WO2023134091A1 - 一种时间敏感网络中路径选择和门控调度的联合优化方法 - Google Patents

一种时间敏感网络中路径选择和门控调度的联合优化方法 Download PDF

Info

Publication number
WO2023134091A1
WO2023134091A1 PCT/CN2022/091458 CN2022091458W WO2023134091A1 WO 2023134091 A1 WO2023134091 A1 WO 2023134091A1 CN 2022091458 W CN2022091458 W CN 2022091458W WO 2023134091 A1 WO2023134091 A1 WO 2023134091A1
Authority
WO
WIPO (PCT)
Prior art keywords
path
tsn
transmission
flow
paths
Prior art date
Application number
PCT/CN2022/091458
Other languages
English (en)
French (fr)
Inventor
魏旻
方兴斌
尤梦飞
霍承杰
王平
徐威
Original Assignee
重庆邮电大学
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 重庆邮电大学 filed Critical 重庆邮电大学
Publication of WO2023134091A1 publication Critical patent/WO2023134091A1/zh

Links

Images

Classifications

    • HELECTRICITY
    • H04ELECTRIC COMMUNICATION TECHNIQUE
    • H04LTRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L45/00Routing or path finding of packets in data switching networks
    • H04L45/302Route determination based on requested QoS
    • H04L45/306Route determination based on the nature of the carried application
    • HELECTRICITY
    • H04ELECTRIC COMMUNICATION TECHNIQUE
    • H04LTRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L45/00Routing or path finding of packets in data switching networks
    • H04L45/22Alternate routing
    • HELECTRICITY
    • H04ELECTRIC COMMUNICATION TECHNIQUE
    • H04LTRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L45/00Routing or path finding of packets in data switching networks
    • H04L45/24Multipath
    • HELECTRICITY
    • H04ELECTRIC COMMUNICATION TECHNIQUE
    • H04LTRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L45/00Routing or path finding of packets in data switching networks
    • H04L45/302Route determination based on requested QoS
    • H04L45/306Route determination based on the nature of the carried application
    • H04L45/3065Route determination based on the nature of the carried application for real time traffic
    • HELECTRICITY
    • H04ELECTRIC COMMUNICATION TECHNIQUE
    • H04LTRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L47/00Traffic control in data switching networks
    • H04L47/10Flow control; Congestion control
    • H04L47/12Avoiding congestion; Recovering from congestion
    • YGENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02DCLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
    • Y02D30/00Reducing energy consumption in communication networks

Definitions

  • the invention belongs to the technical field of time-sensitive networks, and relates to a joint optimization method of path selection and gate control scheduling in time-sensitive networks.
  • Time-Sensitive Networking is a set of standards developed by the Time-Sensitive Networking Task Group (TSN TG) of the IEEE 802.1 working group for real-time communication over Ethernet networks with bounded delay and jitter.
  • TSN brings industrial-grade robustness to Ethernet to facilitate real-time, safety-critical applications (e.g., Industry 4.0 and cyber-physical systems) and supports time-critical and best-effort traffic over a common communications infrastructure. transmission.
  • IEEE 802.1Qbv defines a programmable gating mechanism, the time-aware shaper, which uses time-transfer gates and a gate control list (GCL) to decide which queue is selected for transmission.
  • GCL gate control list
  • the time of all devices should be synchronized based on IEEE 802.1AS-Rev to guarantee the successful deployment of time-aware shaper (TAS).
  • IEEE 802.1 Qbv clarifies the behavior of the gating mechanism, the scheduling configuration and related methods to achieve deterministic end-to-end delay still need further research.
  • TT streams are commonly seen in periodic real-time applications, such as time-sensitive control applications that need to strictly limit the maximum end-to-end delay.
  • the transmission of TT streams has periodic characteristics, and non-TT streams are used to indicate the transmission of non-periodic streams.
  • the premise of transmission scheduling is the need to specify the transmission path of each TSN flow, and most of the existing research gives the transmission path of each TSN flow in advance, which not only reduces the schedulable solution space, but also It makes it impossible to obtain the global optimal solution, and may also cause the schedulable network to obtain an unschedulable solution.
