Time-sensitive network flow routing and scheduling method based on centralized control

Abstract

The invention belongs to the technical field of network transmission, and particularly relates to a time-sensitive network flow routing and scheduling method based on centralized control. Comprising the steps that CNC constructs network topology and generates global topology information; the CUC collects demand information of each to-be-transmitted traffic from each source node and divides traffic priorities according to the demand information; the CNC generates an optimal transmission path for each flow according to the global topological information; the CNC distributes a transmission time slot for each flow and dynamically adjusts the transmission time slot; carrying out synchronous distribution on the optimal transmission path and the transmission time slot, optimizing the transmission time slot and controlling the jitter of the transmission time slot; a scheduling table is generated according to the final distribution result of the paths and the time slots, and the CNC distributes the scheduling table to each network device; when the network state is changed, the scheduling table is dynamically adjusted to realize real-time routing and scheduling of the flow; according to the invention, the dynamic management of the flow priority is realized, the utilization rate of network resources is enhanced, and the service quality and high availability of transmission are ensured.

Description

Technical Field

[0001] The present invention belongs to the technical field of network transmission, and in particular relates to a time-sensitive network traffic routing and scheduling method based on centralized control. Background Art

[0002] With the rapid development of intelligent manufacturing and industrial Internet of Things, modern industrial systems have put forward higher requirements for the real-time performance, reliability and resource utilization of networks. Especially in critical tasks that require precise timing control, time-sensitive networks (TSN) have gradually become one of the important technologies to meet the needs of industrial control. By introducing high-precision time scheduling and deterministic routing, TSN achieves low-latency and high-reliability transmission of critical business traffic, thereby supporting the strict requirements of industrial applications for deterministic communication. Traditional TSN scheduling methods are mostly based on distributed control. In complex network scenarios and high-density critical traffic environments, distributed methods have limitations in resource utilization and scheduling flexibility. Faced with uncertain traffic changes and multi-priority business needs, distributed scheduling is difficult to quickly adjust resource allocation, which easily causes uneven utilization of network resources, resulting in increased latency and bandwidth waste. To this end, more and more studies have begun to focus on centralized control models in order to achieve efficient utilization of resources in complex network topologies. Therefore, there is an urgent need for a centralized routing and scheduling method that can not only ensure the deterministic transmission of critical business flows, but also effectively improve the utilization efficiency of remaining resources in the network. Summary of the invention

[0003] In view of the shortcomings of the prior art, the present invention proposes a time-sensitive network traffic routing and scheduling method based on centralized control, the method comprising:

[0004] S1: CNC builds the network topology based on the current network status and generates global topology information;

[0005] S2: CUC collects the demand information of each flow to be transmitted from each source node, and divides the flow priority according to the demand information, including high priority flow, medium priority flow and low priority flow. The flow priority is divided according to the bandwidth demand and delay demand. The high priority flow has the most stringent delay demand, the medium priority flow has a relatively loose delay demand, and the low priority flow tolerates higher delay and jitter.

[0006] S3: CNC generates the optimal transmission path for each traffic based on the global topology information. High-priority traffic is given priority in the path selection process to ensure that its transmission delay is minimized. CNC selects the best path for each priority traffic based on factors such as link bandwidth, latency, and load status. When selecting a path, the path for high-priority traffic will avoid links with high loads and high packet loss rates.

[0007] S4: CNC allocates transmission time slots for each flow and dynamically adjusts the transmission time slots. High-priority flows will be given priority to obtain time slots to ensure that their low latency requirements are met; low-priority flows will be allocated time slots based on the remaining resources and may tolerate higher latency.

[0008] S5: The optimal transmission path and transmission time slot are synchronously allocated, and the transmission delay and jitter are optimized to complete the final allocation of paths and time slots; bandwidth utilization and packet loss rate will be taken into account to ensure efficient transmission of traffic and avoid increased delays caused by insufficient bandwidth or excessive packet loss rate.

[0009] S6: Generate a scheduling table based on the final allocation results of the paths and time slots, and the CNC distributes the scheduling table to each network device. During the generation of the scheduling table, the CNC will refer to the bandwidth utilization and packet loss rate to optimize the scheduling to improve the efficiency of network resource utilization.

