (4) takes effect.

[0042] In the above-mentioned scheduling method for time-triggered flows in a time-sensitive network, in step S8, the process of determining whether the LT flow on the redundant path can complete the scheduling is as follows:

[0043] HT flow m c The activation time of the kth frame is regarded as the key candidate time, and the link is calculated by iterating the sum of all interferences through formula (5). Upper LT flow m i The delay interference caused by HT flow and LT flow of the same priority :

[0044] (5)

[0045] Since the LT flow belongs to the TT flow and cannot be preempted, the iteration starts from RT = ε. Or the iteration ends when the response time exceeds the deadline, where represents the number of iterations, Represents an infinitesimal unit of time;

[0046] From the stream i The maximum response time is selected from the many key candidate moments of all frames contained in the stream s i In the link The worst-case response time on the path is finally obtained by adding up the worst-case response times of all links on the path. i Worst case end-to-end response time , as shown in (6):

[0047] (6)

[0048] For flow i The transmission time of

[0049] Through formula (6), we can get the LT flow on the path The worst-case end-to-end response time on i The scheduling is considered to have failed.

[0050] In the above-mentioned scheduling method for time-triggered flows in a time-sensitive network, the specific process of step S9 is as follows:

[0051] Assuming that the arrival process of the flow follows the Poisson distribution, it means that the arrival process of the first data frame of all flow instances conforms to the Poisson distribution. Therefore, it is assumed that one frame is a flow instance. At the same time, since the processing time of each flow by the TSN switch is non-exponential, in order to be practical, the M / G / 1 queuing theory model is used to analyze the starvation probability of the TT flow.

[0052] The M / G / 1 queuing theory model is adopted, where M represents the arrival process following the Poisson distribution, G represents the service time following the general distribution, and 1 represents a single processing unit;

[0053] First, according to the arrival rate of TT flow and the switch service rate Calculate the utilization of TT flow :

[0054] (7)

[0055] According to the M / G / 1 queuing theory, the average waiting time of high-priority traffic is W T It is expressed as:

[0056] (8)

[0057] in, is the second-order moment of the high-priority traffic service time;

[0058] The average waiting time W of low priority traffic L It consists of two parts: the impact of high-priority traffic on it and its own service waiting time:

[0059] (9)

[0060] in is the second-order moment of the arrival rate, utilization, and service time of the high-priority flow, is the arrival rate of low priority flows;

[0061] Assuming that the waiting time of TT flow follows exponential distribution, D T is the maximum allowed time, then the probability of starvation is calculated as:

[0062] (10).

[0063] The beneficial effects of the present invention are:

[0064] 1. The present invention proposes an end-to-end response time analysis method for LT flows. When the bandwidth capacity is sufficient, the priority of LT flows that exceed the deadline can be increased, and the response time of HT flows can be reduced while ensuring that the LT flows meet the schedulability.

[0065] 2. The present invention proposes a dual-queue transmission (DT) scheduling method for TT flows. Under the premise of satisfying the reliability of TT flows, the interference of redundant TT flows to original TT flows can be reduced, and the schedulability of TT flows can be improved.

[0066] 3. The present invention evaluates and studies the schedulability and reliability of the scheduling method by comparing it with other most advanced methods. The experimental results show that under the premise of meeting the reliability, the schedulability of the present invention is 15% higher than that of other methods, and the average response time is optimized by nearly 40%. BRIEF DESCRIPTION OF THE DRAWINGS

[0067] Figure 1 It is the overall flow chart of the present invention.

[0068] Figure 2 This is the network topology diagram of TSN.

[0069] Figure 3 This is a case diagram of a key candidate moment.

[0070] Figure 4 The schematic diagram of dual-queue transmission.

[0071] Figure 5 This is a comparison chart of the schedulability results of the present invention and link redundancy.

[0072] Figure 6 This is a comparison chart of average response time results between the present invention and link redundancy.

[0073] Figure 7 This is a comparison chart of the reliability results of the present invention and link redundancy.

[0074] Figure 8 The failure probability of the present invention and the non-redundant TSN model is λ=10 -6 The schedulability comparison result diagram.

[0075] Fig. 9 The failure probability of the present invention and the non-redundant TSN model is λ=10 -5 The schedulability comparison result diagram.

[0076] Fig.10 The failure probability of the present invention and the non-redundant TSN model is λ=10 -4 The schedulability comparison result diagram. DETAILED DESCRIPTION

[0077] The present invention will be further described below in conjunction with the accompanying drawings and embodiments.

