Bus network multi-priority traffic admission control and dynamic programming method

Abstract

The application relates to a bus network multi-priority flow admission control and dynamic planning method, and belongs to the technical field of wired communication network resource dynamic scheduling and planning. The method comprises the following steps: solving a first scheduling scheme of TT flows satisfying the condition that the end-to-end worst-case time delays of all RC flows are not more than the corresponding end-to-end time delay deadlines based on the to-be-admitted TT flows and performing; if the solving fails, solving a second scheduling scheme of TT flows in which the number of RC flows with end-to-end worst-case time delays exceeding the corresponding end-to-end time delay deadlines is as small as possible; if the solving fails, the TT flow admission is not passed; and if the solving succeeds, solving a compression scheme of part of the RC flows based on the second scheduling scheme, and reducing the transmission quality by using the bandwidth compression of part of the RC flows to exchange for the communication continuity of all the RC flows. The method minimizes the influence of the newly-admitted TT flows on the RC flows, increases the admission probability of the TT flows, and solves the problem that the newly-admitted TT flows affect the end-to-end delay of the RC flows and cause the end-to-end timeout of the RC flows.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of dynamic scheduling and planning of wired communication network resources, and particularly relates to a bus network multi-priority traffic admission control and dynamic planning method. BACKGROUND

[0002] The combination of electronic system bus network and Ethernet, especially through time-sensitive network (TSN) technology, is an important direction for the development of industrial automation, which can bring many advantages such as data integration, network simplification and interoperability enhancement. TSN is a set of IEEE standards aiming to provide determinism and real-time performance for Ethernet, so that periodic and aperiodic data can be transmitted in the same network while maintaining high bandwidth and low latency. In the TSN network, the reconfiguration scheme of TT (Time-Triggered) flow is the key to ensure the real-time performance and determinism of the network. In the prior art, when reconfiguring the TT flow, the communication performance of low-priority traffic is often not considered, and the impact on the end-to-end delay of rate-limited low-priority traffic (RC traffic) is ignored, which may cause the quality of service of some relatively critical RC traffic to decline or even cause communication interruption. SUMMARY

[0003] In view of the above analysis, the application aims to provide a bus network multi-priority traffic admission control and dynamic planning method to solve the problem of RC flow end-to-end timeout caused by the influence of RC flow by the to-be-admitted TT flow.

[0004] The method of the application specifically comprises the following steps:

[0005] Based on a plurality of TT flows to be admitted, a first scheduling scheme of the plurality of TT flows satisfying that the end-to-end worst-case delay of each RC flow does not exceed the corresponding end-to-end delay deadline is solved, and if the solution is successful, the first scheduling scheme is executed; otherwise,

[0006] A second scheduling scheme of the plurality of TT flows in which the number of RC flows whose end-to-end worst-case delay exceeds the corresponding end-to-end delay deadline is as small as possible is solved, and if the solution fails, the plurality of TT flows is not admitted; otherwise,

[0007] Based on the second scheduling scheme, a partial RC flow compression scheme satisfying that the end-to-end worst-case delay of all RC flows does not exceed the corresponding end-to-end delay deadline is solved, and if the solution is successful, the second scheduling scheme and the compression scheme are executed, otherwise,

[0008] The second scheduling scheme is executed, and a traditional RC flow reconstruction method is implemented for the RC flow whose worst-case delay exceeds the corresponding end-to-end delay deadline.

[0009] Further, when receiving the network withdrawal notification of the TT flow, if there is a compressed RC flow in the network, based on the set of compressed RC flows existing in the network and the network withdrawal flow routing, a decompression scheme of the RC flow is solved and the decompression scheme is executed.

[0010] Further, the particle swarm algorithm is used to solve the partial RC flow compression scheme based on the second scheduling scheme, including:

[0011] Based on the RC flow whose end-to-end worst case delay exceeds the corresponding end-to-end delay deadline and the RC flow sharing the same directed link with the routing thereof, a set of RC flows to be solved is determined;

[0012] Each RC flow corresponding to the set of RC flows to be solved is numbered;

[0013] Each combination of the corresponding compression scheme is taken as a particle, and the position of the particle is described by the combination of the corresponding numbers of each compression scheme in each particle. The particle swarm is randomly initialized based on the compression scheme corresponding to each RC flow in the set of RC flows to be solved;

[0014] Based on the end-to-end delay and the maximum bandwidth constraint of all affected RC flows, a fitness function is constructed;

[0015] The optimal solution of the particle swarm is obtained based on the fitness function, and a partial RC flow compression scheme is obtained.

[0016] Further, the fitness function is expressed as:

[0017] F(X)=Q(X)+γE(X);

[0018] Wherein, Q(X) is an objective function, Q(X)=max{D ave -min{δ}};D ave represents the average delay time rate of all RC flows; min{δ} represents the minimum value of the delay caused by compression that can satisfy the end-to-end delay not exceeding the corresponding end-to-end delay deadline, and δ satisfies Wherein F RC represents the set of RC flows to be solved, f h represents the hth RC flow in the set, represents the offset of the hth RC flow to the destination, represents the initial offset of the data flow routing vl h of the kth RC flow, dl h represents the end-to-end delay deadline of the hth RC flow;

[0019] γ is a penalty weight;

[0020] E(X) is a penalty function,

[0021]

[0022] wherein, denotes the number of RC flows affected; vl j denotes the route of the jth RC flow to be admitted or in the network; e αβ denotes the directed edge connecting network nodes v α , v β ; denotes the worst-case time interval for any frame of the RC flow to be transmitted from network node v α to network node v β ; dl i denotes the end-to-end delay bound of the ith RC flow; E denotes the set of directed communication links between network topology nodes; e denotes a directed communication link between network topology nodes; l j denotes the data flow frame length of the jth RC flow to be admitted or in the network; p j denotes the data flow period of the jth RC flow to be admitted or in the network; C j denotes the maximum bandwidth service rate of the directed link; denotes that the RC flows to be admitted and in the network satisfy the maximum bandwidth constraint.

