Wide area network mixed flow scheduling method and system based on CSQF and electronic equipment

Abstract

The invention provides a CSQF-based wide area network mixed flow scheduling method and system and an electronic device, and the method combines the scheduling demands of a TT flow and an AVB flow through the reasonable division of a periodic resource block, guarantees that the timeliness requirement of the TT flow is satisfied, and provides a flexible bandwidth allocation strategy for the AVB flow. And a dynamic period scaling mechanism is introduced to adapt to the change of the network load, and conflicts among different traffic types are reduced through a sliding window preemption strategy, so that the overall scheduling effect of the mixed traffic is optimized.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of wide area network environment time sensitive network mixed traffic scheduling, and more particularly to a wide area network mixed traffic scheduling method and system based on cycle specified queuing and forwarding CSQF and an electronic device. BACKGROUND

[0002] In emerging fields such as industrial internet, intelligent manufacturing and autonomous driving, time sensitive network (TSN) has become a key technology to support real-time communication due to its high precision delay control and large-scale data transmission capability. However, the deployment of TSN in wide area networks faces significant challenges. Traditional scheduling mechanisms rely on the zero propagation delay assumption of local area networks and are limited by the seven-hop range constraint of IEEE 802.1D protocol, making it difficult to adapt to the delay characteristics of long-distance networks. In addition, the lack of time synchronization accuracy, high traffic dynamics, and intense resource competition in wide area networks further limit the large-scale application of TSN.

[0003] To expand the wide area applicability of TSN, the IETF DetNet working group proposes a cycling queuing and forwarding (CSQF) mechanism, which divides equal-length cycles and labels the cycle tag in the data packet header to achieve deterministic control of end-to-end delay (theoretical jitter does not exceed 2 cycles). However, CSQF still has the following limitations: (1) the cycle label stack only supports single-flow linear path scheduling, making it difficult to cope with multi-flow mixed scenarios; (2) scheduling tasks have NP-hard characteristics, with high modeling complexity; (3) the differentiated quality of service (QoS) requirements of mixed traffic exacerbate resource allocation conflicts. For example, time triggered (TT) flows require strict periodicity and low latency guarantees, while audio video bridging (AVB) flows focus more on bandwidth utilization, and traditional static scheduling algorithms cannot accommodate both characteristics.

[0004] Mixed traffic includes three different types of time triggered (TT) flows, audio video bridging (AVB) flows, and best effort (BE) flows. TT flows have strict periodicity and timeliness requirements, while AVB flows have certain requirements for bandwidth and transmission delay. Existing research has made some progress in optimizing mixed traffic scheduling. For example, K-shortest path and GRASP meta-heuristic algorithm are combined to optimize TT / AVB flow joint scheduling; a hybrid mechanism (HSTC) based on TAS and CQF is used to improve resource utilization through time slot adjustment and offset planning; probability coloring algorithm, group weighted round robin strategy, and improved particle swarm optimization algorithm (PSO-OMA) are designed to reduce latency and conflicts.

[0005] However, the above methods are mostly focused on local area network scenarios, and have high algorithm complexity, which is difficult to adapt to the scale requirements of wide area networks. SUMMARY

[0006] The present application aims to solve the problems of high method complexity, difficulty in coping with multi-flow mixed scenarios, resource allocation conflicts, and difficulty in adapting to the scale requirements of wide area networks in the prior art of time-sensitive network mixed traffic scheduling in wide area network environments, and provides a wide area network mixed traffic dynamic collaborative scheduling method based on cycle specified queuing and forwarding (CSQF). This method combines the scheduling requirements of TT flows and AVB flows by reasonable division of cycle resource blocks, ensures that the timeliness requirements of TT flows are met, and provides flexible bandwidth allocation strategies for AVB flows. Moreover, this method introduces a dynamic cycle stretching mechanism to adapt to changes in network load, and reduces conflicts between different traffic types through a sliding window preemption strategy, optimizing the overall scheduling effect of mixed traffic.

[0007] The first aspect of the present application discloses a wide area network mixed traffic scheduling method based on CSQF, comprising:

[0008] Step S1: The TSN switch discovers the TSN network topology according to the link discovery protocol, and abstracts the TSN network topology into a network directed graph through a network modeling method, obtains a network node set, link bandwidth, transmission delay, and characteristics of a set of traffic to be scheduled; the network directed graph is represented as G=(V, E), where V represents the network node set, including TSN switches and terminal devices ES={es1, es2, …, es n}. E represents a physical link set, where the nodes V a and the physical link between V b correspond to two one-way links [V a , V b ] and [V b , V a ];

[0009] Step S2: Based on the obtained network node set, link bandwidth, transmission delay, and characteristics of the set of traffic to be scheduled, perform traffic classification and feature analysis to obtain the traffic type; construct a mixed traffic scheduling model based on CSQF, allocate independent queue resources for each traffic type using CSQF, determine the traffic type, check whether it is a TT flow or an AVB flow, if it is a TT flow, jump to step S3; if it is an AVB flow, go to step S4;

[0010] Step S3: Run the ILP-TT flow scheduling method based on the shortest weighted routing, and go to step S5;

[0011] Step S4: running the AVB flow scheduling method based on dynamic cycle adjustment and elastic resource block allocation, and entering step S5;

[0012] Step S5: checking whether there is a conflict in time slot allocation, checking whether the resource allocation conflicts with the successfully scheduled TT flow; if there is a resource conflict, postponing the AVB flow to the next safe time window, entering step S4 to re-adjust the resource allocation strategy; if there is no conflict, directly entering step S6;

[0013] Step S6: generating and outputting the final scheduling result, including: the transmission path of each flow, the accurate time slot allocation table of each flow in the link, the resource occupation rate and the real-time guarantee index.

[0014] According to the CSQF-based wide-area network mixed flow scheduling method of the first aspect of the present application, the step S2 of constructing a mixed flow scheduling model according to the CSQF is specifically:

[0015] The CSQF divides the sending time of the output port into equal time intervals, and defines the queue resource block at different time periods of each port. The mixed flow scheduling model is constructed according to the CSQF mechanism, and the resource block type is allocated to different flow types for flow scheduling.