[0078] like Figure 1 As shown, a scheduling method for time-triggered flows in a time-sensitive network includes the following steps:

[0079] Step S1: construct a topology map of the entire time-sensitive network according to the actual time-sensitive network structure.

[0080] Time-sensitive network is TSN. TSN is generally composed of multiple nodes and links connecting the nodes, where the nodes are switches SW or terminal systems ES. Figure 2 A small TSN network topology diagram is shown, in which the double arrows represent links with fixed transmission bandwidth, and the general link uses a full-duplex twisted pair as a transmission link. The dotted line represents the two transmission paths for the terminal system ES1 to send messages to the terminal system ES4.

[0081] The TSN topology is modeled as a directed graph G = (V, L), where V represents the node set, denoted as V = ES ∪ SW, and L represents the link set between nodes, denoted as ;use Represents a node, =[ u , v ] indicates a slave node u To Node v The link u , v ∈V,ρ∈L,[ u , v ]= [ v , u ], ESs communicate with each other through paths composed of multiple links. ESs can also communicate with each other through multiple paths, such as Figure 2 ES1 in the example can send messages to ES4 through two paths: path I and path II. Path I can be expressed as .

[0082] Step S2: Obtain the parameters of the time-triggered flow, ie, the TT flow, to be sent by each node in the topology graph. The parameters of the TT flow include a period, a deadline, a sending address, and a destination address.

[0083] Step S3: Based on the IEEE 802.1CB frame duplication and elimination protocol, the redundant frames and the original frames are transmitted simultaneously through different paths to determine the number of redundant flows required for each TT flow from the sending address to the destination address.

[0084] Step S4: A path set that meets the redundant flow quantity requirement in step S3 is configured for each TT flow through the shortest path algorithm. The shortest path in the path set is the best path, that is, the original path, and the rest are redundant paths.

[0085] The specific steps to form a path set are:

[0086] Step S41: Selecting the original path: determining the original path of each TT flow according to the shortest response time or the shortest route, which is the default path for data transmission when redundancy is not considered.

[0087] Step S42: Select the safety requirement level: According to the application scenario, select the safety requirement PFH according to the safety level in the DO-178B standard. The levels range from A to E. Level A corresponds to mission-critical scenarios and requires the highest reliability, while level E corresponds to situations with no safety impact. By selecting different levels, the safety and reliability of different flows can be ensured.

[0088] Step S43: Select redundant paths: After determining the required security level, determine redundant paths, which must be disjoint with the original paths to ensure that the redundant paths do not share any critical links or nodes that may cause simultaneous failures; use the shortest path algorithm or the disjoint path algorithm to generate redundant paths, and finally integrate the path set consisting of the original path and the redundant path into the routing configuration of the time-sensitive network;

[0089] Step S44: Path set evaluation: The number of paths n in the path set must satisfy the following formula:

[0090] n≥log Fi PFH

[0091] Where Fi is the failure probability value of the path with the highest failure probability in the path set, and PFH is the probability of error per hour. The obtained path set is evaluated to ensure that the reliability requirements are met. If the reliability requirements are not met, the redundant path is reselected.

[0092] Step S45: Select the final path set: Generate a path set for each TT flow, which includes the original path and the redundant path, to ensure that the time-sensitive TSN can maintain the required performance and reliability even in the event of a failure. The original TT flow selects the original path in the path set for transmission, and the redundant TT flow selects the redundant path in the path set for transmission.

[0093] Step S5: When each TT stream passes through the first switch, the frame is copied and each copied frame is transmitted through a different path in the path set. The receiving end detects whether the frame transmission is erroneous according to the frame check code FCS and eliminates erroneous and duplicate frames.

[0094] The frame duplication and elimination in IEEE 802.1CB enables the transmitter to duplicate the time-triggered frame into two frames with the same content and send them to the receiver at the same time. The receiver checks whether there is an error based on the frame check code. If there is no error in the frame transmission, the frame is received, otherwise the frame is discarded. At the same time, the receiver eliminates frames with duplicate content through the recovery function IRF and the sequence recovery function SRF.

[0095] Step S6: Set a preliminary scheduling plan, and make the time-triggered flow transmitted on the best path the HT flow, and the time-triggered flow transmitted on the redundant path the LT flow. Let the HT flow be at the highest priority and the LT flow be at the second priority, so that the HT flow has the shortest response time to ensure the real-time requirements. The LT flow is transmitted through the redundant path to ensure the reliability of the time-sensitive TSN.