[0023] Further, the decompression scheme of the RC flows is solved based on the set of RC flows in the network in which compression state exists and the route of the decommissioned traffic, comprising:

[0024] determining the RC flows involved in the decommissioned traffic based on the route of the decommissioned traffic;

[0025] solving the decompression scheme of the RC flows using the particle swarm algorithm on the intersection of the set of RC flows involved in the decommissioned traffic and the set of RC flows in the network in which compression state exists.

[0026] Further, the first scheduling scheme of the multiple TT flows to be admitted is solved based on the multiple TT flows to be admitted, so that the end-to-end worst-case delay of each RC flow does not exceed the corresponding end-to-end delay bound, comprising:

[0027] obtaining multiple optional routing schemes one by one based on the multiple TT flows to be admitted;

[0028] solving the first routing scheme based on the optional routing schemes, so that the end-to-end worst-case delay of each RC flow does not exceed the corresponding end-to-end delay bound, and if the solution is successful, the first scheduling scheme of the multiple TT flows is solved based on the first routing scheme; otherwise, the solution of the first scheduling scheme fails.

[0029] Further, the first routing scheme satisfying the end-to-end worst-case delay of each RC flow not exceeding the corresponding end-to-end delay deadline is solved based on the optional routing schemes, including:

[0030] An optimal routing scheme is solved based on the optional routing schemes, aiming at optimizing the delay time rate of each RC flow affected by the TT flow.

[0031] In each iteration, it is determined whether the end-to-end worst-case delay of each RC flow exceeds the corresponding end-to-end delay deadline based on the optimal routing scheme obtained in the iteration:

[0032] If yes, the optimal routing scheme obtained in the iteration is the first routing scheme.

[0033] If no, the next iteration is continued until the maximum number of iterations is reached.

[0034] Further, the particle swarm algorithm is used to solve the optimal routing scheme based on the optional routing schemes, aiming at optimizing the delay time rate of each RC flow affected by the TT flow, including:

[0035] The optional routing schemes are numbered.

[0036] The combination of the routing schemes of each to-be-admitted TT flow is taken as a particle, and the combination of the corresponding numbers of the routing schemes of each to-be-admitted TT flow in each particle is used to describe the particle position, and the particle swarm is randomly initialized based on the multiple optional routing schemes.

[0037] The fitness function is constructed based on the delay time rate of the affected RC flow and the maximum bandwidth constraint.

[0038] In each iteration, the optimal solution of the particle swarm is obtained based on the fitness function, and the optimal routing scheme of the iteration is obtained.

[0039] Further, the fitness function is constructed based on the delay time rate of the affected RC flow and the maximum bandwidth constraint, including:

[0040] The objective function is constructed based on the delay time rate of the affected RC flow.

[0041] The penalty function is constructed based on the maximum bandwidth constraint.

[0042] The fitness function is constructed based on the objective function, the penalty function, and the weight coefficient of the penalty function.

[0043] Further, the second scheduling scheme of the to-be-admitted TT flow is solved, including:

[0044] solving a second routing scheme based on the each preferred routing scheme, in each iteration round:

[0045] selecting the preferred routing scheme in which the number of RC flows whose end-to-end delay exceeds the corresponding worst delay is the least;

[0046] solving a second scheduling scheme based on the preferred routing scheme and scheduling constraints:

[0047] if successful, the preferred routing scheme is the second routing scheme, and the iteration is ended;

[0048] if failed, the next iteration round is continued until all the preferred routing schemes are traversed.

[0049] The present application can achieve at least one of the following beneficial effects:

[0050] By considering the influence of the to-be-admitted TT flow on the end-to-end delay of the RC flow in the network to determine the scheduling scheme of the TT flow, the influence of the TT flow on the RC flow in the TSN network is minimized, and the problem of RC flow end-to-end timeout caused by the influence of the newly-admitted TT flow on the end-to-end delay of the RC flow is solved.

[0051] By solving the first routing scheme that satisfies the condition that the end-to-end worst-case delay of all RC flows does not exceed the corresponding end-to-end delay deadline, packet loss and communication interruption of the RC flow are avoided from the source; and in the case that the first routing scheme fails to be solved, the admission requirement is reduced and the admission probability of the new TT flow is increased, the second scheduling scheme that can satisfy the condition that the end-to-end worst-case delay of most RC flows does not exceed the corresponding end-to-end delay deadline is determined, and on the basis of the second scheduling scheme, the RC flow is reconstructed elastically by compressing the bandwidth of part of the RC flow, the service quality of part of the RC flow is reduced in exchange for the communication continuity of all the RC flow, the robustness and determinacy of the network are guaranteed, and the situation of secondary reconstruction (i.e. re-allocation) of the RC flow is avoided as much as possible.

[0052] By decompressing the RC flow in the case of receiving the TT flow network exit notification, the network bandwidth resources are fully utilized, and the influence on the transmission quality of the RC flow is further reduced.

[0053] By constructing the fitness function based on the delay time and the maximum bandwidth constraint of the RC flow, the particle swarm algorithm is used to solve the first and second routing schemes, the compression scheme and the decompression scheme of part of the RC flow, respectively, to ensure that the optimal solution of the corresponding scheme is obtained at each link of the method.

[0054] The embodiment considers the influence of the to-be-admitted TT flow on the end-to-end delay of the RC flow in the network to determine the scheduling scheme of the TT flow, minimizes the influence of the TT flow on the RC flow in the TSN network, can reduce the admission requirement in the case of failure of solving the first scheduling scheme, increase the admission probability of the new TT flow, and realize the elastic reconstruction of the RC flow by bandwidth compression of part of the RC flow, exchange the service quality of part of the RC flow for the communication continuity of all the RC flow, avoid the packet loss and communication interruption of the RC flow, and thus guarantee the robustness and determinacy of the network.