[0016] According to the CSQF-based wide-area network mixed flow scheduling method of the first aspect of the present application, the step S3 specifically includes:

[0017] Step S31: for each TT flow, the optimal path is selected by the shortest weighted routing algorithm; the selection of the path considers the physical structure of the network topology, and is weighted in combination with the weight of the link;

[0018] Step S32: the time slot is solved by using ILP, and the sending time interval of the output port is T slot , the cycle label of the current path is allocated to each node V a , the link propagation delay LD (Va,Va+1) between the node V a+1 and the next node V a is calculated, and the cycle offset amount is calculated according to the link propagation delay.

[0019] If the current node V a is the source node, the initial cycle label <T i , Q i > is allocated, wherein T i is the cycle identifier, and Q i is the queue identifier; if the node V a is not the source node, the cycle label T a of the previous hop is inherited.a-1 + Δ a-1 , and the queue identifier Q a of the previous node is used to assign a queue position for the current node according to Q a-1 = (Q TT + 1) mod M.

[0020] After the periodic tag is assigned, the capacity C a (Q TT ) of the current queue is checked to see if it can meet the requirement of the TT flow, i.e., if Q a (T a , Q (Va,Va+1) ) + LD TT > C a (Q TT ), then the periodic offset and the periodic tag are recalculated.

[0021] Step S33: Update the resource reservation matrix R i according to the calculated periodic tag and the queue identifier, save the information of the flow f elastic in the corresponding queue position, and update the information of the successfully scheduled flow in the successfully scheduled TT flow set .

[0022] According to the CSQF-based wide-area network mixed flow scheduling method of the first aspect of the present application, the step S4 is specifically:

[0023] Step S41: Calculate the network load factor according to formula (1), and then obtain the dynamically adjusted dynamic periodic length T j according to formula (2) to adapt to the current network state: when the network load is large, the period is extended to ensure that the AVB flow can obtain sufficient bandwidth in the heavy load condition; when the network load is small, the period is shortened to improve the resource utilization rate.

[0024]

[0025] wherein FB is the fixed resource block, EB is the elastic resource block, a is the periodic correction factor, a and b are the load setting limits, and Load is the network load.

[0026] Step S42: Calculate the required resource block number N block for the AVB flow according to formula (3), and use the sliding window mechanism to find available elastic resource blocks in the future T periods: if enough available resource blocks are found in the sliding window, the resource is allocated to the AVB flow, and the resource allocation matrix R AVB; if there are not enough resource blocks available, trigger the token bucket throttling mechanism, i.e. adjust the bandwidth requirement of the AVB flow, through the token bucket quota, and recalculate the required number of resource blocks N j ;

[0027]

[0028] wherein, denotes the minimum bandwidth of the AVB flow;

[0029] Check if the resource allocation conflicts with the successfully scheduled TT flows, if there is a resource conflict, the AVB flow will be postponed to the next safe time window;

[0030] Step S43: After completing the scheduling, broadcast the updated periodic T elastic and resource allocation matrix R AVB to all network nodes through the centralized network architecture, to ensure the coordination of the entire network.

[0031] The CSQF-based wide area network mixed traffic scheduling method according to the first aspect of the present application, step S1 further comprises:

[0032] The set of traffic to be scheduled F is expressed as: F = {F TT , F AVB}, the TT flow can be represented by a seven-tuple:

[0033]

[0034] wherein, denote the source node and the destination node of the TT flow respectively, denotes the period of the TT flow, denotes the packet size, denotes the deadline, denotes the path information of the flow, denotes the offset time of the flow;

[0035] The AVB flow can be represented by a five-tuple:

[0036]

[0037] wherein denote the source node and the destination node of the AVB flow respectively, denotes the bandwidth requirement of the flow, denotes the deadline of the flow, denotes the path information of the flow.

[0038] The CSQF-based wide area network mixed traffic scheduling method according to the first aspect of the present application, step S3 further comprises:

[0039] The ILP-TT flow scheduling method based on the shortest weighted routing needs to satisfy the following constraint conditions:

[0040] (1) Strict deadline constraint

[0041] The end-to-end delay f from the source node to the destination node i_delay Strictly less than or equal to the deadline That is,

[0042]

[0043] Where f i_delay represents the end-to-end delay of the flow, represents the maximum allowed delay of the flow, i.e. the deadline;

[0044] (2) Period lower bound constraint

[0045] The period lower bound constraint must ensure that all data packets stored in the queue can be scheduled within its duration, and the constraint condition is:

[0046]

[0047] Where f i_proc represents the processing delay, f i_queue represents the queuing delay, f i_trans represents the transmission delay, L represents the maximum capacity of the queue, MTU is the maximum transmission unit of the data packet size, and BW is the link capacity;

[0048] (3) Period upper bound constraint

[0049] The upper bound constraint of the period length ensures that the period of all TT flows in the system can be divided, and the upper bound is consistent with the greatest common divisor of the transmission periods of all flows;

[0050] (4) Offset time constraint

[0051] In the offset time constraint, the offset amount, i.e. the transmission time from the first switch of the source node to the destination node, cannot exceed the period of the flow, and the TT flow should be completely forwarded within the current transmission period, i.e.

[0052]

[0053] Where, is the offset time of the TT flow, is the data packet size of the TT flow, T slot is the transmission time interval of the output port;

[0054] (5) Bounded queue length constraint

[0055] In the bounded queue length constraint, the number of buffered TT flow packets in a single queue cannot exceed the total capacity L of the queue, preventing packet loss and ensuring that each fixed resource block queue is not overloaded in any period.

[0056] According to the CSQF-based wide-area network mixed traffic scheduling method of the first aspect of the application, the step S4 further comprises:

[0057] The AVB flow scheduling method based on dynamic period adjustment and elastic resource block allocation needs to meet the following constraint conditions:

[0058] (1) Bandwidth constraint:

[0059] The minimum bandwidth of the AVB flow needs to be guaranteed during transmission In the elastic resource block, continuous or discrete period blocks are allocated, the number of which can be obtained by bandwidth demand conversion calculation, wherein, T slot is the transmission time interval of the output port:

[0060]

[0061] (2) Elastic resource block constraint:

[0062] Define the resource block allocation matrix R(t), R(t, m) = 1 indicates that the queue m is occupied at period t, otherwise 0;

[0063] (3) Preemption window constraint:

[0064] The AVB flow is only allowed to preempt the elastic resource in the future limited period through the sliding window strategy, and W is the preemption window size, which is the number of periods, if the current period is t, the AVB flow can only preempt the EB resource block in the range of [t, t+W];

[0065] (4) Delay constraint:

[0066] The maximum delay of the AVB flow from entering the queue to being scheduled needs to meet the upper limit, which prevents low-priority traffic from excessive queuing, and it cannot exceed the maximum allowed delay of the AVB flow; combined with the queue mapping rule, the AVB flow is inserted into the rear queue, which will obtain a larger queuing delay: the number of waiting periods x T slot .