[0096] Step S7: Determine whether the bandwidth occupied by the TT flow exceeds 70% of the total bandwidth. If so, do not perform schedulability analysis on the LT flow, and directly use the preliminary scheduling plan as the final plan for scheduling. If not, proceed to step S8.

[0097] Step S8: Analyze the response time of each LT flow in an iterative manner to determine whether the LT flow on the redundant path can complete the scheduling. If it can complete the scheduling, continue to maintain the second priority for transmission. If not, upgrade it to the first priority and turn it into an HT flow. Re-analyze the schedulability of all LT flows until all LT flows meet the schedulability requirements and obtain the final scheduling plan.

[0098] Since the transmission of HT flows in the present invention is less affected by delay interference and other factors than that in the classic link redundancy TSN, there is no need to perform schedulability analysis on HT flows. Instead, the focus should be on the schedulability of LT flows. The key to obtaining a feasible solution that allows a higher acceptance rate for TT flows while meeting reliability is to analyze the schedulability of LT flows and decide whether to convert LT flows into HT flows.

[0099] Before performing schedulability analysis on the LT flow, analyze the type of delay interference suffered by the transmission of the LT flow. The delay interference suffered by the transmission of the LT flow includes the delay interference of the highest priority HT flow and the delay interference of the LT flow with the same priority. The analysis process of the delay interference of the highest priority HT flow is as follows:

[0100] First, an offset is determined for each frame in the stream instance of all TT streams; suppose there is a stream instance belonging to stream s i Frames i j , where j is the index of the frame, and there is another frame s a b , and s ab has a lower priority than stream s i Frames in

[0101] Then, analyze s a b The worst response time; when s a b With Flows i When a frame in the same time is activated, the worst delay interference may occur, so this time is called s a b The key candidate moment of i The activation time of any flow instance in the frame is a key candidate time. In addition, there may be other time points in the time-sensitive network. a b Higher priority flows, these are higher than s a b The activation time of the frame of the stream instance in the higher priority stream will be a key candidate time. The period of all TT streams constitutes a super period; there are multiple s in this super period. a b The key candidate moment is calculated as s a b The interference delay at each key candidate time, among which the largest interference delay is s a b The worst case interference delay encountered; i j For a b There are two types of interference. The first is that a b The frame activated before the key candidate moment of s, but due to the delay, a b Activate at the same time; the second is in s a b activation that occurs after the critical candidate moment;

[0102] Since the flow being analyzed in the model also belongs to TT flow, the retention and release mechanism is not used. Figure 3As shown in the figure, all the key candidate moments of the LT flow occur at the activation moment of the HT flow. HT1 and HT2 are the original flows, that is, the HT flows, and the periods of HT1 and HT2 are T1 and T2 respectively. LT1 is the redundant flow, that is, the LT flow. The transmission time required for HT1, LT1 and HT2 are 2, 3 and 3 time units respectively. Focus on the LT1 frame. In case I, the key moment candidate is considered to be the activation moment of the first frame of HT1. At the same time, since the priority of HT1 and HT2 is higher than LT, HT2 is activated immediately after HT1 transmission is completed. LT needs to wait for HT2 transmission to start transmission, making the response time of LT 8 time units. In case II, the key moment candidate is considered to be the activation moment of the HT2 frame. In this case, the response time of LT is 6 time units. In case III, the key moment candidate is considered to be the activation moment of the second frame of HT1. At this time, the response time of LT is 5 time units. From these three cases, it can be known that the activation moment of the first frame of HT1 is the key moment of LT, and the worst response time will occur at this moment. Therefore, the activation of any HT frame in an LCM may cause the worst-case response time of the analyzed frame. Based on this conclusion, it is necessary to determine the phase of multiple HT frames in an LCM.

[0103] Next, in the link The minimum common period of all flows on ,link HT frame m h It needs to be transmitted N times within the minimum common period. The kth transmission is the kth frame. The period is , considering m h Offset and protection band G, G=0 without holding and release mechanism, m h Phase set Through formula (1), we can get:

[0104] (1)

[0105] Assume that HT stream frame m c The activation time of m is the key candidate time, and m h As the interference frame of the LT flow being analyzed, equation (2) is expressed by m h Offset Subtract m c Phase Then m h The period modulo, we get m h With m c The phase difference between :

[0106] (2)

[0107] Among them, mod is the remainder function.