[0055] Other features and advantages of the present application will be set forth in the following description, and in part will become apparent to those skilled in the art upon examination of the following or can be learned by practice of the present application. The objects and other advantages of the present application can be realized and attained by the instrumentalities expressed in the specification, claims and drawings. BRIEF DESCRIPTION OF DRAWINGS

[0056] The accompanying drawings are included to provide a further understanding of the application and are incorporated in and constitute a part of this specification, illustrate embodiments of the application and together with the description serve to explain the principles of the application, and should not be considered limiting of the application's scope as it is encompassed by the appended claims.

[0057] Figure 1 Flow chart of the traffic admission control and dynamic planning method of the present application;

[0058] Figure 2 Schematic diagram of the RC flow decompression process of the method of the present application. DETAILED DESCRIPTION

[0059] The preferred embodiments of the present application will be described in detail below with reference to the drawings, wherein the drawings constitute a part of this application and are used to explain the principles of the embodiments of the present application, and are not considered limiting the scope of the present application as it is encompassed by the appended claims.

[0060] One specific embodiment of the present application discloses a bus network multi-priority traffic admission control and dynamic planning method, specifically comprising the following steps:

[0061] Step S1, based on a plurality of to-be-admitted TT flows, solving a first scheduling scheme of the plurality of TT flows satisfying that the end-to-end worst-case delay of each RC flow does not exceed the corresponding end-to-end delay deadline, if the solving is successful, executing the first scheduling scheme;

[0062] Solving a second scheduling scheme of the plurality of TT flows with the number of RC flows whose end-to-end worst-case delay exceeds the corresponding end-to-end delay deadline being as few as possible, if the solving fails, the plurality of TT flows are not admitted; otherwise, entering step S2,

[0063] Step S2, based on the second scheduling scheme, a partial RC flow compression scheme is solved, which satisfies all RC flow end-to-end worst case delay does not exceed the corresponding end-to-end delay deadline, if the solution is successful, the second scheduling scheme and the compression scheme are executed, otherwise, the second scheduling scheme is executed, and the traditional RC flow reconstruction method is implemented for the RC flow whose worst case delay exceeds the corresponding end-to-end delay deadline;

[0064] Step S3, when receiving the network withdrawal notification of the TT flow, if there is a RC flow in the compression state in the network, based on the compression state RC flow set existing in the network and the network withdrawal flow routing, a decompression scheme of the RC flow is solved and the decompression scheme is executed.

[0065] In the embodiment, the scheduling scheme of the TT flow is determined by considering the influence of the to-be-admitted TT flow on the end-to-end delay of the RC flow in the network, the influence of the TT flow on the RC flow in the TSN network is minimized, in the case that the first scheduling scheme fails to be solved, the admission requirement is reduced, the admission probability of the new TT flow is increased, and the RC flow is elastically reconstructed by compressing the bandwidth of the partial RC flow, the communication continuity of all RC flows is exchanged for the service quality of the partial RC flow, the packet loss and communication interruption of the RC flow are avoided, and thus the robustness and determinacy of the network are ensured.

[0066] It should be noted that the bus network multi-priority flow admission control and dynamic planning method of the application adopts a TAS scheduling model based on IEEE 802.1Qbv, and the flow includes time-triggered (TT) flow, rate-constrained (RC) flow (such as audio and video AVB flow) and traditional Ethernet best effort (BE) flow. In the application, only the TT flow and the RC flow are considered. In order to be general, the RC flow in the application is only periodically sent at the source end system, and no additional scheduling algorithm is executed in the ports of the passing switches, and only the basic first-in-first-out principle (FIFO) is followed. This assumption is reasonable and downward compatible with more complex scheduling algorithms, and thus the most widely applicable.

[0067] Specifically, the TSN network topology is represented as a directed graph G(V, E), where V represents a set of terminal systems (ES) and switch (SW) network nodes, and E represents a set of directed communication links between network nodes. Let v1 and v2 represent two adjacent nodes, and satisfy v1, v2∈V, then [v1, v2] and [v2, v1] can be used to represent a bidirectional data flow link. The edge e st can be described as e st =[v s ,v t ]∈E,v s ,v t∈V. Multi-priority traffics in Time-Sensitive Network (TSN) mainly include TT, Rate Constrained (RC) and Best Effort (BE) traffics. TT traffics are forwarded according to an offline predetermined schedule, with the highest priority. RC traffics are periodically transmitted at the source end system, with a rate limited by Bandwidth Allocation Gap (BAG), and with a lower priority than TT traffics. BE traffics are traditional Ethernet traffics without any QoS guarantee, with the lowest priority, and are not discussed in this invention.

[0068] Further, all communication tasks in the network are transmitted through data flows, including data frame instances. Let F = F TT ∪F RC represent the set of all data flows. Any TT flow f m ∈F TT and any RC flow f n ∈F RC can be represented by the following seven-tuple and six-tuple, respectively:

[0069] f m = {s m , d m , l m , p m , dl m , vl m , O m} and f n = {s n , d n , l n , p n , dl n , vl n} ;

[0070] where s m , d m , l m , p m , dl m and vl m represent the source ES, the destination ES, the frame length in time dimension (frame length divided by link bandwidth), the traffic period, the End-to-End (ETE) latency bound and the routing information, respectively. O m is unique to TT traffics, representing the set of offsets between the first data frame instance on each traversed directed link and the start time of the period, which collectively form the time schedule of the network. In simple terms, the admission control and dynamic planning of TT traffics mean re-allocating the routing information and the corresponding offset time of TT traffics

[0071] Specifically, in step S1, based on the multiple TT flows to be admitted, solving the first scheduling scheme of the multiple TT flows that satisfies the worst-case end-to-end delay of each RC flow does not exceed the corresponding end-to-end delay period includes steps S11-S12.