[0067] The second aspect of the application discloses a CSQF-based wide-area network mixed traffic scheduling system, which comprises a processing unit configured to perform the steps of the CSQF-based wide-area network mixed traffic scheduling method of the first aspect.

[0068] The third aspect of the present application discloses an electronic device, comprising a memory and a processor, the memory stores a computer program, and the processor is used to execute the program in the memory to realize the CSQF-based wide area network mixed traffic scheduling method of the first aspect.

[0069] The fourth aspect of the present application discloses a computer readable storage medium, which stores the CSQF-based wide area network mixed traffic scheduling method of the first aspect.

[0070] In summary, the present application proposes a dynamic collaborative scheduling scheme based on cycle-specific queuing and forwarding (CSQF) mechanism for wide area network mixed traffic scheduling problem. By designing a deterministic scheduling algorithm based on integer linear programming (ILP) for TT flow, combined with fixed resource block isolation mechanism, the strict delay constraint of time-sensitive flow is guaranteed; at the same time, the dynamic elastic cycle adjustment strategy and sliding window preemption mechanism are introduced, which optimizes the bandwidth utilization and resource fragmentation problem of AVB flow, and provides theoretical support and practical reference for efficient collaborative transmission of mixed traffic in wide area network. BRIEF DESCRIPTION OF DRAWINGS

[0071] In order to more clearly illustrate the specific embodiments of the present application or the technical solutions in the prior art, the drawings needed in the specific embodiments or prior art description will be briefly introduced below. Obviously, the drawings in the following description are some embodiments of the present application, and those skilled in the art can also obtain other drawings according to these drawings without creative labor.

[0072] Figure 1 The figure is a schematic diagram of resource block division and isolation mechanism based on the present application;

[0073] Figure 2 The figure is a mixed traffic scheduling example diagram of TT flow and AVB flow based on the embodiment of the present application;

[0074] Figure 3 The figure is a collaborative transmission flow chart of TT flow and AVB flow in the embodiment of the present application;

[0075] Figure 4 The figure is a mixed traffic scheduling method flow chart of the present application. DETAILED DESCRIPTION

[0076] In order to make the purposes, technical solutions and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only some of the embodiments of the present application, but not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative efforts fall within the scope of protection of the present application.

[0077] The present scheme combines the scheduling requirements of TT flow and AVB flow by reasonable division of periodic resource blocks, ensures that the timeliness requirement of TT flow is met, and provides a flexible bandwidth allocation strategy for AVB flow. The algorithm introduces a dynamic period scaling mechanism to adapt to changes in network load, and reduces conflicts between different traffic types through a sliding window preemption strategy, optimizing the overall scheduling effect of mixed traffic.

[0078] The present application provides a wide area network mixed traffic scheduling method based on CSQF, which comprises the following steps:

[0079] Step S1: The TSN switch discovers the TSN network topology according to the link discovery protocol, and abstracts the TSN network topology into a network directed graph through a network modeling method, obtains a network node set, a link bandwidth, a transmission delay and a to-be-scheduled traffic set characteristics; the network directed graph is represented as G=(V, E), wherein V represents the network node set, including TSN switches and terminal devices ES={es1, es2, …, es n}. E represents a physical link set, wherein the physical link between nodes V a and V b corresponds to two one-way links [V a , V b ] and [V b , V a ].

[0080] The TSN switch discovers the TSN network topology according to the link discovery protocol, and abstracts the TSN network into a network directed graph through a network modeling method. The time-sensitive network is abstracted into a directed graph G=(V, E), wherein V represents a network node set, including TSN switches and terminal devices ES={es1, es2, …, es n}. E represents a physical link set, since the TSN is established on a full-duplex Ethernet, the physical link between nodes V a and V b corresponds to two one-way links [V a , V b ] and [V b , V a ].

[0081] Step S2: Based on the obtained network node set, link bandwidth, transmission delay and the characteristics of the set of traffic to be scheduled, traffic classification and feature analysis are performed to obtain traffic types; a mixed traffic scheduling model is constructed according to the CSQF, the CSQF is used to allocate independent queue resources for each traffic type, by judging the traffic type, it is checked whether it is a TT flow or an AVB flow, if it is judged as a TT flow, it jumps to step S3; if it is judged as an AVB flow, it enters step S4.

[0082] The network traffic is divided into three categories: TT flow, AVB flow and BE flow, where the BE flow is a best-effort transmission flow, which is not considered in this method, and only the scheduling of TT flow and AVB flow is considered. The set of traffic to be scheduled, the set of TT flow and the set of AVB flow are represented, and the network modeling of the system is completed.

[0083] The system modeling of the mixed traffic scheduling model is performed. The CSQF divides the sending time of the output port into a series of equal time intervals, and defines the queue resource block in each port at different time periods. The mixed traffic scheduling model is designed according to the CSQF mechanism, and the resource block type is allocated to different traffic types for traffic scheduling.

[0084] After the system model is built, the specific scheduling algorithm design should be performed to balance the transmission efficiency and resource allocation of TT flow and AVB flow. In this paper, the scheduling scheme based on the cycle specified queuing and forwarding (CSQF) mechanism is proposed to perform accurate resource allocation and dynamic scheduling strategy, which guarantees the timeliness of TT flow and maximizes the bandwidth utilization of AVB flow, and finally realizes the efficient collaborative transmission of the two types of traffic.

[0085] Set different types of traffic transmission constraints. For the design of scheduling algorithm, the corresponding scheduling algorithm is designed according to different traffic types. For TT flow, an integer linear programming (ILP) based TT flow scheduling algorithm is designed, and for AVB flow, an AVB flow scheduling optimization algorithm based on dynamic elastic cycle is designed, and the two types of algorithms are analyzed in detail.