[0108] It is worth noting that since other frames do not apply jitter to the time-triggered frame, there is no need to consider jitter in equation (2);

[0109] Finally, in m c When the activation time is the key candidate time, the transmission duration under duration t is The worst-case delay interference caused by HT frames to LT frames , see formula (3):

[0110] (3)

[0111] The analysis process of delay interference of LT flows with the same priority is as follows:

[0112] Based on the analysis of the phase difference, the phase difference between the interference frame and the analyzed frame is obtained. Then, based on the transmission time required for the interference frame, the interference frame m with the same priority is obtained by formula (4): h For the analyzed frame m i The delay impact :

[0113] (4)

[0114] In the formula is m i and m h The phase difference, and m h and m i are of the same priority, is frame m i The path set; when ,and When , formula (4) takes effect.

[0115] The process of determining whether the LT flow on the redundant path can complete scheduling is as follows:

[0116] HT flow m c The activation time of the kth frame is regarded as the key candidate time, and the link is calculated by iterating the sum of all interferences through formula (5). Upper LT flow m i The delay interference caused by HT flow and LT flow of the same priority :

[0117] (5)

[0118] Since the LT flow belongs to the TT flow and cannot be preempted, the iteration starts from RT = ε. Or the iteration ends when the response time exceeds the deadline, where represents the number of iterations, Represents an infinitesimal unit of time.

[0119] From the stream i The maximum response time is selected from the many key candidate moments of all frames contained in the stream s i In the link The worst-case response time on the path is finally obtained by adding up the worst-case response times of all links on the path. i Worst case end-to-end response time , as shown in (6):

[0120] (6)

[0121] For flow i The transmission time of

[0122] Through formula (6), we can get the LT flow on the path The worst-case end-to-end response time on i The scheduling is considered to have failed.

[0123] In order to improve the acceptance rate of TT flows, different from the classic link redundant TSN transmission model, e.g. Figure 4 As shown, the present invention uses dual transmission queues (Q7 and Q6) to transmit HT streams and LT streams respectively, Q5 and Q4 to transmit AVB streams, and other transmission queues to transmit BE streams. Initially, the stream transmitted on the original path is the HT stream, and the stream transmitted on the redundant path is the LT stream.

[0124] As Figure 4As shown in the figure, the dark boxes represent the TT stream sent by ES1, and the white boxes represent the TT stream sent by ES2. At the same time, it is assumed that the solid line {[ES1, SW1], [SW1, SW2], [SW2, SW4], [SW4, ES3]} is the best path for the TT stream sent by ES1 and is also the original path, and the remaining solid line paths are redundant paths. The dotted line {[ES2, SW1], [SW1, SW3], [SW3, SW4], [SW4, ES3]} is the best path for the TT stream sent by ES2 and is also the original path, and the remaining dotted line paths are redundant paths. The edge switches SW1 and SW4 respectively implement the functions of frame replication and elimination. In the intermediate switch SW2, the TT stream sent by ES1 belongs to the HT stream and is located in the Q7 transmission queue, and the TT stream sent by ES2 belongs to the LT stream and is located in the Q6 transmission queue. This is because switch SW2 is on the best path for the TT stream sent by ES1, while for the TT stream sent by ES2, SW2 is only on the redundant path. Similarly, in SW3, the TT stream of ES2 is in the Q7 transmission queue, and the TT stream of ES1 is in the Q6 transmission queue. In this way, the TT stream sent by each node can have a transmission path in the path set with the highest priority. From another perspective, by reducing the number of TT frames in the highest priority transmission queue, the interference between TT frames can be reduced, thereby achieving the purpose of reducing the response time of TT frames.

[0125] Step S9: Use the M / G / 1 queuing theory model to analyze the starvation probability of the TT flow and verify the reliability of the final scheduling solution.

[0126] Assuming that the arrival process of the flow follows the Poisson distribution, it means that the arrival process of the first data frame of all flow instances conforms to the Poisson distribution. Therefore, it is assumed that a frame is a flow instance. At the same time, since the processing time of each flow by the TSN switch is non-exponential, in order to be practical, the M / G / 1 queuing theory model is used to analyze the starvation probability of the TT flow.