[0072] S11. Based on multiple TT flows to be admitted and the network topology, use Dijkstra's algorithm to calculate and obtain multiple optional routing schemes.

[0073] S12. Based on the optional routing scheme, solve for a first routing scheme that satisfies the worst-case end-to-end delay of each RC flow not exceeding the corresponding end-to-end delay period. If the solution is successful, then solve for a first scheduling scheme of the multiple TT flows based on the first routing scheme; otherwise, the solution for the first scheduling scheme fails.

[0074] Furthermore, step S12 includes S121-S123.

[0075] S121. With the goal of optimizing the delay rate of each RC flow affected by the TT flow to be admitted, the particle swarm optimization algorithm is used to iteratively solve the optimal routing scheme based on the available routing schemes.

[0076] Specifically, the available routing schemes are numbered from 1 to k.

[0077] Furthermore, the combination of routing schemes for each TT flow to be admitted is used as a particle, and the particle position is described by the combination of the corresponding numbers of the routing schemes for each TT flow in each particle. The position of each particle is represented as an M-dimensional vector X = [x1, x2, ..., x...]. M ], where x m (m∈[1,M]) represents the pending TT flow f. m The routing scheme number is determined based on the range of available routes, and its value is an integer from 1 to k. The particle's velocity is represented as V = [v1, v2, ..., v...]. M ], where v m This represents the distance the particle moves in the m-th dimension, and its value ranges from -(k-1) to (k-1).

[0078] Furthermore, an objective function is constructed based on the delay rate of the affected RC stream.

[0079] The objective function is expressed as:

[0080]

[0081] Where Q(X) is the objective function used to calculate the time delay rate of the affected RC flow; X is a c-dimensional vector representing the particle position, X = [x1, x2, ..., x...]. m ]; c represents the number of TT streams to be admitted; denotes the number of affected RC flows; denotes the worst end-to-end delay of the i-th RC flow, where vl i denotes the route of the i-th RC flow, e αβ denotes the directed edge connecting network node v α , v β , v denotes the worst time interval for the RC flow to transmit any frame from network node v α to network node v β ; dl i denotes the end-to-end delay limit of the i-th RC flow; D ave denotes the average delay time rate of all RC flows, F RC denotes the total number of RC flows in the network.

[0082] Further, a penalty function is constructed based on the maximum bandwidth constraint.

[0083] The penalty function is represented as:

[0084]

[0085] where E denotes the set of directed communication links between network topology nodes; vl j denotes the route of the j-th to-be-admitted TT flow or RC flow in the network; e denotes the directed communication link between network topology nodes; l j denotes the data flow frame length of the j-th to-be-admitted TT flow or RC flow in the network; p j denotes the data flow period of the j-th to-be-admitted TT flow or RC flow in the network; C j denotes the maximum bandwidth service rate of the directed link; denotes the expectation that the to-be-admitted TT flow and the RC flow in the network satisfy the maximum bandwidth constraint.

[0086] Further, a fitness function is constructed based on the objective function, the penalty function, and the weight coefficient of the penalty function, for evaluating the performance of the routing scheme represented by the particle position. The lower the fitness function value, the better the performance of the selected route.

[0087] The fitness function is represented as: F(X) = Q(X) + γE(X); where γ is the penalty weight, which varies according to different topologies and traffic, and is preferably 300.

[0088] Further, in each round of iteration, the optimal solution of the particle swarm is obtained based on the fitness function, and the preferred routing scheme of the round of iteration is obtained and recorded.

[0089] Specifically, in the process of searching the optimal solution, the position and velocity of the particle are updated according to two values: the current optimal solution pBest of the single particle and the optimal solution gBest of all particles in the population; the search process is represented as:

[0090] V t+1 = ωV t + c1r1(pBest - X t ) + c2r2(gBest - X t ) ;

[0091] X t+1 = X t + V t ;

[0092] wherein ω is an inertia weight, used to adjust the degree of the particle to keep its original velocity, and the value range is between 0-1, and preferably set to 0.9; c1 and c2 respectively represent self-learning and social learning factors, used to balance the local and global search ability, c1 controls the speed and direction of the particle to update itself in the search space, the greater the value, the more inclined to search along the direction of its own historical optimal solution, and the smaller the value, the more inclined to random exploration; c2 controls the speed of the particle to update in the space according to the optimal solution of the neighbor particle, the greater the value, the more inclined to search affected by the neighbor particle, and the smaller the value, the more inclined to independent search; the sum of c1 and c2 is not greater than 4, and preferably, c1 and c2 are set to 2 and 0.9 respectively; r1 and r2 are random numbers between 0 and 1; t represents the search number in the solution process.

[0093] S122, judging whether the worst-case end-to-end delay of each RC flow is less than the corresponding end-to-end delay limit based on the preferred routing scheme obtained in each iteration: if yes, the preferred routing scheme obtained in the iteration is the first routing scheme; if no, the next iteration is continued until the maximum iteration number is reached.

[0094] Specifically, the calculation method of the worst-case end-to-end delay of each RC flow is as follows:

[0095] calculating the worst transmission interval between each two nodes of the network link based on the preferred routing scheme;

[0096] calculating the worst-case delay of the corresponding RC flow based on all the worst transmission intervals of the RC flow.