[0010] S7: Monitors the network status. When the network status changes, CNC will dynamically adjust the scheduling table based on information such as the link's bandwidth utilization, packet loss rate, and port status, and prioritize the transmission path and time slot of high-priority traffic to ensure stable transmission of critical traffic.

[0011] Preferably, the process of generating global topology information includes:

[0012] S11: CNC scans all nodes and links in the network and generates a topology diagram;

[0013] S12: CNC records the basic attributes of each link, including bandwidth, delay, load status, packet loss rate, and port status;

[0014] S13: The CNC sets a fixed time interval to refresh the link status to obtain the latest link information.

[0015] Preferably, the process of prioritizing traffic according to demand information includes:

[0016] Demand information includes bandwidth demand BW req and delay requirement D req ;

[0017] According to bandwidth requirement BW req , delay requirement D req Prioritize traffic: high priority traffic meets D req ≤D th And BW req ≥BW th , medium priority traffic is D th <D req ≤D max, low priority traffic satisfies D req >D max ; BW th represents the bandwidth requirement threshold, D th Denotes the delay requirement threshold, D max Indicates the maximum delay requirement.

[0018] Preferably, the process of generating an optimal transmission path for each flow includes:

[0019] S31: Generate a set of candidate paths for each flow according to the topology diagram, and each candidate path meets the bandwidth requirement BW of the flow req and delay demand D req ;

[0020] S32: Eliminate candidate paths that do not meet the link load and packet loss rate requirements from the candidate path set;

[0021] S33: Calculate the fitness and use the PSO algorithm to find the optimal transmission path from the candidate path set according to the fitness.

[0022] Furthermore, the formula for calculating fitness is:

[0023] F path =α·L total (P)+β·D total (P)

[0024] Among them, F path represents fitness, L total (P) represents the total path load of the candidate path set P,

[0025] D total (P) represents the total path delay of the candidate path set P; α and β represent the load weight and delay weight.

[0026] Preferably, the process of CNC allocating a transmission time gap for each flow and dynamically adjusting the transmission time slot includes:

[0027] S41: CNC allocates initial transmission time slots for each priority traffic; high-priority traffic has the most stringent latency requirements, while low-priority traffic can tolerate higher latency;

[0028] S42: using the SAC algorithm and optimizing the time slot offset of each flow in descending order of priority to obtain an optimal time slot offset; adjusting the transmission time slot according to the optimal time slot offset to obtain a transmission gap allocated to each flow;

[0029] S43: Calculate the score of each type of traffic based on latency requirements, bandwidth requirements, and jitter tolerance; sort the traffic based on the score; the sorted traffic will be allocated time slots in order of priority to ensure that high-priority traffic has priority in obtaining low-latency transmission resources.

[0030] Preferably, the process of synchronously allocating the optimal transmission path and transmission time slot and optimizing the transmission delay and jitter includes:

[0031] S51: synchronously allocating the optimal transmission path and transmission time slot for each flow in descending order of priority to ensure that the transmission time slot of each flow is arranged within a suitable time window;

[0032] S52: Optimize the transmission time slot according to the delay requirements of the traffic, and reduce the transmission delay of the key traffic by dynamically adjusting the position and length of the time slot.

[0033] S53: Control transmission time slot jitter T jitter , so that it meets the transmission demand target Minimize T jitter =|T slot -T req |, T slot Represents the transmission time slot, T req Indicates the demand time slot.

[0034] Preferably, the time-sensitive network traffic routing and scheduling method further includes: when a link or node fails, the CNC recalculates the backup path; the formula for calculating the backup path is:

[0035]

[0036] Among them, P new Indicates the backup path, L link (i) represents the load of path i, D limit (i) represents the delay of path i, F usa Indicates the set of non-link failures.