[0086] The network traffic is divided into three categories: TT flow, AVB flow and BE flow, where the BE flow is a best-effort transmission flow, which is not considered in this method, and only the scheduling of TT flow and AVB flow is considered. Therefore, the set of traffic to be scheduled F is represented as: F = {F TT , F AVB}, if f i ∈ F TT , i.e. belongs to the TT flow in the network traffic, it is represented as a seven-tuple

[0087]

[0088] (3) where, denote the source and destination nodes of a TT flow, respectively, denote the period of a TT flow, denote the packet size, denote the deadline, denote the path information of a flow, denote the offset time of a flow; if f j ∈ F AVB denote the five-tuple of an AVB flow if it belongs to the network traffic,

[0089]

[0090] wherein denote the source and destination nodes of an AVB flow, respectively, denote the bandwidth requirement of a flow, denote the deadline of a flow, denote the path information of a flow.

[0091] The hybrid traffic scheduling model is used to model the system. The CSQF divides the transmission time of the output port into a series of equal time intervals, each interval is called a period T slot . In multiple queues participating in scheduling, only one queue is the transmission queue (Tx) in each period, and the rest are reception queues (Rx). The waiting time is different according to the distance from the transmission queue. The queue resource block of each port at different time periods can be defined as wherein T(t) is a time slice in a specific period sequence, and (Va, Vb) refers to the link between two nodes in the network.

[0092] The hybrid traffic scheduling model is designed based on the cycle-specific queuing and forwarding (CSQF) mechanism, which needs to consider the quality of service (Qos) required by different types of traffic. The resources are divided into two types: fixed blocks (FB) and elastic blocks (EB), which are allocated to different traffic types. Among them, the fixed resource block (FB) is used to guarantee the deterministic transmission of time-sensitive (TT) flow, and the elastic resource block (EB) is allocated to bandwidth-sensitive traffic (i.e. AVB flow).

[0093] When the TT flow enters the queue, it is preferentially entered into the reception queue close to the transmission queue, so that it can wait for a smaller number of T sloti.e. less latency, which meets its higher latency requirement. While AVB flows, due to their lower priority, have less stringent latency requirements and are received by the back queue. The capacity of the queue is L, when a certain amount of data packets are accumulated, the current queue cannot continue to receive, at this time, the queue needs to be selected behind and has capacity, otherwise, overflow will cause packet loss.

[0094] In the wide area network environment, the scheduling demand of mixed traffic mainly reflects how to balance the transmission efficiency and resource allocation of TT flow and AVB flow. TT flow usually requires strict latency control, while AVB flow pays more attention to bandwidth utilization. Under the driving of these demands, the traditional scheduling method is difficult to meet the different characteristics of the two types of traffic at the same time, especially in the case of limited resources. In order to meet the needs of the two types of traffic and realize the transmission of mixed traffic, this chapter proposes a scheduling scheme based on the cycle specified queuing and forwarding (CSQF) mechanism. Through accurate resource allocation and dynamic scheduling strategy, the scheme can guarantee the efficient collaborative transmission of the two types of traffic. The CSQF mechanism allocates independent queue resources for each traffic type, and on the basis of dynamic adjustment, guarantees the timeliness of TT flow and maximizes the bandwidth utilization of AVB flow.

[0095] In order to achieve this goal, the following optimization objectives are set:

[0096] (1) Ensure that all TT flows can complete end-to-end transmission within their deadline. That is, TT flow must transmit all data packets on time within the specified period, and the delay should not exceed the set delay.

[0097] (2) Maximize the number of successfully scheduled AVB flows, and schedule as many AVB flows as possible to meet their bandwidth requirements. AVB flow is usually tolerant to latency, so the scheduling of AVB flow should make full use of the remaining bandwidth resources as much as possible without affecting the transmission of TT flow. Define S n S identifies whether the AVB flow is successfully scheduled, as shown in equation (3). And further define the optimization objective, as shown in equation (4):

[0098]

[0099] Maximize∑S n (4)

[0100] In order to ensure that the scheduling process meets the dual demands of real-time and bandwidth utilization, a series of constraint conditions need to be met. The following are the main constraint conditions for different traffic types:

[0101] TT flow constraint conditions:

[0102] (1) Strict deadline constraint:

[0103] All data packets of each TT flow must be transmitted within its specified latency limit. For TT flow f i , the end-to-end latency f i_delay from the source node to the destination node must be strictly less than or equal to the deadline latency , i.e. all TT flows must satisfy this latency constraint:

[0104]

[0105] where f i_delay denotes the end-to-end latency of the flow, denotes the maximum allowed latency of the flow, i.e. the deadline latency.

[0106] (2) Period lower bound constraint:

[0107] Data packets of each TT flow must be transmitted within its scheduled period, ensuring that the data packets arrive at the destination on time and meet their latency requirements. The lower bound must ensure that all data packets stored in the queue can be scheduled within its duration. This time includes the processing latency f i_proc , the queuing latency f i_queue and the transmission latency f i_trans . The lower bound constraint is given by:

[0108]

[0109] where L denotes the maximum capacity of the queue, MTU is the maximum transmission unit of the data packet size, and BW is the link capacity.

[0110] (3) Period upper bound constraint:

[0111] To ensure the stability of the system while allowing all TT flows to be effectively scheduled, the upper bound of the period length should guarantee that the periods of all TT flows in the system can be evenly divided, and the upper bound is consistent with the greatest common divisor of the transmission periods of all flows. The upper bound constraint is given by:

[0112]

[0113] where T denotes the period of the TT flow. Typically, T slot will be greater than its minimum value and is a number that needs to be divisible by the super period (calculated as shown in equation 8), as shown in equation (9).