  • the purpose of the present invention is to provide a joint optimization method for path selection and gating scheduling in a time-sensitive network in order to solve the problem of transmission performance degradation caused by single-path congestion or failure, and to meet the needs of large-scale time-sensitive networks.
  • Transport needs By planning the optimal transmission path for the TT flow in the TSN network, and planning the transmission path for the non-TT flow at the same time, setting traffic transmission constraint conditions, and configuring the gating list for the optimal transmission path of the TT flow, it reduces the time-consuming factors in the transmission process of the TT flow. Traffic conflicts caused by overlapping transmission links.
  • the present invention provides the following technical solutions:
  • a joint optimization method for path selection and gating scheduling in a time-sensitive network comprising the following steps:
  • the centralized network configuration module CNC discovers the TSN network topology, and abstracts the TSN network topology into a network directed graph;
  • the terminal device sends a TSN connection request to the centralized user configuration module CUC through the user configuration protocol, and the CUC sends the connection request to the CNC through the user network interface UNI;
  • CNC selects the K alternative paths selected in step S3 using the path criticality ⁇ k to select m optimal paths;
  • the CNC takes the m optimal paths selected in step S4 as the input of the path selection stage. According to the characteristics of the TSN traffic type, based on the link transmission cost and the pheromone update method, it is a pair of TT streams from the sender to the receiver. Find an optimal transmission path, store this path in the path information table ⁇ , and find a suitable transmission path for non-TT flows;
  • step S6 CNC traverses whether there is a path that has not been calculated between the sending end and the receiving end. If there is, return to step S3-S6, and save the calculated optimal path from each pair of sending end to receiving end TT flow in the path information table ⁇ , if not, execute step S7;
  • step S7 Taking the optimal transmission path information table ⁇ of TT traffic calculated in step S6 as input, setting traffic transmission constraint conditions, configuring a gating list for the optimal transmission path of TT traffic for each pair of terminal devices;
  • the CNC encapsulates the calculation result into a gate control schedule, and configures it to the TSN switch, and then sends the flow transmission calculation result to the TSN terminal device through the CUC.
  • step S1 specifically includes: the CNC discovers the TSN network topology according to the Link Discovery Protocol (LLDP), and abstracts the TSN network topology into a network directed graph through a network modeling method;
  • LLDP Link Discovery Protocol
  • each link (BR i , BR j ) ⁇ E is a list of measurements denoted by the tuple (b, ld), where is the remaining bandwidth of the link (BR i , BR j ); is the link delay, given by and composition; is bounded;
  • the ordered data sequence that starts from the sender and is transmitted to the receiver according to certain requirements is called a stream, and the set of all TSN streams is denoted as F; for different types of streams, the main parameters include the transmission path R i of the TSN stream , the end-to-end delay D i of the TSN stream, the transmission period T i of the TSN stream, and the size S i of the TSN stream ; ,S i );
  • step S2 specifically includes: the terminal device selects the number of alternative paths K, the number of optimal paths m selected by the path criticality ⁇ k through the user configuration protocol, the maximum number of cycles N cyc of the algorithm, the maximum number of ants N ant , the total number of pheromones Quantity Q, period T i of TSN flow, size S i and delay D i are sent to CUC, and CUC sends connection request to CNC through user network interface UNI.