[0037] The beneficial effects of the present invention are as follows: the present invention realizes the global optimization configuration of traffic in a centralized control mode by introducing particle swarm optimization (PSO) and soft actor critic (SAC) algorithms. The centralized user configuration manager (CUC) is responsible for collecting and managing the demand information of each business traffic, including bandwidth, delay and jitter tolerance, and classifying the traffic by priority. The central network controller (CNC) generates a set of candidate paths for each type of traffic based on the real-time updated network topology information and link load conditions, and uses the PSO algorithm to select a suitable optimal path from the candidate paths. The SAC algorithm further optimizes the time slot allocation of critical traffic to ensure that high-priority traffic is transmitted in low-delay time slots. Combining the time-aware scheduling (QBV) and time-sensitive scheduling (TAS) mechanisms, delay and jitter are reduced in the synchronous optimization of paths and time slots to ensure the transmission reliability of critical traffic. The present invention realizes the dynamic management of traffic priority, enhances the utilization rate of network resources, and ensures the service quality and high availability of transmission. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] Figure 1 It is a flow chart of the time-sensitive network traffic routing and scheduling method based on centralized control in the present invention;

[0039] Figure 2 It is a schematic diagram of the network topology structure of the present invention;

[0040] Figure 3 This is a diagram of the experience playback and feedback mechanism of TSN network routing and scheduling in the present invention. DETAILED DESCRIPTION

[0041] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0042] This paper proposes a time-sensitive network traffic routing and scheduling method based on centralized control, aiming to solve the contradiction between the deterministic transmission requirements of key business flows and the efficient use of network resources. This method combines the particle swarm optimization (PSO) algorithm and the soft actor critic (SAC) algorithm to achieve global path selection and time slot scheduling optimization, effectively improving the utilization rate of remaining network resources while ensuring low-latency and high-reliability transmission of high-priority traffic; Figure 1 As shown, the method includes:

[0043] S1: CNC builds the network topology according to the current network status and generates global topology information.

[0044] CNC builds the topology of the network, and the collaborative working mode between CNC and CUC is as follows Figure 2 As shown, the CNC (central network controller) systematically scans all nodes and links in the entire network, and constructs a global topology diagram G = (V, E) by obtaining the connection status and link characteristics of the devices, where V is the node set in the network, including switches, routers and other network devices, and E is the link set.

[0045] CNC collects and records the basic attributes of each link, including bandwidth B link (e ij ) is the maximum transmission capacity of the link, delay D link (e ij ) is the data transmission delay from one end of the link to the other end, and the load state L usage (e ij ) is the current usage of the link, the packet loss rate is the ratio of the number of lost data packets to the number of sent data groups, and the port status S port (e ij ) record the connectivity and operation status of the port;

[0046] The CNC sets a fixed time interval to refresh the link status to obtain the latest link information and provide real-time feedback on link status changes. The updated information will be fed back to the central network controller (CNC) and related network devices to enable dynamic routing adjustments and time slot allocation based on the current network status.

[0047] S2: CUC collects the demand information of each flow to be transmitted from each source node and divides the flow priorities according to the demand information, including high priority flow, medium priority flow and low priority flow.

[0048] Traffic is prioritized based on bandwidth requirements and latency requirements. High-priority traffic has the most stringent latency requirements and higher bandwidth requirements, and links and time slots with sufficient bandwidth must be allocated first. Medium-priority traffic has relatively loose latency requirements and more moderate bandwidth requirements. Low-priority traffic tolerates higher latency, but its bandwidth requirements can be appropriately reduced.

[0049] CUC collects the demand information of each flow to be transmitted from each source node, including the bandwidth demand BW req and delay requirement D req Based on these requirements, CUC divides traffic into high priority, medium priority, and low priority. Specifically, high priority traffic meets D req ≤D th And BW req ≥BW th , medium priority traffic is D th <D req ≤D max, low priority traffic satisfies D req >D max ; BW th represents the bandwidth requirement threshold, D th Denotes the delay requirement threshold, D max Indicates the maximum delay requirement.

[0050] S3: CNC generates the optimal transmission path for each flow based on the global topology information.

[0051] Consider the bandwidth utilization, bandwidth demand, and packet loss rate of the link. CNC prioritizes links with sufficient bandwidth and low packet loss rate to ensure that traffic with large bandwidth demand can obtain the required resources and optimize the transmission quality of traffic.

[0052] S31: Generate a candidate path set P for each flow based on the topology graph G=(V,E)

[0053] {p1,p2,...,p n}, each candidate path meets the bandwidth requirement BW of the traffic req and delay demand D req .