[0127] The M / G / 1 queuing theory model is adopted, where M represents the arrival process following the Poisson distribution, G represents the service time following the general distribution, and 1 represents a single processing unit;

[0128] First, according to the arrival rate of TT flow and the switch service rate Calculate the utilization of TT flow :

[0129] (7)

[0130] According to the M / G / 1 queuing theory, the average waiting time of high-priority traffic is W T It is expressed as:

[0131] (8)

[0132] in, is the second-order moment of the high-priority traffic service time;

[0133] The average waiting time W of low priority traffic L It consists of two parts: the impact of high-priority traffic on it and its own service waiting time:

[0134] (9)

[0135] in is the second-order moment of the arrival rate, utilization, and service time of the high-priority flow, is the arrival rate of low priority flows.

[0136] Assuming that the waiting time of TT flow follows exponential distribution, D T is the maximum allowed time, then the probability of starvation is calculated as:

[0137] (10)

[0138] Example Calculation

[0139] A simple example is used to prove that the probability of TT flows sent by nodes in the present invention being starved of queues is lower than that of classic TSN. Assume that each node sends a TT flow with There are redundant paths, and each path passes through only one switch node. The parameters of the switch are as follows: the service rate of the switch μ = 100 frames / second, the maximum allowed time D T = 0.05 seconds, the arrival rate of TT flow in classic TSN = 10 frames / second, the arrival rate of the HT stream in the present invention is , the arrival rate of LT flow is ,in It is positively correlated with the ratio of the number of HT and LT flows in the present invention, and 0≤ ≤1.

[0140] First, we calculate the probability of TT flow starvation in the classic model, that is, =10 frames / second is substituted into the above formula (7), and the average waiting time is calculated by formula (8), and finally the probability of TT flow starving on a path in the classic TSN model is obtained as exp(-4.5045); then the probability of HT flow starving on a path in the model of the present invention is calculated in the same way, and the probability of HT flow starving on a path in the model of the present invention is obtained as Finally, the starvation probability of the LT flow needs to be calculated. The only difference is that the average waiting time is calculated using formula (9). It can be concluded that the starvation probability of the LT flow on a path in the present invention is Since each node has n redundant paths for the TT flow it sends, the probability of starvation of the TT flow sent by the node in the classic TSN is , in the present invention, the probability of starvation of the TT flow sent by the node is .

[0141] The starvation probability of TT flows in the two scheduling models can be compared by the following polynomial:

[0142]

[0143] And after analysis and calculation, there is no 0< A value <1 makes Δ less than 0. It can be concluded that:

[0144] Conclusion: For any flow instance in any TT flow set, it is sent from the sending node, received by the receiving node after passing through the FRER of the intermediate switch, where the probability of starvation caused by queuing in the classic link redundant TSN switch is higher than the probability of starvation caused by queuing in the present invention.

[0145] Based on this conclusion, it can be explained that the TT flow schedulability in the present invention (DT-TSN) is always higher than that in the classic link redundancy TSN.

[0146] Through simulation experiments, the acceptance rate, average response time and other aspects of the latest scheduling scheme are compared and analyzed to highlight the advantages of DT-TSN.

[0147] First, the scheduling scheme proposed in the present invention is compared with the link redundancy (LR) scheduling scheme. In LR, the TT flow uses multiple different paths to send redundant flows with the same content, and they are all at the same priority. Figure 5 , Figure 6 and Figure 7 The comparisons are in terms of acceptance rate, average response time and reliability. The acceptance rate refers to the ratio of the number of TT flows that have completed scheduling to the total number of TT flows, which can reflect the schedulability of TT flows in time-sensitive TSN.

[0148] exist Figure 5In the LR scheme, when the number of TT flow types doubles to 24 flows, the schedulability of LR begins to decline. In contrast, even when the number of TT flows reaches 36, the DT-TSN proposed in the present invention can maintain a nearly 100% acceptance rate, which is 25% better than LR. When the number of TT flow types increases to 48, the acceptance rate of DT-TSN also decreases. This can be attributed to the following reasons: although the number of highest priority TT flows in DT-TSN is reduced compared to the LR scheme, the absolute number of highest priority TT flows is increased. Coupled with limited bandwidth resources, this may result in some of the highest priority TT flows failing to complete transmission within the specified deadline.