[0114]

[0115] T Hypercycle mod T slot = 0 (9)

[0116] (4) Offset time constraint:

[0117] The bias determines the initial period of each flow transmission time, i.e. the time when the first switch from the source node to the terminal node starts transmission, which cannot exceed the period of the flow. Moreover, TT flows should be forwarded in the current transmission period to avoid flow interference that may occur in the adjacent transmission period. The formula is expressed as

[0118]

[0119] (5) Bounded queue length constraint:

[0120] The number of TT flow packets cached in a single queue cannot exceed the total capacity L of the queue to prevent overflow and packet loss. Since TT flows use fixed resource blocks (FB), their queue capacity needs to be controlled independently. This constraint ensures that each fixed resource block queue is not overloaded in any period, and its expression is:

[0121]

[0122] AVB flow constraint conditions:

[0123] (1) Bandwidth constraint:

[0124] The minimum bandwidth of the AVB flow needs to be guaranteed during transmission It is required to allocate continuous or discrete period blocks in the elastic resource block, the number of which can be obtained by bandwidth demand conversion calculation, and the mathematical expression is:

[0125]

[0126] (2) Elastic resource block constraint:

[0127] Since the transmission of TT flows needs to be prioritized, while achieving dynamic resource acquisition of AVB flows, they cannot occupy the fixed resource blocks of TT flows to avoid interfering with the deterministic transmission of time-sensitive flows. Define the resource block allocation matrix R(t), where R(t,m) = 1 indicates that queue m is occupied at period t, otherwise 0. Let M TT be the set of fixed resource blocks dedicated to TT flows, then the constraint is:

[0128]

[0129] This constraint ensures that the elastic resource block (EB) is completely isolated from the fixed resource block (FB).

[0130] (3) Preemption window constraint:

[0131] AVB flow only allows preemption of elastic resources in a limited period in the future through a sliding window strategy, avoiding long-term occupation affecting future reserved resources of TT flow. Let W be the preemption window size (its unit is the number of cycles), if the current cycle is t, then AVB flow can only preempt EB resource blocks in the range of [t, t+W]. The number of EB resource blocks allocated to the flow in cycle t is denoted as Nt.

[0132]

[0133] (4) Delay constraint:

[0134] The maximum delay of AVB flow from entering the queue to being scheduled needs to meet the upper limit to prevent low-priority traffic from excessive queuing. Let The maximum allowed delay of AVB flow is denoted as Dmax, and the constraint is:

[0135]

[0136] Combined with the queue mapping rule, AVB flow is inserted into the rear queue, which will obtain a larger queuing delay, which is: waiting cycle number x T slot .

[0137] Step S3: Run the ILP-TT flow scheduling method based on the shortest weighted routing, and enter step S5;

[0138] Step S4: Run the AVB flow scheduling method based on dynamic cycle adjustment and elastic resource block allocation, and enter step S5;

[0139] Step S5: Check if there is a conflict in time slot allocation, check if the resource allocation conflicts with the successfully scheduled TT flow; if there is a resource conflict, postpone the AVB flow to the next safe time window, enter step S4, and adjust the resource allocation strategy; if there is no conflict, directly enter step S6.

[0140] Different scheduling algorithms are designed for different traffic types. The core idea of the ILP-TT flow scheduling algorithm based on integer linear programming (ILP) is to ensure that time-sensitive flow (TT) can achieve high-precision end-to-end transmission in a wide area network, while considering low complexity and high reliability of scheduling. This algorithm combines the shortest weighted routing selection and cycle scheduling mechanism, solves the delay fluctuation and resource competition problems of time-sensitive flow in the transmission process through link delay compensation, cycle label allocation and queue capacity check. The ILP-TT flow scheduling scheme based on the shortest weighted routing specifically includes:

[0141] First, for each TT flow, the algorithm selects the optimal path using a shortest weighted routing algorithm. This path selection considers not only the physical structure of the network topology but also the weights of the links (such as link delay) to ensure that the selected path provides low latency and high bandwidth utilization during transmission.

[0142] The next step is to solve for the time slots using ILP. Periodic labels are assigned to each node V on the current path. The time slots from node V are then calculated. a To the next node V a+1 Link propagation delay LD (Va,Va+1) Then calculate the period offset based on the link propagation delay.

[0143] If the current node V a If it is the source node, then assign an initial periodic label <T. i Q i >where T i It is a periodic identifier, Q i It is a queue identifier. If node V a If it is not the source node, then it inherits the periodic label T from the previous hop. a =T a-1 +Δ a-1 And based on the queue identifier Q of the previous node a =(Q a-1 +1)modM assigns a queue position to the current node.

[0144] After completing the periodic label allocation, the algorithm checks the current queue capacity C. TT (Q a Does the queue meet the requirements of the TT stream? If the queue capacity is insufficient to hold the current stream's data volume (i.e., Q) TT (T a Q a )+LD (Va,Va+1) >C TT (Q a If this happens, the cycle offset and cycle label will be recalculated to ensure the effectiveness of scheduling.

[0145] Finally, the algorithm updates the resource reservation matrix R based on the calculated periodic labels and queue identifiers. TT , will flow f i Information is stored in the corresponding queue position to ensure that the stream can be transmitted as scheduled in future cycles. Information on successfully scheduled streams is also updated in the successfully scheduled TT stream set. middle.

[0146] In order to maximize the number of AVB flow scheduling successes in a mixed traffic scheduling problem, the application proposes an AVB flow scheduling algorithm based on dynamic cycle adjustment and elastic resource block allocation. The algorithm can dynamically adjust the cycle time according to the network load, reasonably allocate resource blocks, and effectively deal with the problem of resource fragmentation, thereby improving the transmission performance of AVB flow in a mixed traffic environment.

[0147] The AVB flow scheduling scheme based on dynamic cycle adjustment and elastic resource block allocation specifically includes:

[0148] The method first calculates the network load factor according to formula (16), and then obtains the dynamically adjusted T according to formula (17) elastic to adapt to the current state of the network. When the network load is large, i.e. Load > a, the cycle is extended to ensure that AVB flow can obtain sufficient bandwidth in heavy load conditions; when the network load is small, i.e. Load < b, the cycle is shortened to improve resource utilization. Referring to the congestion threshold design of TCP Vegas and DCTCP, when the link utilization is less than 30%, it is considered to be in a light load state, and when the link utilization exceeds 70%, the congestion avoidance mechanism is triggered. Therefore, in the experimental design of this chapter, a = 0.7 and b = 0.3.

[0149]

[0150] Then, for AVB flow, the required resource block number N j is calculated according to formula (18) block , where S block is the bandwidth of a single resource block. And use the sliding window mechanism to find available elastic resource blocks in the future T cycles: if enough available resource blocks are found in the sliding window, the algorithm will allocate these resources to the AVB flow and update the resource allocation matrix R AVB ; if there are not enough resource blocks available, the algorithm will trigger the token bucket throttling mechanism, i.e. adjust the bandwidth requirement of the AVB flow, through the token bucket quota, and recalculate the required resource block number N j .