[0097] Further, the calculation of the worst transmission interval between each two nodes of the network link based on the preferred routing scheme comprises:

[0098]

[0099] wherein, denotes the worst time interval for the transmission of an arbitrary frame of the RC flow from network node v α to network node v β ; and denote the time instants for the transmission of an arbitrary frame of the RC flow to network nodes v β and v α , respectively; denotes the technical delay introduced by each data frame; denotes the delay impact of TT flows on the RC flow, where F TT denotes the set of other TT flows that are not admitted TT flows, f k denotes the other TT flows that are admitted TT flows, denotes the time offset between the transmission of an arbitrary frame of the RC flow from network node v α to network node v β , l k denotes the data flow frame time length of the k-th admitted TT flow, p k denotes the flow period of the k-th admitted TT flow; denotes the delay impact of other RC flows on the RC flow, where F RC denotes the set of RC flows, f l denotes the admitted TT flows, F TTRec denotes the set of admitted TT flows, l l denotes the frame time length of the l-th other RC flow, p k denotes the flow period of the l-th other RC flow; denotes the additional delay of TT flows on the RC due to the adoption of the pre-emption integration strategy; l i denotes the frame time length of the RC flow.

[0100] The following explains the principle of calculating the worst transmission interval of each RC flow between each two nodes of the network link:

[0101] In the method of the present application, it is reasonably assumed that all RC flows have passed the worst-case end-to-end delay verification in the design stage, i.e. (wcd n denotes the worst-case end-to-end delay of the n-th RC flow). When judging whether the end-to-end worst-case delay of each RC flow exceeds the corresponding end-to-end delay limit, it is only necessary to judge the RC flows that have intersection with the route of the admitted TT flow. This is because in the delay analysis, the end-to-end delay of the non-intersecting RC flow may change, but its worst-case end-to-end delay wcd does not change (regardless of the delay analysis method such as network calculus or trajectory method).

[0102] Further, the end-to-end delay of RC flow is mainly affected by four aspects: ① the delay caused by the related technology, such as the delay caused by the data validity and integrity check between nodes, the delay caused by the error containment interval of bandwidth allocation, etc.; ② the delay caused by the bandwidth resource competition between RC flows; ③ the delay caused by the high priority TT flow to RC flow; and ④ the delay caused by the integration of TT and RC flows, which is mainly divided into three kinds of shuffling, preemption and timely blocking according to different integration strategies. In the method of the present application, the most common preemption strategy is adopted (this strategy is the most general, representative and practical).

[0103] Further, the end-to-end delay of RC flow is mainly composed of two parts: ① the queuing delay of source ES; and ② the delay of each intermediate node (link).

[0104] First, the queuing delay at the source ES is considered. Specifically, the set of RC flows intersecting with the route of the to-be-admitted TT flow is identified as F RC . Since the initial phase (offset) of the route corresponding to the RC flow can be allocated by a conventional offset allocation algorithm, for the RC flow f , the initial offset value of f i is denoted as Further, the transmission time (denoted as ) of f j from the end node into the network can be calculated as follows:

[0105]

[0106] where FIFO(j, i) represents that the data frame of TT flow f j enters the buffer earlier than the data frame of RC flow f i ; when FIFO(j, i) is an empty set, Therefore, the offset of RC flow f i entering the network can be represented by several possible discrete values.

[0107] Second, the delay calculation of each intermediate node (link) is considered. Specifically, the present application adopts an iterative strategy of overall timing analysis, and uses the concept of busy period to analyze the delay of the affected RC flow in the network. For any data frame of RC flow, its busy period on a certain link can be defined as a time interval, i.e. the time interval from the frame arriving at node v α to the frame being transmitted to the next node v β is denoted as The time interval from the frame arriving at node v α to the frame being transmitted to the next node v β is denoted as The end-to-end delay of the RC flow is denoted as To ensure that the end-to-end worst-case delay of each RC flow does not exceed the corresponding end-to-end delay deadline dl i worst-case delay interval is described as (for any known ):

[0108]

[0109] wherein, represents the technical delay introduced by each data frame; represents the delay impact of TT flows on the RC flow; represents the delay impact of other RC flows on the RC flow; represents the additional delay of the RC flow due to the preemption integration strategy of TT flows.

[0110] Since the above formula is a recursive function containing on both sides, the recursive tree method can be used to solve it. Thus obtained represents the worst-case end-to-end delay of RC flow f αβ in link e i from time t , which can guarantee the transmission of all TT flows and other accumulated RC flows that arrive first in this time period. The obtained can be used as the starting time of the next link in vl i to solve the busy period until is obtained.

[0111] The above is the principle of calculating the worst-case transmission interval of each RC flow between each two nodes of the network link.

[0112] Further, in S1022, it is judged whether the end-to-end worst-case delay of each RC flow exceeds the corresponding end-to-end delay deadline, which can be expressed as:

[0113]

[0114] S123, based on the first routing solution obtained by solving in step S122, solving a first scheduling solution of the plurality of TT flows.

[0115] Specifically, if the solving of the first routing solution in step S122 fails, it means that the solving of the first scheduling solution fails; if the solving of the first routing solution in step S122 succeeds, an integer linear programming solver is used to solve a first scheduling solution that satisfies the scheduling constraints based on the first routing solution. The solver can be selected, such as gurobi, cplex, etc. Since the solver and the solving method are prior art, they will not be discussed here.

[0116] Further, if the first scheduling scheme is solved successfully based on the first routing scheme, the first scheduling scheme is dispatched as a new configuration scheme A to configure the corresponding nodes in the network to perform the first scheduling scheme. If it is not successful, it is determined whether the maximum iteration number of the particle swarm algorithm is reached. If the maximum iteration number is not reached, the first routing scheme of the next iteration is solved. If the maximum iteration number is reached, it means that the first scheduling scheme is failed to be solved. It should be noted that, generally, the solving of the first scheduling scheme based on the first routing scheme is related to the network load, and the failure probability is extremely small in actual application.