[0054] S32: Eliminate candidate paths that do not meet the link load and packet loss rate requirements from the candidate path set.

[0055] S33: Calculate the fitness and use the PSO algorithm to find the optimal transmission path from the candidate path set according to the fitness.

[0056] CNC uses PSO and SAC algorithms to optimize the path and time slot respectively, thus achieving joint optimization. In the candidate path set, CNC uses particle swarm optimization (PSO) algorithm for path selection. The PSO algorithm regards each candidate path as a "particle" and its fitness function is:

[0057] F path =α·L total (P)+β·D total (P)

[0058] Among them, L total (P) = ∑ i∈p L link (i) represents the total path load, D total (P) = ∑ i∈p D link (i) represents the total path delay, L link (i) represents the load of link i, D link (i) represents the delay of link i; α and β represent the load weight and delay weight.

[0059] Each particle is based on its individual optimal value Pbest and the global optimal value g best Update speed and position, expressed as:

[0060]

[0061] in, is the speed of the particle in the tth iteration, is the specific solution or path of the current particle in the path selection space, ω is the inertia weight, c1 and c2 are learning factors, r1 and r2 are random numbers. After multiple iterations, the algorithm finally converges to the optimal path.

[0062] According to the fitness function result after PSO optimization, the candidate path set P cand Sort and select the path P with the best fitness opt =argmin p∈pcand F path as the optimal transmission path.

[0063] S4: CNC allocates transmission time slots to each traffic flow according to the traffic priority and dynamically adjusts the transmission time slots.

[0064] S41: CNC allocates initial transmission time slot T for each priority traffic slot .

[0065] Consider the delay requirement of traffic D req , jitter tolerance J tol and bandwidth requirement BW req High-priority traffic will be given priority in low-latency time slots to ensure that its transmission latency requirements are met, and appropriate time slot lengths are allocated based on bandwidth requirements. Traffic with lower jitter tolerance will also be given priority in time slot allocation to reduce the impact of time slot jitter. Time slot allocation for low-priority traffic will be based on the remaining resources and may tolerate higher latency and jitter.

[0066] S42: Adopt the SAC algorithm and optimize the time slot offset of each flow in descending order of priority to obtain the optimal time slot offset; adjust the transmission time slot according to the optimal time slot offset to obtain the transmission gap allocated to each flow.

[0067] The SAC algorithm optimizes the transmission time slot through reinforcement learning to minimize the delay. In SAC, the actor network is based on the current state s t Calculate action a t , i.e., the time slot offset; the critic network estimates the value of the action and gives a corresponding reward. The reward design takes into account factors such as latency, bandwidth utilization, and jitter.

[0068] After optimization, the information such as status, action, and reward is stored in the experience replay pool. By updating the loss function The allocation of each time slot is optimized to improve the overall performance of the network.

[0069] S43: Calculate the score of each type of traffic, which is comprehensively evaluated by factors such as latency requirement, bandwidth requirement, and jitter tolerance. Sort the traffic according to the score, and the sorted traffic will be allocated time slots in order of priority to ensure that high-priority traffic gets low-latency transmission resources first, and low-priority traffic can be delayed appropriately under high load conditions.

[0070] S5: Synchronously allocate the optimal transmission path and transmission time slot, optimize the transmission time slot and control the transmission time slot jitter to complete the final allocation of the path and time slot.

[0071] like Figure 3 As shown in the figure, the system realizes dynamic adjustment of paths and time slots through a multi-layer feedback mechanism combined with QBV and TAS strategies.

[0072] S51: Combined with the time-aware scheduling (QBV) mechanism, the optimal transmission path and transmission time slot for each flow are synchronously allocated in descending order of priority to ensure that each flow is arranged in an appropriate time window, thereby ensuring that the transmission time slot matches the demand time slot. The demand time slot is the target value for calculating the ideal transmission time slot of the flow, which is used to ensure that the flow completes data transmission on time if the network load permits.

[0073] S52: Optimize the transmission time slot according to the delay requirements of the traffic, and reduce the transmission delay of the key traffic by dynamically adjusting the position and length of the time slot.

[0074] For traffic with low latency requirements, earlier time slots are allocated first and overlap with other traffic is minimized; for traffic with higher latency tolerance, later time slots can be allocated. Use SAC to optimize time slot offsets to further reduce latency and improve traffic transmission efficiency.