[0149] exist Figure 6 In the present invention, when the number of TT flow types reaches 36, the acceptance rate of the present invention is still close to 100%, while the response time is 20% lower than that of the LR scheme. These results can be attributed to two main factors. First, in the method proposed in the present invention, the number of the highest priority TT flow types remains unchanged, but the number of flows of each type is reduced, thereby effectively reducing the interference faced by the highest priority TT flows. Second, each type of TT flow has at least one path in the highest priority state, which ensures that each type of TT flow can maintain a low response time. Despite the increasing number of TT flow types, the method of the present invention always maintains a response time lower than the LR scheme. In addition, as the number of TT flow types increases, the method of the present invention can still maintain a high degree of schedulability.

[0150] exist Figure 7 In the LR, the lower the FPH value, the higher the reliability of the system. As the number of flows increases, the number of redundant frames that can be correctly scheduled gradually decreases, and the reliability of the system also decreases. Especially when the number of flows reaches 30, the reliability provided by the redundant frames that can complete the scheduling in LR is not enough to meet the reliability requirements. However, the experimental results show that the failure probability of DT-TSN is slightly lower than that of LR, especially when the number of flows is 30, DT-TSN can still meet the reliability requirements.

[0151] In addition, the present invention is compared with a non-link redundant scheduling scheme (NR-TSN), in which each TT flow is transmitted through only one path and no redundancy is performed. The present invention compares the TT flow acceptance rate under different failure probability parameters λ.

[0152] exist Figure 8 In the results, the acceptance rate of DT-TSN is close to 100% with that of NR-TSN, but the acceptance rate of DT-TSN is slightly higher. The reason for this scheduling result is that when the failure probability λ=10 -6In the DT-TSN model, although each data stream of NR-TSN is sent only once, almost all data frames can be received correctly due to the extremely low probability of failure, and the bandwidth capacity can just meet the transmission requirements of all data frames, so the acceptance rate is close to 100%. In the DT-TSN model, in order to meet the reliability requirements, each TT stream requires 1-2 redundant streams. When a failure occurs, the present invention has enough redundancy to ensure that the data frame can be correctly scheduled, so that the acceptance rate of DT-TSN is slightly higher than that of NR-TSN.

[0153] exist Fig. 9 In the equation, since λ=10 -5 , the data frame is compared with λ=10 -6 Therefore, the acceptance rate of NR-TSN decreases. For the TT flow of DT-TSN, the calculation results show that 3 redundant flows are needed to meet the requirements of formula (5). Although the acceptance rate begins to decrease when the number of flows reaches 18, it is still close to 100% until 36 flows. The reason for the decrease in acceptance rate is that the bandwidth is not enough to support the transmission of some redundant flows at this time, resulting in some redundant flows unable to complete scheduling to compensate for the faulty frames, so the acceptance rate decreases. However, some redundant flows can still complete scheduling and compensate for the faulty frames, making the acceptance rate of DT-TSN higher than that of NR-TSN.

[0154] exist Fig.10 In the experiment, λ is increased to 10 -4 , the probability of failure during frame transmission further increases. In this case, the acceptance rate of NR-TSN drops to 98%. For DT-TSN, when the number of streams is 12 and 18, the acceptance rate is still close to 100%, but as the number of streams continues to increase, the acceptance rate approaches 99%, until it drops below 99% at 36 streams. Despite the significant drop in acceptance rate, DT-TSN still improves its acceptance rate by 1-2% compared to NR-TSN.

Claims

1. A scheduling method for time-triggered flows in a time-sensitive network, characterized in that: The following steps are involved: Step S1: construct a topology map of the entire time-sensitive network according to the actual time-sensitive network structure; Step S2: Obtain the parameters of the time-triggered flow to be sent by each node in the topology graph, that is, the TT flow; Step S3: Based on the IEEE 802.1CB frame duplication and elimination protocol, the redundant frames and the original frames are transmitted simultaneously through different paths to determine the number of redundant flows required for each TT flow from the sending address to the destination address; Step S4: A path set that meets the requirement of the number of redundant flows in step S3 is configured for each TT flow through the shortest path algorithm. The shortest path in the path set is the best path, that is, the original path, and the rest are redundant paths. Step S5: When each TT flow passes through the first switch, the frame is duplicated, and each duplicated frame is transmitted through a different path in the path set. The receiving end detects whether the frame transmission is wrong according to the frame check code FCS, and eliminates the wrong and duplicate frames. Step S6: setting a preliminary scheduling scheme, making the time-triggered flow transmitted on the best path the HT flow, and the time-triggered flow transmitted on the redundant path the LT flow, making the HT flow the highest priority and the LT flow the second priority; Step S7: Determine whether the bandwidth occupied by the TT flow exceeds 70% of the total bandwidth. If so, do not perform schedulability analysis on the LT flow, and directly use the preliminary scheduling plan as the final plan for scheduling. If not, proceed to step S8. Step S8: Analyze the response time of each LT flow in an iterative manner to determine whether the LT flow on the redundant path can complete the scheduling. If the scheduling can be completed, the LT flow continues to maintain the second priority for transmission. If not, it is promoted to the first priority and becomes an HT flow. The schedulability analysis is performed on all LT flows again until all LT flows meet the schedulability and the final scheduling solution is obtained. Step S9: Use the M / G / 1 queuing theory model to analyze the starvation probability of the TT flow and verify the reliability of the final scheduling solution.