[0117] Specifically, in step S1, if the first scheduling scheme is failed to be solved, step S13 is performed.

[0118] Step S13, a second scheduling scheme of the plurality of TT flows with the least number of RC flows whose end-to-end worst-case delay exceeds the corresponding end-to-end delay deadline is solved. If the solving fails, the plurality of TT flows is rejected.

[0119] Specifically, based on all the preferred routing schemes obtained in step S121, the second routing scheme is iteratively solved. Specifically, in each iteration:

[0120] The preferred routing scheme with the least number of RC flows whose end-to-end delay exceeds the corresponding worst-case delay is selected from the preferred routing schemes;

[0121] The solver of integer linear programming is used to solve the second scheduling scheme based on the preferred routing scheme and the scheduling constraints:

[0122] If the solving is successful, the preferred routing scheme is the second routing scheme, and the iteration is ended.

[0123] If the solving fails, it is determined whether all the preferred routing schemes are traversed. If not, the next iteration is continued. If yes, it means that the second scheduling scheme is failed to be solved, and the admission request of the plurality of TT flows to be admitted is rejected.

[0124] It should be noted that, in each iteration, the solving of the second scheduling scheme based on the preferred routing scheme selected in the iteration and the scheduling constraints is related to the network load. In actual application, the failure probability of the solving of the second scheduling scheme is extremely small.

[0125] Specifically, if the second scheduling scheme is solved in step S1, in step S2, a partial RC flow compression scheme satisfying all the RC flows whose end-to-end worst-case delay does not exceed the corresponding end-to-end delay deadline is solved based on the second scheduling scheme.

[0126] Further, a particle swarm algorithm is used to solve a partial RC flow compression scheme based on the second scheduling scheme, which satisfies that the end-to-end worst-case delay of all RC flows does not exceed the corresponding end-to-end delay deadline. Specifically, S21-S25 are included.

[0127] S21, determining a set of RC flows to be solved based on the RC flows whose end-to-end worst-case delay exceeds the corresponding end-to-end delay deadline and the RC flows whose routing shares the same directed link.

[0128] Specifically, based on the second scheduling scheme, the RC flows whose end-to-end worst-case delay exceeds the corresponding end-to-end delay deadline and the RC flows whose routing shares the same directed link are determined to form a set F RC . The number of RC flows in the set F RC ' is denoted as M'.

[0129] S22, numbering the corresponding compression scheme of each RC flow in the set of RC flows to be solved, and the total number of period scheme numbers of each RC flow in the set F RC ' is denoted as k'.

[0130] S23, taking the combination of each corresponding compression scheme as a particle, and describing the particle position with the combination of the corresponding numbers of each corresponding compression scheme in each particle, and initializing the particle swarm randomly based on the corresponding compression scheme of each RC flow in the set F RC ' of RC flows to be solved.

[0131] Specifically, the position of each particle is represented as an M' dimensional vector X' = [x'1, x'2,..., x' M ], where x' m' (m' ∈ [1, M']) represents the compression scheme number of the RC flow f m' to be solved, and the moving range of the particle is -(k'-1) to (k'-1).

[0132] S24, constructing a fitness function based on the end-to-end delay and the maximum bandwidth constraint of all affected RC flows.

[0133] Specifically, a target function is constructed based on the end-to-end delay of all affected RC flows. The target function is represented as:

[0134] Q(X) = max{D ave -min{δ}};

[0135] where D ave represents the average delay time rate of all RC flows; min{δ} represents the minimum value of the delay caused by compression that can satisfy that the end-to-end delay of all affected RC flows does not exceed the corresponding end-to-end delay deadline, and δ satisfies where F RCRC flow set to be solved, f h represents the hth RC flow in the set, represents the offset of the hth RC flow to the destination, the calculation method and principle of which have been described in step S1022 and will not be discussed here. represents the data flow route vl h of the kth RC flow, h represents the end-to-end delay limit of the hth RC flow.

[0136] Further, a penalty function is constructed based on the maximum bandwidth constraint. The penalty function is the same as that in step S121 and is represented as:

[0137]

[0138] wherein, represents the number of affected RC flows; vl j represents the route of the jth new TT flow or RC flow in the network; e αβ represents the directed edge connecting network nodes v α and v β ; represents the worst time interval for transmission of any frame of the RC flow from network node v α to network node v β ; dl i represents the end-to-end delay limit of the ith RC flow; E represents the set of directed communication links between network topology nodes; e represents a directed communication link between network topology nodes; l j represents the data flow frame length of the jth new TT flow or RC flow in the network; p j represents the data flow period of the jth new TT flow or RC flow in the network; C j represents the maximum bandwidth service rate of the directed link; represents the satisfaction of the maximum bandwidth constraint by the expected TT flow to be admitted and the RC flow in the network.

[0139] Further, a fitness function is constructed based on the objective function, the penalty function, and the weight coefficient of the penalty function. The fitness function is represented as:

[0140] F(X) = Q(X) + γE(X);

[0141] wherein γ is the penalty weight; preferably, the value is 300.

[0142] S25, obtaining a partial RC flow compression scheme based on the optimal solution of the particle swarm according to the fitness function.

[0143] Specifically, in each round of particle swarm iteration reaching the maximum iteration number, the current optimal solution pBest of a single particle and the optimal solution gBest of all particles in the population are solved based on the fitness function. According to the fitness function constructed in S24, it is known that the solution obtained needs to satisfy Further, in the desired iteration reaching the maximum iteration number, the optimal solution is determined based on the best fitness function value, and a partial RC flow compression scheme is obtained.