[0075] S53: Control transmission time slot jitter T jitter , so that it meets the transmission demand target Minimize T jitter =|T slot -T req |, T slot Represents the transmission time slot, T req Indicates the demand time slot.

[0076] S6: Generate a scheduling table based on the final allocation results of the paths and time slots, and the CNC distributes the scheduling table to each network device.

[0077] When generating the scheduling table, CNC will refer to the bandwidth utilization and packet loss rate to optimize the scheduling to improve the efficiency of network resource utilization.

[0078] After completing the path and time slot optimization, the system generates a scheduling table S sched , including the best path P for each flow opt , Priority P k and the corresponding time slot T slot ; Each item in the schedule is S i = {P opt ,T slot, P k}, indicating the flow rate F i On path P opt and time slot T slot transmission on .

[0079] S7: Monitor the network status. When the network status changes, CNC will dynamically adjust the scheduling table based on the link bandwidth utilization, packet loss rate, port status and other information to achieve real-time routing and scheduling of traffic. It will give priority to adjusting the transmission path and time slot of high-priority traffic to ensure stable transmission of critical traffic.

[0080] CNC and CUC monitor the link load L of the network in real time usage (e ij ), topology G, and traffic change F. When the network status changes significantly (such as link failure, new traffic joining, etc.), CNC and CUC will repeat steps S1 to S6 according to the latest network status to recalculate the optimal transmission path and time slot allocation scheme. CUC can extract historical experience from the experience playback pool and optimize time slot offset to reduce delay and jitter, and quickly respond to the time slot optimization needs of traffic. At the same time, bandwidth demand BW adjusted =

[0081] f(L ratio ,Q alloc ) will be recalculated, and CNC will dynamically adjust path selection and priority allocation based on bandwidth requirements, link load, and time slot requirements to ensure that high-priority traffic obtains sufficient network resources and efficiently utilizes the remaining resources.

[0082] The time-sensitive network traffic routing and scheduling method also includes: when a link or node fails, the CNC recalculates the backup path; when calculating the backup path, it is ensured that the backup path can meet the bandwidth requirements and its performance is not lower than the current path. The formula for calculating the backup path is:

[0083]

[0084] Among them, P new Indicates the backup path, L link(i) represents the load of path i, D limit represents the delay of path i, F set represents the link failure set F set , E represents the link set, F usa =EF set , F usa Indicates the set of non-link failures.

[0085] In summary, by introducing particle swarm optimization (PSO) and soft actor critic (SAC) algorithms, global optimal configuration of traffic is achieved in centralized control mode. The centralized user configuration manager (CUC) is responsible for collecting and managing the demand information of each business traffic, including bandwidth, delay and jitter tolerance, and classifying the traffic by priority. The central network controller (CNC) generates a set of candidate paths for each type of traffic based on real-time updated network topology information and link load conditions, and uses the PSO algorithm to select a suitable optimal path from the candidate paths. The SAC algorithm further optimizes the time slot allocation of critical traffic to ensure that high-priority traffic is transmitted in low-delay time slots. Combined with time-aware scheduling (QBV) and time-sensitive scheduling (TAS) mechanisms, delay and jitter are reduced in the synchronous optimization of paths and time slots to ensure the transmission reliability of critical traffic. The present invention realizes dynamic management of traffic priority, enhances the utilization of network resources, and ensures the service quality and high availability of transmission.

[0086] The above embodiments further illustrate the purpose, technical solutions and advantages of the present invention in detail. It should be understood that the above embodiments are only preferred implementation modes of the present invention and are not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made to the present invention within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A time-sensitive network traffic routing and scheduling method based on centralized control, characterized in that: include: S1: CNC builds the network topology based on the current network status and generates global topology information; S2: CUC collects the demand information of each flow to be transmitted from each source node and divides the flow priorities according to the demand information, including high priority flow, medium priority flow and low priority flow. S3: CNC generates the optimal transmission path for each flow based on the global topology information. S4: CNC allocates transmission time slots to each traffic flow according to the traffic priority and dynamically adjusts the transmission time slots. S5: Synchronously allocate the optimal transmission path and transmission time slot, optimize the transmission time slot and control the transmission time slot jitter to complete the final allocation of the path and time slot; S6: Generate a scheduling table based on the final allocation results of the paths and time slots, and the CNC distributes the scheduling table to each network device; S7: Monitor the network status. When the network status changes, CNC dynamically adjusts the scheduling table to achieve real-time routing and scheduling of traffic.