[0151]

[0152] In addition, the algorithm also checks whether the resource allocation conflicts with the successfully scheduled time-sensitive flow (TT flow). If there is a resource conflict, the AVB flow will be postponed to the next safe time window.

[0153] Step S6: generating and outputting the final scheduling result, including: the transmission path of each flow in the traffic set, the accurate time slot allocation table of each flow in the link, the resource occupancy rate and the real-time guarantee index. After completing the scheduling, the algorithm broadcasts the updated cycle T elastic and the resource allocation matrix RAVB to the full network node to ensure the coordination and consistency of the entire network.

[0154] Figure 1 The hybrid traffic scheduling method proposed by the application realizes the resource block division and isolation mechanism. The CSQF divides the sending time of the output port into a series of equal length periods T slot , and divides the resource block into two types of fixed resource block (FB) and elastic resource block (EB), guarantees the deterministic transmission of time sensitive (TT) flow and the bandwidth allocation of audio video bridge (AVB) flow, and the specific hybrid traffic scheduling model structure is as follows:

[0155] (1) Assuming that the CSQF scheduling model uses N queue resources in total, among which the first K queues are allocated to TT flow, and the remaining N-K queues are allocated to AVB flow. One of the multiple queues participating in scheduling is taken as a sending queue (Tx), and the rest are taken as receiving queues (Rx);

[0156] (2) When the TT flow enters the queue, it is preferentially received by the receiving queue close to the sending queue, so as to wait for a smaller number of T slot , that is, a smaller waiting time, which meets the higher latency requirement;

[0157] (3) The AVB flow has lower priority and has a relatively easy latency requirement, and is received by the rear receiving queue;

[0158] (4) The capacity of the queue is L, when a certain amount of data packets are accumulated, the current queue cannot continue to receive, at this time, the queue behind and with capacity needs to be selected, otherwise overflow will cause packet loss;

[0159] Figure 2 It is explained that when Figure 1The mixed traffic scheduling of TT flow and AVB flow in two networks when the queue N in the figure is 3. The dotted line in the figure represents waiting for transmission, and the solid line represents transmitting current traffic. It is assumed that two flows (one is TT flow and one is AVB flow) are transmitted from the A node to the C node through the B node, and since the TT flow has a higher priority, the TT flow is transmitted preferentially in the T0 period, and the AVB flow waits in the next queue. According to the principle of the CSQF mechanism, the queue state period is repeated, and each queue takes turns to open the gate to become a sending queue. Moreover, in the wide area network scenario, the propagation delay cannot be ignored, and the propagation delay after a period is considered to arrive at the next node. Therefore, at T1, the queue of the AVB flow at the A node becomes a sending queue, and it starts to transmit; at the B node, the TT flow enters the first receiving queue (this queue becomes a sending queue in the next period). During T2, the AVB flow arrives at the B node and enters the receiving queue, and if there is no TT flow in the current flow to be scheduled, it can enter the first receiving queue preferentially; at this time, the TT flow is sent out at the B node. The scheduling process during T3 is consistent with that during T1, and at this time, the TT flow arrives at the C node and completes the scheduling. At T4, the AVB flow also arrives at the C node and completes the scheduling. Thus, the period label of the TT flow is <0, 2>. In the current example, initially, two types of traffic enter the queue at the same time and wait, and since the AVB flow is in the next queue, the final scheduling delay is longer than that of the TT flow by Δt, which is equal to 1T slot .

[0160] It can be seen that the CSQF mixed traffic scheduling mechanism proposed in the application guarantees the independent queue of each type of traffic, dynamically adjusts the resource allocation under the traffic demand and network load, and maximizes the network resource utilization efficiency.

[0161] The scheduling scheme based on the cycle specified queuing and forwarding (CSQF) mechanism proposed in the application. The main of the scheme is to guarantee the efficient collaborative transmission of two types of traffic through accurate resource allocation and dynamic scheduling strategy.

[0162] The collaborative transmission process of the TT flow and the AVB flow is as shown in Figure 3 , and the specific process is as follows:

[0163] Step S1: The CSQF mechanism allocates independent queue resources for each type of traffic, so that the scheduling of the TT flow and the AVB flow is carried out separately, but they are also related to each other.

[0164] Step S2: The TT flow scheduling mainly relies on the ILP integer linear programming to accurately allocate the fixed resource block (FB), and the period label is aligned to guarantee the timeliness of the TT flow.

[0165] Step S3: The AVB flow scheduling is flexible, and the elastic period T elastic can be dynamically adjusted, and the sliding window is used to preempt the elastic resource block (EB), so as to maximize the bandwidth utilization rate of the AVB flow.

[0166] Step S4: Check if the resource allocation conflicts with the successfully scheduled TT flows. If there is a resource conflict, the AVB flow will be postponed to the next safe time window. After the scheduling is completed, the algorithm broadcasts the updated cycle T elastic and the resource allocation matrix R AVB to all network nodes through the centralized network architecture to ensure the coordination and consistency of the entire network.

[0167] The definitions of the terms in the present application are shown in Table 1:

[0168]

[0169]

[0170]

[0171]

[0172] Table 1

[0173] Figure 4 For the complete flowchart of the algorithm of the present application, the detailed execution steps are as follows:

[0174] Step S1: Input the topology information and flow set information, including the directed graph G=(V, E), the network node set V, the link bandwidth, the transmission delay, and other physical connection relationships and the characteristics of all to-be-scheduled flows.

[0175] Step S2: Perform traffic classification and characteristic analysis to determine the flow type and check if it is a TT flow or an AVB flow. If it is determined to be a TT flow, go to step S3; if it is determined to be an AVB flow, go to step S4.

[0176] Step S3: TT flow scheduling algorithm based on fixed resource blocks. Assign fixed resource blocks to the highest priority and use the shortest weighted routing method to calculate the optimal transmission path of the TT flow, giving priority to low delay. By using ILP to solve the time slot, periodic label allocation is performed for each node on the current path. Calculate the link propagation delay between the node and the next node, and then calculate the period offset according to the link propagation delay. Finally, the algorithm also checks if the capacity of the current queue meets the demand of the TT flow. If the capacity of the queue is insufficient to accommodate the data volume of the current flow, the period offset and the period label will be recalculated to ensure the effectiveness of the scheduling.