[0144] Specifically, in steps S21-S25, if the partial RC flow compression scheme is successfully solved, the second scheduling scheme and the compression scheme are distributed as a new configuration scheme B to the corresponding nodes in the network for configuration switching to execute the second scheduling scheme and the compression scheme. Otherwise,

[0145] Step S26 is performed, the second scheduling scheme is distributed as a new configuration scheme B' to the corresponding nodes in the network for configuration switching to execute the second scheduling scheme, and a traditional RC flow reconstruction method is implemented for RC flows whose worst-case latency exceeds the corresponding end-to-end latency deadline.

[0146] It should be noted that in actual application, the probability of failure of the partial RC flow compression scheme in steps S21-S25 is extremely small, and the service quality of the partial RC flow can be maximally reduced to exchange for the communication continuity of the entire RC flow, avoiding packet loss and communication interruption of the RC flow.

[0147] The traditional RC flow reconstruction method in the prior art is described below: The traditional RC flow reconstruction method is a post-hoc flow reconfiguration method, that is, only the routing and scheduling scheme of the newly added TT flow is considered to meet the related delay and jitter constraints in the design phase; after the newly added TT flow is distributed to the nodes, whether the specified RC flow needs to be reconfigured is determined according to the specific frame loss situation (continuous periodic frame loss or communication timeout) in the network nodes; at this time, the communication interruption (i.e., stopping data transmission) of these timeout RC flows is stopped, and the routing scheme is re-solved and distributed according to the link load information; the process is repeated until there is no RC flow in the network that appears communication timeout, or after a sufficient number of iterations and the problem cannot be solved, a subset of the set of RC flows is gradually attempted to be restored; after further iteration for a sufficient number of times and still unable to solve the problem, the subset is gradually attempted to be reduced until the subset is empty (the empty set indicates that all specified RC flows are not restored and all are no longer in communication). The strategy of reducing the subset is based on expert experience and the importance ranking of RC flows, which is predefined at the initial design time.

[0148] Further, in the bus network multi-priority traffic admission control and dynamic programming method described in the embodiment, when steps S1 and S2 are implemented, the admission control and dynamic programming for the TT flows to be admitted in this round have been completed (as shown in Figure 1

[0149] Further, in step S3 (as shown in Figure 2 When the network withdrawal notification of the TT flow is received, if there is a compressed RC flow in the network, the decompression scheme of the RC flow is solved and implemented based on the set of compressed RC flows existing in the network and the routing of the network withdrawal flow.

[0150] Further, in step S3, the RC flow involved in the network withdrawal flow is determined based on the routing of the network withdrawal flow; the intersection of the set of RC flows involved in the network withdrawal flow and the set of compressed RC flows existing in the network is the set of RC flows to be solved in the decompression scheme, denoted as F RC ”.

[0151] Further, the particle swarm algorithm is used to solve the decompression scheme of F RC The solving process is consistent with the process of solving the compression scheme of part of the RC flow in step S2.

[0152] Further, if the decompression scheme is solved within the maximum number of iterations, the decompression scheme is distributed as a new configuration scheme C to the corresponding nodes in the network for configuration switching to implement the decompression scheme; if the solution fails, the existing configuration scheme B in the network is not changed and remains unchanged.

[0153] The bus network multi-priority traffic admission control and dynamic programming method disclosed in the embodiment determines the scheduling scheme of the TT flow by considering the influence of the TT flow to be admitted on the end-to-end delay of the RC flow in the network, minimizes the influence of the TT flow in the TSN network on the RC flow, and solves the problem of RC flow end-to-end timeout caused by the influence of newly admitted TT flow on the end-to-end delay of RC flow.

[0154] ​By solving the first routing scheme satisfying all RC flow end-to-end worst case delay does not exceed the corresponding end-to-end delay deadline, the packet loss and communication interruption of RC flow are avoided from the source; and in the case of failure to solve the first routing scheme, the admission requirement is reduced, the admission probability of new TT flow is increased, the second scheduling scheme which can satisfy most RC flow end-to-end worst case delay does not exceed the corresponding end-to-end delay deadline is determined, and on the basis of the second scheduling scheme, the RC flow is reconstructed by bandwidth compression, the service quality of part of RC flow is reduced to exchange for the communication continuity of all RC flow, the robustness and determinacy of network are ensured, and the RC flow secondary reconstruction (i.e. re-allocation) is avoided as far as possible.

[0155] By decompressing the RC flow in the case of receiving the TT flow network exit notification, the network bandwidth resources are fully utilized, and the influence on the transmission quality of RC flow is further reduced.

[0156] By constructing the fitness function based on the delay time and the maximum bandwidth constraint of RC flow, the particle swarm algorithm is used to solve the first, second routing scheme, part of RC flow compression scheme and decompression scheme, and the optimal solution of the corresponding scheme is ensured in each link of the method.

[0157] The above is only the preferred specific embodiment of the present application, but the protection scope of the present application is not limited to this, any person skilled in the art can easily think of the changes or replacements within the technical range disclosed by the present application, which should be covered in the protection scope of the present application.