  • K shortest paths are selected as alternative paths, specifically: using the K shortest path algorithm KSP, for all ES i , ES′ i ⁇ H, based on their shortest path incremental sorting, through the input network
  • KSP K shortest path algorithm
  • step S4 specifically includes:
  • the hop count HC of the path p k is the number of TSN switches of the path p k except the sending end and the receiving end, defined as follows:
  • the end-to-end delay DL of path p k is its link delay The sum, expressed as;
  • ⁇ (HC, SBW, DL) is used to map its path criticality ⁇ k to a value between 0 and 1; in all alternative path sets p k , select the path with the largest ⁇ k as the path selection stage input path; for the path p k ⁇ P sd , ⁇ k is expressed as follows:
  • SBW max represents the maximum SBW among all paths p k ⁇ P sd ;
  • HC min is the smallest HC among all p k ⁇ P sd ;
  • DL min is the smallest DL among all p k ⁇ P sd ; the larger the value of ⁇ k It means that the smaller the number of hops in the network, the larger the remaining bandwidth, and the smaller the network delay;
  • the optimal transmission path is found based on the ant colony algorithm, which specifically includes the following steps:
  • S51 First set the initialization parameters, and output the initial path table R ant of the fusion path selection and gating scheduling algorithm from the pre-selection path stage; set the ant taboo table R b , the maximum number of cycles of the algorithm N cyc , the maximum number of ants N ant , and the pheromone Total Q;
  • S52 Determine the type of traffic, and set influence factors ⁇ and ⁇ , link weight factors ⁇ and ⁇ , pheromone volatilization coefficient ⁇ , and pheromone increment ⁇ according to TT flow and non-TT flow respectively.
  • the ant is placed on the sending end, and the sending end is stored in R b ;
  • S56 judge whether to reach the receiving end, if not reach the receiving end, return to S55; if arrive at the receiving end, store the path traveled by the ant in the path table R te , store the path node in the taboo table R b , Avoid intersecting with the path of the next ant, and then jump to S57;
  • represents the volatilization coefficient of pheromone, 0 ⁇ 1, ⁇ (BR i , BR j ) represents the amount of pheromone in link (BR i , BR j ), ⁇ (BR i , BR j ) represents the link (BR i , BR j ) i , BR j ) pheromone increment;
  • ⁇ (BR i , BR j ) The value of ⁇ (BR i , BR j ) is determined according to the type of TSN data flow, and its expression is as follows:
  • Q represents the total quantity of pheromones, Indicates the transmission delay of the path (BR i , BR j ), the path selection coefficient
  • S59 According to the TSN traffic type, respectively output the optimal transmission path of TT flow and non-TT flow; if it is TT flow f i , select the path with the smallest delay from the path set R as the optimal path, which can be expressed as If it is a non-TT flow f i ′, select N non-TT flow paths from the path set R as the output optimal solution, and forward them.
  • step S7 the cycle period GC of the gating list is expressed as:
  • lcm represents the least common multiple
  • T is the period of all data streams
  • T ⁇ f 0 .T 0 ,f 1 .T 1 ,...,f n .T n ⁇
  • f n .T n represents the period T n of flow f n ;
  • TT flow f i transmits offset at (BR i , BR j ) TT stream
  • the stored queue ID is (BR i , BR j ) ⁇ E, then the time taken for flow f i to transmit any data frame on the link (BR i , BR j )
  • the time slot length LOS is analyzed, and the maximum time slot length is:
  • the maximum value of LOS is the greatest common divisor of the data flow period
  • the minimum slot length is:
  • LOG is the length of the transmission queue, That is, when all queue resources are occupied, the time it takes for the last 1 byte to be sent to the link;
  • the transmission path is ⁇ ES i ,BR 1 ,BR 2 ,...,BR n ,ES′ i ⁇ , the GCL cycle start time of switch BR n
  • the calculation formula is as follows:
  • BR n Denotes the processing delay in BR n , Denotes the transmission delay in BR n-1 , Indicates the propagation delay in (BR n-1 , BR n );
  • the end-to-end delay constraint is expressed as:
  • the frame isolation constraint without transmission duration is expressed as:
  • the frame isolation constraint with transmission duration is expressed as:
  • the traffic transmission constraints are expressed as:
  • the optimization goal is:
  • the beneficial effect of the present invention is that: the present invention aims at the transmission characteristics of TT flow and non-TT flow in TSN, selects an optimal path in the network to ensure the transmission of TT flow first, and also plans the transmission path for non-TT flow, thereby realizing Better network transmission performance.