2. A time-sensitive network traffic routing and scheduling method based on centralized control according to claim 1, characterized in that: The process of generating global topology information includes: S11: CNC scans all nodes and links in the network and generates a topology diagram; S12: CNC records the basic attributes of each link, including bandwidth, delay, load status, packet loss rate, and port status; S13: The CNC sets a fixed time interval to refresh the link status to obtain the latest link information.

3. The method for time-sensitive network traffic routing and scheduling based on centralized control according to claim 1, characterized in that: The process of prioritizing traffic based on demand information includes: Demand information includes bandwidth demand BW req and delay requirement D req ; According to bandwidth requirement BW req , delay requirement D req Prioritize traffic: high priority traffic meets D req ≤D th And BW req ≥BW th , medium priority traffic is D th <D req ≤D max , low priority traffic satisfies D req >D max ; BW th represents the bandwidth requirement threshold, D th Denotes the delay requirement threshold, D max Indicates the maximum delay requirement.

4. The method for time-sensitive network traffic routing and scheduling based on centralized control according to claim 1, characterized in that: The process of generating the optimal transmission path for each flow includes: S31: Generate a set of candidate paths for each flow according to the topology diagram, and each candidate path meets the bandwidth requirement BW of the flow req and delay demand D req ; S32: Eliminate candidate paths that do not meet the link load and packet loss rate requirements from the candidate path set; S33: Calculate the fitness and use the PSO algorithm to find the optimal transmission path from the candidate path set according to the fitness.

5. A time-sensitive network traffic routing and scheduling method based on centralized control according to claim 4, characterized in that: The formula for calculating fitness is: F path =α·L total (P)+β·D total (P) Among them, F path represents fitness, L total (P) represents the total path load of the candidate path set P, D total (P) represents the total path delay of the candidate path set P; α and β represent the load weight and delay weight.

6. A time-sensitive network traffic routing and scheduling method based on centralized control according to claim 1, characterized in that: The process of CNC allocating transmission time slots for each flow and dynamically adjusting the transmission time slots includes: S41: CNC allocates initial transmission time slots for each priority traffic; high-priority traffic has the most stringent latency requirements, while low-priority traffic can tolerate higher latency; S42: using the SAC algorithm and optimizing the time slot offset of each flow in descending order of priority to obtain an optimal time slot offset; adjusting the transmission time slot according to the optimal time slot offset to obtain a transmission gap allocated to each flow; S43: Calculate the score of each type of traffic based on latency requirements, bandwidth requirements, and jitter tolerance; sort the traffic based on the score; the sorted traffic will be allocated time slots in order of priority to ensure that high-priority traffic has priority in obtaining low-latency transmission resources.

7. The method for time-sensitive network traffic routing and scheduling based on centralized control according to claim 1, characterized in that: The process of synchronously allocating the optimal transmission path and transmission time slot and optimizing the transmission delay and jitter includes: S51: synchronously allocating the optimal transmission path and transmission time slot for each flow in descending order of priority to ensure that the transmission time slot of each flow is arranged within a suitable time window; S52: Optimize the transmission time slot according to the delay requirements of the traffic, and reduce the transmission delay of the key traffic by dynamically adjusting the position and length of the time slot. S53: Control transmission time slot jitter T jitter , so that it meets the transmission demand target Minimize T jitter =|T slot -T req |, T slot Represents the transmission time slot, T req Indicates the demand time slot.

8. The method for time-sensitive network traffic routing and scheduling based on centralized control according to claim 1, characterized in that: The time-sensitive network traffic routing and scheduling method further includes: when a link or node fails, the CNC recalculates the backup path; the formula for calculating the backup path is: Among them, P new Indicates the backup path, L link (i) represents the load of path i, D limit (i) represents the delay of path i, F usa Indicates the set of non-link failures.

Images