2. The scheduling method for time-triggered flows in a time-sensitive network according to claim 1, characterized in that: In step S1, a time-sensitive network is a TSN, which is composed of multiple nodes and links connecting the nodes, wherein the nodes are switches SW or terminal systems ES, and the switches and the terminal systems are connected through links with fixed transmission bandwidth; The TSN topology is modeled as a directed graph G = (V, L), where V represents the node set, denoted as V = ES ∪ SW, and L represents the link set between nodes, denoted as ;use Represents a node, =[ u , v ] indicates a slave node u To Node v The link, u , v ∈V,ρ∈L,[ u , v ]= [ v , u ], ESs communicate with each other through a path composed of multiple links or through multiple paths.

3. The scheduling method for time-triggered flows in a time-sensitive network according to claim 1, characterized in that: In step S2, the parameters of the TT flow include a period, a deadline, a sending address, and a destination address.

4. The scheduling method for time-triggered flows in a time-sensitive network according to claim 1, characterized in that: In step S4, the specific steps of forming a path set are: Step S41: Select the original path: determine the original path of each TT flow according to the shortest response time or the shortest route. This is the default path for data transmission when redundancy is not considered. Step S42: Select the safety requirement level: According to the application scenario, select the safety requirement PFH according to the safety level in the DO-178B standard. The levels range from A to E. Level A corresponds to mission-critical scenarios and requires the highest reliability, while level E corresponds to situations with no safety impact. By selecting different levels, the safety and reliability of different flows can be ensured; Step S43: Select redundant paths: After determining the required security level, determine redundant paths, which must be disjoint with the original paths to ensure that the redundant paths do not share any critical links or nodes that may cause simultaneous failures; use the shortest path algorithm or the disjoint path algorithm to generate redundant paths, and finally integrate the path set consisting of the original path and the redundant path into the routing configuration of the time-sensitive network; Step S44: Path set evaluation: The number of paths n in the path set must satisfy the following formula: n≥log Fi PFH Where Fi is the failure probability value of the path with the highest failure probability in the path set, and PFH is the probability of error per hour; the obtained path set is evaluated to ensure that the reliability requirements are met. If the reliability requirements are not met, the redundant path is reselected; Step S45: Select the final path set: Generate a path set for each TT flow, which includes the original path and the redundant path, to ensure that the time-sensitive TSN can maintain the required performance and reliability even when a failure occurs. The original TT flow selects the original path in the path set for transmission, and the redundant TT flow selects the redundant path in the path set for transmission.

5. The scheduling method for time-triggered flows in a time-sensitive network according to claim 1, characterized in that: In step S5, the frame duplication and elimination in IEEE 802.1CB enables the sending end to duplicate the time-triggered frame into two frames with the same content and send them to the receiving end at the same time. The receiving end checks whether there is an error based on the check code of the frame. If there is no error in the transmission of the frame, the frame is received, otherwise the frame is discarded; at the same time, the receiving end eliminates the frames with duplicate content through the recovery function IRF and the sequence recovery function SRF.