  • FIG 1 is an architecture diagram of the TSN network
  • Fig. 2 is the flowchart of CNC execution fusion path selection and gating scheduling algorithm
  • Fig. 3 is the shortest path tree
  • Fig. 4 is a flow chart of the shortest path algorithm
  • Figure 5 is a delay analysis diagram
  • Figure 6 is an example diagram of frame isolation constraints
  • Figure 1 describes the architecture of the TSN network proposed in this scheme. Its workflow is as follows, when the sender wants to send a TSN stream to the receiver.
  • the sender uses a user configuration protocol (such as OPC UA) to send a TSN connection request to the centralized user configuration module (CUC), and the request information includes information such as TSN network topology and traffic transmission requirements;
  • a user configuration protocol such as OPC UA
  • CRC centralized user configuration module
  • the CUC converts the TSN connection request into a TSN connection requirement, and forwards it to the centralized network configuration module (CNC) through the user network interface (UNI);
  • CNC sends the transmission schedule to CUC, and forwards it to the network sending end;
  • the sending end starts to transmit the TSN flow to the receiving end along the selected path according to the calculated transmission schedule.
  • the innovation of the method proposed in this scheme is to propose a fusion path selection and gating scheduling algorithm for the CNC path calculation and traffic scheduling in the above step (3).
  • the execution of this method is divided into four stages, which are respectively the network modeling stage, Alternate Paths Phase, Path Selection Phase, and Gating Design Phase.
  • the proposed method is outlined as follows:
  • CNC performs network modeling according to network topology and TSN traffic characteristics, which is the basis for normal execution of subsequent stages;
  • the CNC In the path selection phase, when the CNC receives a connection request from the CUC, the CNC first selects K candidate paths using the fusion path selection and gating scheduling algorithm according to the network topology, and then selects m (m ⁇ K ) links, using these m paths as the input of the path selection stage;
  • CNC calculates the optimal transmission path for each pair of TT traffic from the sender to the receiver based on the characteristics of TT traffic and non-TT traffic in TSN traffic, based on the cost of link transmission and the way of pheromone update. And use other valid paths as transmission paths for non-TT traffic;
  • CNC executes the fusion path selection and gating scheduling algorithm to calculate the gating scheduling list for the transmission of TT traffic, and sends the calculation result to the TSN switch .
  • the centralized user configuration management system includes two parts, CNC and CUC.
  • the CNC includes a TSN network topology discovery module, a TSN network configuration module, a TSN network update module, a path selection module and a gating calculation module.
  • the path selection module and the gate The control calculation module jointly executes the fusion path selection and gating scheduling algorithm.
  • the TSN network topology discovery module mainly generates the network topology of the TSN network.
  • the main function of the TSN network configuration module is to configure the TSN domain device data and add the configuration data to the database.
  • the main function of the TSN network update module is to update all device configuration data in the TSN domain, and save the update operation in the database.
  • CNC perceives the resource information of the TSN network through the NETCONF/YANG model, and sends configuration information.
  • the CUC delivers the TSN flow configuration information to the terminal devices (sender and receiver) on the TSN network through the user configuration protocol.
  • the CNC uses the link discovery protocol (LLDP) to discover the TSN network topology.
  • the terminal device will send the TSN network connection request to the CUC through the user configuration protocol; the CUC converts the TSN connection request into TSN
  • the connection requirements are forwarded to the centralized network configuration module (CNC) through the user network interface (UNI); at this time, the CNC calculates the optimal transmission of each pair of TT streams from the sender to the receiver according to the TSN network topology information and connection requirements path and non-TT flow transmission path, in order to ensure the traffic conflict caused by transmission link overlap during TT flow transmission, this solution sets traffic transmission constraint conditions, and configures a gating list for the optimal transmission path of TT flow, so as to To reduce the problem of traffic conflict, the CNC then uses the NETCONF protocol to send the scheduling table to the TSN switch to complete the configuration management of the TSN network.