[0177] Step S4: AVB flow scheduling optimization algorithm based on dynamic elastic period. According to the sliding window mechanism, the elastic resource block is allocated, that is, the time sliding window model is used to dynamically divide the resource block and allocate the elastic bandwidth for the AVB flow. Then, the sliding window preemption strategy optimizes the scheduling, and the sliding window mechanism is used to find the available elastic resource block in the future T period: if enough available resource blocks are found in the sliding window, the algorithm will allocate these resources to the AVB flow, and update the resource allocation matrix R AVB ; if there is not enough resource block available, the algorithm will trigger the token bucket throttling mechanism, that is, adjust the bandwidth requirement of the AVB flow, through the token bucket quota, and recalculate the required resource block number N j .

[0178] Step S5: Check if there is a conflict in the time slot allocation. The algorithm checks whether the resource allocation conflicts with the successfully scheduled TT flow. If there is a resource conflict, the AVB flow will be postponed to the next safe time window, and then enter step S4 to adjust the resource allocation strategy again. If there is no conflict, it directly enters step S6.

[0179] Step S6: Output the scheduling scheme. The final scheduling result is generated and output, including: the transmission path of each flow; the accurate time slot allocation table of each flow in the link; the resource occupation rate and real-time guarantee index.

[0180] The time complexity of the application mainly depends on the dynamic period stretching and elastic resource allocation process. In a single network node, the dynamic period stretching traverses the resource block state in the current period window, and its time complexity is O(N); the sliding window preemption only needs to query the state of K queues, and the complexity is O(K); the elastic resource allocation uses a proportional-integral controller to realize real-time bandwidth adjustment, and the calculation overhead is constant O(1). Therefore, the single-node complexity of the algorithm is O(N+K), which is significantly lower than the traditional centralized integer linear programming (ILP) scheduling method (O(2 N ). For large-scale scenarios, when the network scale expands to M nodes, the overall complexity is O(M(N+K)), which grows linearly with the number of nodes, meeting the design requirements for large-scale scenarios.

[0181] The application also provides a multi-switch TSN dynamic scheduling system based on fuzzy control, which is used to realize the aforementioned CSQF-based wide area network mixed flow scheduling method.

[0182] The application also provides an electronic device, which includes a memory and a processor, and the memory stores a computer program. The processor is used to execute the program in the memory to realize the aforementioned CSQF-based wide area network mixed flow scheduling method.

[0183] The application further provides a computer readable storage medium, which stores a computer program for implementing the aforementioned CSQF-based wide-area network mixed traffic scheduling method.

[0184] Please note that the technical features of the above embodiments can be combined in any manner, and for the sake of brevity, not all possible combinations of the technical features in the above embodiments are described, however, as long as the combinations of the technical features do not contradict each other, they should be considered as within the scope of the present disclosure. The above embodiments only express several implementation manners of the present application, which are described in a more specific and detailed manner, but should not be understood as a limitation on the scope of the patent. It should be pointed out that for those skilled in the art, several modifications and improvements can be made without departing from the concept of the present application, which are all within the protection scope of the present application. Therefore, the protection scope of the patent of the present application should be subject to the appended claims.

Claims

1. A CSQF-based method for wide area network hybrid traffic scheduling, characterized in that, The method includes: Step S1: The TSN switch discovers the TSN network topology according to the link discovery protocol, and abstracts the TSN network topology into a directed network graph using network modeling methods, obtaining the characteristics of the network node set, link bandwidth, transmission delay, and the set of traffic to be scheduled; the directed network graph is represented as G = (V, E), where V represents the network node set, including the TSN switch and terminal devices ES = {es1, es2, ..., es n }; E represents the set of physical links, where node V a With V b The physical links between them correspond to two unidirectional links [V] a V b ] and [V b V a ]; Step S2: Based on the acquired network node set, link bandwidth, transmission delay, and characteristics of the traffic set to be scheduled, perform traffic classification and feature analysis to obtain traffic types; construct a hybrid traffic scheduling model according to CSQF, allocate independent queue resources for each traffic type using CSQF, and check whether the traffic type is a TT flow or an AVB flow. If it is determined to be a TT flow, proceed to step S3; if it is determined to be an AVB flow, proceed to step S4. Step S3: Run the ILP-TT flow scheduling method based on the shortest weighted route, and proceed to step S5; Step S4: Run the AVB stream scheduling method based on dynamic periodic adjustment and elastic resource block allocation, and proceed to step S5; Step S5: Check if there is a conflict in time slot allocation, and check if the resource allocation conflicts with the successfully scheduled TT stream; if there is a resource conflict, postpone the AVB stream to the next safe time window, proceed to step S4, and readjust the resource allocation strategy; if there is no conflict, proceed directly to step S6. Step S6: Generate and output the final scheduling result, including: the transmission path of each flow in the traffic set, the precise time slot allocation table of each flow in the link, and the resource utilization rate and real-time guarantee indicators.

2. The CSQF-based WAN hybrid traffic scheduling method according to claim 1, characterized in that, The specific steps in step S2, namely, constructing a hybrid traffic scheduling model based on CSQF, are as follows: CSQF divides the transmission time of the output port into equal time intervals and defines queue resource blocks for each port under different time periods. Based on the CSQF mechanism, a hybrid traffic scheduling model is constructed, and resource block types are assigned to different traffic types for traffic scheduling.

3. The CSQF-based WAN hybrid traffic scheduling method according to claim 2, characterized in that, Step S3 specifically includes: Step S31: For each TT flow, select the optimal path using the shortest weighted routing algorithm; the path selection takes into account the physical structure of the network topology and is weighted in combination with the link weights. Step S32: Solve for the time slot using ILP, the transmission time interval of the output port is T. slot Periodic labels are assigned to each node V on the current path, and the values ​​from node V are calculated. a To the next node V a+1 Link propagation delay LD (Va,Va+1) Then calculate the period offset based on the link propagation delay. If the current node V a If it is the source node, then assign an initial periodic label <T. i Q i >, where T i Qi is the period identifier, and Qi is the queue identifier; if node V a If it is not the source node, then it inherits the periodic label T from the previous hop. a =T a-1 +Δ a-1 And based on the queue identifier Q of the previous node a =(Q a-1 +1)modM assigns a queue position to the current node; After completing the periodic label allocation, check the current queue capacity C. TT (Q a Does it meet the requirements of the TT stream? If the queue capacity is insufficient to accommodate the current stream's data volume Q? TT (T a Q a )+LD (Va,Va+1) >C TT (Q a If this happens, the cycle offset and cycle label will be recalculated. Step S33: Update the resource reservation matrix R based on the calculated periodic label and queue identifier. TT , will flow f i The information is stored in the corresponding queue position, and the information of successfully scheduled flows is updated in the set of successfully scheduled TT flows. middle.