6. The scheduling method for time-triggered flows in a time-sensitive network according to claim 2, characterized in that: In step S8, before performing schedulability analysis on the LT flow, the type of delay interference suffered by the transmission of the LT flow is analyzed, and the delay interference suffered by the transmission of the LT flow includes the delay interference of the highest priority HT flow and the delay interference of the LT flow with the same priority, wherein the analysis process of the delay interference of the highest priority HT flow is: First, an offset is determined for each frame in the stream instance of all TT streams; suppose there is a stream instance belonging to stream s i Frames i j , where j is the index of the frame, and there is another frame s a b , and s a b has a lower priority than stream s i Frames in Then, analyze s a b The worst response time; when s a b With Flows i When a frame in the same time is activated, the worst delay interference may occur, so this time is called s a b The key candidate moment of i The activation time of any flow instance in the frame is a key candidate time. In addition, there may be other time points in the time-sensitive network. a b Higher priority flows, these are higher than s a b The activation time of the frame of the stream instance in the higher priority stream will be a key candidate time. The period of all TT streams constitutes a super period; there are multiple s in this super period. a b The key candidate moment is calculated as s a b The interference delay at each key candidate time, among which the largest interference delay is s a b The worst case interference delay encountered; i j For a b There are two types of interference. The first is that a b The frame activated before the key candidate moment of s, but due to the delay, a b Activate at the same time; the second is in s a b activation that occurs after the critical candidate moment; Next, in the link The minimum common period of all flows on ,link HT frame m h It needs to be transmitted N times within the minimum common period. The kth transmission is the kth frame. The period is , considering m h Offset and protection band G, G=0 without holding and release mechanism, m h Phase set Through formula (1), we can get: （1） Assume that HT stream frame m c The activation time of m is the key candidate time, and m h As the interference frame of the LT flow being analyzed, equation (2) is expressed by m h Offset Subtract m c Phase Then m h The period modulo, we get m h With m c The phase difference between : （2） Among them, mod is the remainder function; Since other frames do not apply jitter to the time-triggered frame, there is no need to consider jitter in equation (2); Finally, in m c When the activation time is the key candidate time, the transmission duration under duration t is The worst case delay interference caused by HT frames to LT frames , see formula (3): （3）。 7. The scheduling method for time-triggered flows in a time-sensitive network according to claim 6, characterized in that: The analysis process of delay interference caused by LT frames of the same priority level is as follows: Based on the analysis of the phase difference, the phase difference between the interference frame and the analyzed frame is obtained. Then, based on the transmission time required for the interference frame, the interference frame m with the same priority is obtained by formula (4): h For the analyzed frame m i The delay impact : （4） In the formula is m i and m h The phase difference, and m h and m i are of the same priority, is frame m i The path set; when ,and When , formula (4) takes effect.

8. The scheduling method for time-triggered flows in a time-sensitive network according to claim 7, characterized in that: In step S8, the process of determining whether the LT flow on the redundant path can complete scheduling is as follows: HT flow m c The activation time of the kth frame is regarded as the key candidate time, and the link is calculated by iterating the sum of all interferences through formula (5). Upper LT flow m i The delay interference caused by HT flow and LT flow of the same priority : （5） Since the LT flow belongs to the TT flow and cannot be preempted, the iteration starts from RT = ε. Or the iteration ends when the response time exceeds the deadline, where represents the number of iterations, Represents an infinitesimal unit of time; From the stream i The maximum response time is selected from the many key candidate moments of all frames contained in the stream s i In the link The worst-case response time on the path is finally obtained by adding up the worst-case response times of all links on the path. i Worst case end-to-end response time , as shown in (6): （6） For flow i The transmission time of Through formula (6), we can get the LT flow on the path The worst-case end-to-end response time on i The scheduling is considered to have failed.

9. The scheduling method for time-triggered flows in a time-sensitive network according to claim 1, characterized in that: The specific process of step S9 is as follows: Assuming that the arrival process of the flow follows the Poisson distribution, it means that the arrival process of the first data frame of all flow instances conforms to the Poisson distribution. Therefore, it is assumed that one frame is a flow instance. At the same time, since the processing time of each flow by the TSN switch is non-exponential, in order to be practical, the M / G / 1 queuing theory model is used to analyze the starvation probability of the TT flow. The M / G / 1 queuing theory model is adopted, where M represents the arrival process following the Poisson distribution, G represents the service time following the general distribution, and 1 represents a single processing unit; First, according to the arrival rate of TT flow and the switch service rate Calculate the utilization of TT flow : （7） According to the M / G / 1 queuing theory, the average waiting time of high-priority traffic is W T It is expressed as: （8） in, is the second-order moment of the high-priority traffic service time; The average waiting time W of low priority traffic L It consists of two parts: the impact of high-priority traffic on it and its own service waiting time: （9） in is the second-order moment of the arrival rate, utilization, and service time of the high-priority flow, is the arrival rate of low priority flows; Assuming that the waiting time of TT flow follows exponential distribution, D T is the maximum allowed time, then the probability of starvation is calculated as: (10)。

Images