Claims

1. A bus network multi-priority traffic admission control and dynamic scheduling method, characterized in that, The method comprises the following steps: Solving a first scheduling scheme of the multiple TT flows satisfying that the end-to-end delay of each RC flow does not exceed the corresponding end-to-end delay deadline based on the multiple TT flows to be admitted, comprising: obtaining multiple routing schemes based on the multiple TT flows to be admitted; solving a first routing scheme satisfying that the end-to-end delay of each RC flow does not exceed the corresponding end-to-end delay deadline based on the multiple routing schemes, if the solving is successful, solving the first scheduling scheme of the multiple TT flows based on the first routing scheme; otherwise, the solving of the first scheduling scheme fails; if the solving of the first scheduling scheme is successful, the first scheduling scheme is executed; otherwise, Solving a second scheduling scheme of the multiple TT flows in which the number of RC flows whose end-to-end delay exceeds the corresponding end-to-end delay deadline is the least, if the solving fails, the multiple TT flows are not admitted; otherwise, Solving a partial RC flow compression scheme satisfying that the end-to-end delay of all RC flows does not exceed the corresponding end-to-end delay deadline based on the second scheduling scheme, if the solving is successful, the second scheduling scheme and the compression scheme are executed, otherwise, The second scheduling scheme is executed, and an RC flow reconstruction method is implemented on the RC flow whose delay exceeds the corresponding end-to-end delay deadline; The solving of the first routing scheme satisfying that the end-to-end delay of each RC flow does not exceed the corresponding end-to-end delay deadline based on the multiple routing schemes comprises: solving a target routing scheme based on the multiple routing schemes with the optimization of the delay time rate of each RC flow affected by the TT flow as the target; in each iteration, it is judged whether the end-to-end delay of each RC flow exceeds the corresponding end-to-end delay deadline based on the target routing scheme obtained in the iteration; if yes, the target routing scheme obtained in the iteration is the first routing scheme; if no, the next iteration is continued until the maximum number of iterations is reached; The solving of the second scheduling scheme of the multiple TT flows in which the number of RC flows whose end-to-end delay exceeds the corresponding end-to-end delay deadline is the least comprises: solving a second routing scheme based on each target routing scheme, in each iteration: selecting a target routing scheme in which the number of RC flows whose end-to-end delay exceeds the corresponding delay is the least among the target routing schemes; solving a second scheduling scheme based on the target routing scheme and the scheduling constraint; if successful, the target routing scheme is the second routing scheme, and the iteration is ended; if failed, the next iteration is continued until all target routing schemes are traversed.

2. The flow admission control and dynamic planning method of claim 1, wherein, When receiving the network withdrawal notification of the TT flow, if there is an RC flow in the compressed state in the network, a decompression scheme of the RC flow is solved and executed based on the set of RC flows in the compressed state existing in the network and the route of the withdrawn flow.

3. The traffic admission control and dynamic planning method of claim 2, wherein, The solving of the partial RC flow compression scheme based on the second scheduling scheme using the particle swarm algorithm comprises: Determining a set of RC flows to be solved based on the RC flow whose end-to-end delay exceeds the corresponding end-to-end delay deadline and the RC flow sharing the same directed link with the routing thereof; Numbering the corresponding compression schemes of each RC flow in the set of RC flows to be solved; The particle is a combination of the corresponding compression schemes, the particle position is described by the combination of the corresponding compression schemes corresponding to the number in each particle, and the particle swarm is randomly initialized based on the corresponding compression schemes of each RC flow in the to-be-solved RC flow set; An adaptive function is constructed based on the end-to-end delay and maximum bandwidth constraint of all affected RC flows; An optimal solution is obtained for the particle swarm based on the adaptive function, and a partial RC flow compression scheme is obtained.

4. The traffic admission control and dynamic planning method of claim 3, wherein, The adaptive function is expressed as: ； in, Let be the objective function. ; This represents the average delay rate of all RC streams; This means that the delay caused by compression in all affected RC flows can satisfy the condition that the end-to-end delay does not exceed the minimum value of the corresponding end-to-end delay period. satisfy ,in Denotes the set of RC flows to be solved. This represents the h-th RC flow in the set. This represents the offset of the h-th RC stream as it reaches the destination. This indicates the data flow route for the h-th RC flow. The initial offset, Indicates the end-to-end delay duration of the h-th RC stream; penalty weight; is a penalty function, ； wherein, denotes the number of RC flows that are affected; denotes the route of the j-th to-be-admitted TT flow or RC flow in the network; denotes the directed edge connecting the network nodes , ; denotes the time interval for the arbitrary frame of the RC flow to be transmitted from the network node to the network node ; denotes the end-to-end delay limit of the i-th RC flow; denotes the set of directed communication links between the network topology nodes; denotes the directed communication link between the network topology nodes; denotes the data flow frame length of the j-th to-be-admitted TT flow or RC flow in the network; denotes the data flow period of the j-th to-be-admitted TT flow or RC flow in the network; denotes the maximum bandwidth service rate of the directed link; denotes that the to-be-admitted TT flow and the RC flow in the network are expected to satisfy the maximum bandwidth constraint.

5. The traffic admission control and dynamic scheduling method of claim 4, wherein, The RC flow set based on the compression state existing in the network and the off-network traffic are routed, and the decompression scheme of the RC flow is solved, including: The RC flow involved in the off-network traffic is determined based on the off-network traffic routing; The intersection of the RC flow set involved in the off-network traffic and the RC flow set based on the compression state existing in the network is solved by using a particle swarm algorithm to solve the decompression scheme of the RC flow.

6. The traffic admission control and dynamic scheduling method of claim 5, wherein, The particle swarm algorithm is used to optimize the delay time rate of each RC flow affected by the TT flow as an objective, and the particle swarm algorithm is used to iteratively solve the optimal objective based on the multiple routing schemes, including: The multiple routing schemes are numbered; The particle is a combination of the routing schemes of each to-be-admitted TT flow, the particle position is described by the combination of the routing schemes of each to-be-admitted TT flow corresponding to the number in each particle, and the particle swarm is randomly initialized based on the multiple routing schemes; An adaptive function is constructed based on the delay time rate and maximum bandwidth constraint of the affected RC flow; In each iteration, an optimal solution is obtained for the particle swarm based on the adaptive function, and the target routing scheme of this iteration is obtained.

7. The traffic admission control and dynamic planning method of claim 6, wherein, The adaptive function is constructed based on the delay time rate and maximum bandwidth constraint of the affected RC flow, including: A target function is constructed based on the delay time rate of the affected RC flow; A penalty function is constructed based on the maximum bandwidth constraint; The adaptive function is constructed based on the target function, the penalty function, and the weight coefficient of the penalty function.

Images