  • LLDP link discovery protocol
  • the first step CNC discovers the TSN network topology according to the Link Discovery Protocol (LLDP), and abstracts the TSN network topology into a network directed graph through the network modeling method;
  • LLDP Link Discovery Protocol
  • the second step the terminal device selects the number of optimal paths m , the maximum number of cycles N cyc of the algorithm, the maximum number of ants N ant and the total amount of pheromones Q,
  • the period, size and delay D i of the TSN flow are sent to the CUC, and the CUC sends the connection request to the CNC through the user network interface UNI;
  • Step 3 When the CNC receives the request for path selection, in order to increase the diversity of path selection samples, the CNC first performs the fusion path selection and gating scheduling algorithm to select K shortest paths as alternative paths;
  • CNC utilizes path criticality ⁇ k to select m (m ⁇ K) optimal path with the K alternative paths selected in the 3rd step;
  • Step 5 The CNC takes the m optimal paths obtained in the fourth step as the input of the path selection stage.
  • the link transmission cost and the pheromone update method are a pair of sending end to receiving end Find an optimal transmission path for the TT flow at the end, and store the path in the path information table ⁇ , and find a suitable transmission path for the non-TT flow at the same time;
  • Step 6 CNC traverses whether there is a path that has not been calculated between the sending end and the receiving end. If there is, return to step 3 to step 6, and save the calculated optimal path from each pair of sending end to receiving end TT flow. to the path information table ⁇ , if not, execute the seventh step;
  • Step 7 In the gating design stage, in order to avoid traffic conflicts caused by overlapping transmission links during the transmission of TT streams, the optimal transmission path information table ⁇ of TT traffic calculated in the sixth step is used as the gating design In the input stage, set traffic transmission constraints, and configure a gating list for the optimal transmission path of TT streams for each pair of terminal devices to ensure reliable transmission of TT streams;
  • Step 8 CNC encapsulates the calculation result into an XML file, uses the NETCONF protocol to configure the XML-based gating schedule to the TSN switch, and sends the flow transmission calculation result to the TSN terminal device through the CUC.
  • the first step CNC discovers the TSN network topology according to the Link Discovery Protocol (LLDP), and abstracts the TSN network topology into a network directed graph through the network modeling method;
  • LLDP Link Discovery Protocol
  • the main task of TSN path selection is to find a suitable transmission path for the TSN flow and schedule the transmission of the TT flow through the gate control list (GCL).
  • GCL gate control list
  • Factors to be considered in this process are network topology and TSN traffic model.
  • S is a collection of TSN switches
  • H is a collection of terminal devices.
  • E is a set of 2-tuples representing links in the network such that E ⁇ (BR i , BR j )
  • each link (BR i , BR j ) ⁇ E is a list of measurements denoted by the tuple (b, ld), where is the remaining bandwidth of the link (BR i , BR j ), in Mbps, is the link delay in microseconds, which is given by and composition. It should be noted is bounded.
  • Each TSN flow F i is represented as a quadruple F i ⁇ (R i , D i , T i , S i ).
  • Step 2 The terminal device selects the number of optimal paths m , the maximum number of cycles N cyc , the maximum number of ants N ant , the total amount of pheromones Q, The period T i , size S i and delay D i of the TSN flow are sent to the CUC, and the CUC sends the connection request to the CNC through the user network interface UNI.
  • Step 3 When the CNC receives the request for path selection, in order to increase the diversity of path selection samples, the CNC first performs the fusion path selection and gating scheduling algorithm to select K shortest paths as alternative paths.
  • K alternative paths P k between a pair of ES i and ES′ i are obtained through the fusion path selection and gating scheduling algorithm.
  • the principle is to use the K shortest path algorithm (KSP). For all ES i , ES ′ i ⁇ H, based on their shortest paths are sorted incrementally. To find these paths, input the network directed graph G, the sending end ES i , the receiving end ES′ i and the number of paths K, and output K path sets p K , and then take the path set p K as the input of the fourth step.
  • KSP K shortest path algorithm
  • the set of paths from ES i to ES′ i is denoted by P sd
  • K paths are the shortest, that is, for all p ⁇ P sd -P k , c(p k ) ⁇ c(p);
  • the construction of the shortest path tree T k is based on an important concept—deviation path.