4. The CSQF-based WAN hybrid traffic scheduling method according to claim 1, characterized in that, Step S4 specifically involves: Step S41: Calculate the network load factor according to formula (1), and then obtain the dynamically adjusted dynamic period length T according to formula (2). elastic To adapt to the current network conditions, when the network load is high, the period is extended to ensure that the AVB stream can obtain sufficient bandwidth under heavy load; when the network load is low, the period is shortened to improve resource utilization. Where FB is a fixed resource block, EB is an elastic resource block; α is a periodic correction factor, a and b are the load setting limits, and Load is the network load. Step S42: For the AVB stream, calculate the required number of resource blocks N according to formula (3). j S block For a single resource block bandwidth, a sliding window mechanism is used to find available elastic resource blocks over the next T periods: if enough available resource blocks are found within the sliding window, the resources are allocated to the AVB stream, and the resource allocation matrix R is updated. AVB If there are not enough resource blocks available, the token bucket rate limiting mechanism is triggered, which adjusts the bandwidth requirements of the AVB stream through token bucket quotas and recalculates the required number of resource blocks N. j ; in, Indicates the minimum bandwidth of the AVB stream; Check if the resource allocation conflicts with the successfully scheduled TT stream. If a resource conflict exists, the AVB stream will be postponed to the next safe time window. Step S43: After scheduling is completed, broadcast the updated period T through a centralized network architecture. elastic and resource allocation matrix R AVB This extends to all nodes in the network to ensure the overall coordination and consistency of the network.

5. The CSQF-based WAN hybrid traffic scheduling method according to claim 1, characterized in that, Step S1 also includes: The set of traffic to be scheduled, F, is represented as: F = {F} TT， F AVB A TT stream can be represented by a 7-tuple: in, These represent the source and destination nodes of the TT stream, respectively. Indicates the period of the TT stream. Describe the size of the data packet. Indicates the deadline. Indicates the path information of the stream. Indicates the offset time of the stream; An AVB stream can be represented by a 5-tuple: in These represent the source and destination nodes of the AVB stream, respectively. This indicates the bandwidth requirement of the stream. The deadline for the flow. This indicates the path information of the stream.

6. The CSQF-based WAN hybrid traffic scheduling method according to claim 5, characterized in that, Step S3 also includes: The ILP-TT flow scheduling method based on shortest weighted routing needs to meet the following constraints: (1) Strict deadline constraints End-to-end delay f from source node to destination node i_delay Strictly less than or equal to the cutoff delay Right now Where f i_delay Indicates the end-to-end delay of the stream. This indicates the maximum allowed delay for the stream, i.e., the cutoff delay. (2) Periodic lower limit constraint The lower bound constraint must ensure that all data packets stored in the queue can be scheduled within its duration. The constraint condition is as follows: Where f i_proc Indicates processing delay, f i_queue Indicates queuing delay, f i_trans L represents the transmission delay, L represents the maximum queue capacity, MTU is the maximum transmission unit of a data packet, and BW is the link capacity. (3) Periodic upper limit constraint The upper limit constraint on the period length ensures that the period of all TT streams in the system is divisible, and the upper limit is consistent with the greatest common divisor of the transmission periods of all streams; (4) Offset Time Constraint In the offset time constraint, the offset, which is the time from the source node to the first switch at the terminal node when transmission begins, cannot exceed the flow period, and the TT flow should be fully forwarded within the current transmission period. in, The offset time of the TT stream, T is the packet size of the TT stream. slot This is the transmission time interval for the output port; (5) Bounded queue length constraint In the bounded queue length constraint, the number of TT stream packets buffered in a single queue cannot exceed the total queue capacity L, to prevent packet loss and ensure that each fixed resource block queue is not overloaded in any period.

7. The CSQF-based WAN hybrid traffic scheduling method according to claim 1, characterized in that, Step S4 further includes: The AVB stream scheduling method based on dynamic periodic adjustment and elastic resource block allocation needs to meet the following constraints: (1) Bandwidth constraints: The minimum bandwidth of the AVB stream must be guaranteed during transmission. The number of continuous or discrete periodic blocks is allocated within the elastic resource block, and its quantity can be calculated from the bandwidth requirement, where T slot The transmission time interval for the output port: (2) Flexible resource block constraints: Define a resource block allocation matrix R(t), where R(t,m) = 1 indicates that queue m is occupied at period t, otherwise it is 0; (3) Preemptive window constraint: AVB streams are only allowed to preempt elastic resources within a limited future period through a sliding window strategy. Let W be the preemption window size, which is the number of periods. If the current period is t, then the AVB stream can only preempt EB resource blocks within the range of [t, t+W]. (4) Delay constraint: The maximum delay from when an AVB stream enters the queue to when it is scheduled must meet an upper limit to prevent low-priority traffic from being over-queued; this delay cannot exceed the maximum allowable delay of the AVB stream. Combined with queue mapping rules, AVB streams inserted later in the queue will experience a larger queuing delay: waiting periods × T. slot .

8. A CSQF-based wide area network hybrid traffic scheduling system, characterized in that, The system includes a processing unit configured to perform steps for implementing the CSQF-based wide area network hybrid traffic scheduling method according to any one of claims 1-7.

9. An electronic device, characterized in that: It includes a memory and a processor, wherein the memory stores a computer program, and the processor executes the program in the memory to implement the CSQF-based WAN hybrid traffic scheduling method as described in any one of claims 1-7.

10. A computer-readable storage medium, characterized in that, The storage medium stores the CSQF-based WAN hybrid traffic scheduling method as described in any one of claims 1-7.

Images