  • (u x ,u x+1 ) is called the side of p j ’s deviation from p i
  • u x is the node of p j ’s deviation from p i
  • the deviation node of p 2 relative to p 1 is node 1
  • the deviation edge is edge (1,3)
  • the deviation path is (1,3,4,5).
  • the value of represents the weight, and the weight of a path is equal to the sum of the delays of each link in the path.
  • the specific execution steps are as follows:
  • CNC utilizes path criticality ⁇ k to select m (m ⁇ K) optimal path with the K alternative paths selected in the 3rd step;
  • this scheme uses the path criticality ⁇ k to quantify and calculate each path in the path table.
  • K candidate path sets P k are input and m optimal paths are output Set R ant (m ⁇ K).
  • Each path p k has its own path criticality index ⁇ (HC, SBW, DL) . ) equals the number of TSN switches, p k .
  • the end-to-end delay DL (denoted by p k .DL) of the path p k is equal to its link delay
  • the sum can be expressed as;
  • the function ⁇ (HC,SBW,DL) is used to map its path criticality ⁇ k (denoted by p k . ⁇ (HC,SBW,DL)) to a value between 0 and 1.
  • ⁇ k is expressed as follows.
  • SBW max represents the maximum SBW among all paths p k ⁇ P sd ;
  • HC min is the smallest HC among all p k ⁇ P sd , and
  • DL min is the smallest DL among all p k ⁇ P sd .
  • Step 5 The CNC takes the m optimal paths obtained in the fourth step as the input of the path selection stage.
  • the link transmission cost and the pheromone update method are a pair of sending end to receiving end Find an optimal transmission path for the TT flow at the end, and store the path in the path information table ⁇ , and find a suitable transmission path for the non-TT flow.
  • the principle of performing path selection function in the fusion path selection and gating scheduling algorithm is based on the ant colony algorithm.
  • the ant colony algorithm (ACO) is a swarm intelligence algorithm. Collaboratively exhibit intelligent behavior, thus providing a new possibility for solving complex problems.
  • the transmission probability of ant a from node BR i to BR j is:
  • ⁇ (BR i , BR j ) The value of ⁇ (BR i , BR j ) is determined according to the type of TSN data flow, and its expression is as follows.
  • this solution sets the link transmission cost and pheromone update method in the fusion path selection and gating scheduling algorithm respectively, and reduces the packet loss rate by calculating the optimal transmission path of TT flow and non-TT flow.
  • This algorithm outputs the transmission path of R TT flow and non-TT flow f i ′ by inputting R ant , R b , N cyc , N ant and Q.
  • the specific algorithm flow is as follows.
  • TSN traffic type output the optimal transmission path of TT flow and non-TT flow respectively. If it is a TT flow f i , choose the path with the smallest delay from the path set R as the optimal path, which can be expressed as If it is a non-TT flow f i ′, select N non-TT flow paths from the path set R as the output optimal solution, and forward them.
  • Step 6 CNC traverses whether there are senders and receivers that have not yet been calculated, and if so, execute steps 3 to 6, and save the calculated optimal path from each pair of sender to receiver TT flow in In the path information table ⁇ , if not, execute the seventh step.
  • the network topology applicable to this solution is a multi-input multi-output network
  • the subsequent gating scheduling mechanism is only for TT flows
  • the obtained The optimal path is also the optimal path for the TT flow. It is necessary to repeat the third step to the sixth step to calculate the optimal path of each pair of receiving end and sending end and store them in the path information table ⁇ .
  • the optimal path of each TT flow in the path information table ⁇ will be calculated as a gating list and scheduled. Table 3 illustrates the calculation results of the path selection phase.
  • the IEEE 802.1Qbv standard specifies a gating mechanism called the Time-Aware Shaper (TAS), which enables and disables the gates to be connected to the associated egress ports based on a statically generated periodic schedule called the Gating Control List.
  • TAS Time-Aware Shaper
  • Queue selection Communication in a network is accomplished by periodically sending a stream of data from a sender to a receiver. The communication path of the data flow in the network at this stage has been calculated in the path selection stage.
  • TT traffic is the most delay-sensitive TSN flow category. In order to ensure the traffic conflict problem caused by transmission link overlap during TT stream transmission, this scheme uses the deadline, traffic transmission constraints and the optimal transmission path information table of TT stream ⁇ , construct periodic GCL scheduling for TT flow to reduce the delay impact caused by traffic conflict.
  • Link (BR i ,BR j ) ⁇ E represents a communication direction, so a pair of links (BR i ,BR j ), (BR j ,BR i ) ⁇ E represents the communication between node BR i and node BR j Full-duplex physical links, from the perspective of scheduling, these two links are two different resources. It is assumed that all devices in the network are time synchronized. The worst-case synchronization error, the maximum difference between the local clocks of any two devices in the network, is denoted as synPre.
  • T n of f n , and the cycle period GC of the gating list is equal to the least common multiple (lcm) of all flows, which can be expressed as:
  • the TT flow of TSN may appear multiple times within a period of a gating list.
  • This solution adopts the integrated path selection and gating scheduling algorithm to search for the transmission path of the calculation flow.
  • This algorithm can calculate the optimal path for TT flow transmission. The focus of this section is on the scheduling of TT flow. If flow f i passes link (BR i , BR j ), then in Can be an empty set.
  • Step 7 In the gating design stage, in order to avoid traffic conflicts caused by overlapping transmission links during the transmission of TT streams, the optimal transmission path information table ⁇ of TT traffic calculated in the sixth step is used as the gating design
  • the input of the stage sets traffic transmission constraints, and configures a gating list for the optimal transmission path of TT streams for each pair of terminal devices to ensure reliable transmission of TT streams.
  • the time slot length LOS is analyzed. Since the data is offset with the time slot as the basic unit, it is assumed that the LOS can evenly divide the period of all data streams. Then the maximum slot length is:
  • the maximum value of LOS is the greatest common divisor of the period of the data flow.
  • the TSN standard stipulates that the sending time slots and receiving time slots of packets at two adjacent nodes are the same. Then, the length of the time slot needs to at least ensure that the sending time slot of the last message in the queue is the same as the receiving time slot between adjacent nodes, so the minimum time slot length is:
  • the TT stream is transmitted from the sender to the receiver along the calculated optimal path.
  • the end-to-end delay includes propagation delay, processing delay and transmission delay.
  • the delay analysis is shown in Figure 5, where, t n represents the sending time of the switch BR i , and the expression of D is as follows.
  • End-to-end delay constraint the end-to-end delay of flow f i from the sending end to the receiving end through j switches should be less than or equal to D i , let (ES i , BR j ) and (BR t , ES′ i ) denote respectively
  • the first and last link of the transmission path of flow f i can be expressed as:
  • Frame Constraint This constraint ensures that transmission of a frame on the first link of the transmission path cannot start before the send time and must be completed on the last link before the deadline.
  • Link constraint The transmission of two different frames on a link cannot overlap in time. That is, for each pair of distinct frames on the same link, the transmission of one frame must complete before the transmission of the other frame begins, and vice versa. This constraint must take into account all frame occurrences in the cycle cycle GC. Due to the zero-jitter requirement, the transmission offset occurring at the kth frame of flow f i on link (BR i , BR j ) ⁇ E is equal to Therefore, link constraints can be written as follows:
  • Frame isolation constraints In order to avoid the problem of frame loss during streaming, the solution proposed in this scheme is to increase frame isolation constraints.
  • the purpose of the frame isolation constraint without transmission duration is to isolate two non-identical frames such that one frame can only be sent to the shared queue after the other frame has been scheduled from the queue.
  • There are three ways to satisfy this constraint the first is to schedule f i on the link (BR a , BR b ) before f j in switch BR a arrives, the second is to schedule f i in switch BR a Scheduling f j on the link (BR a , BR b ) before arrival, the third method is to store frames in different queues, it should be noted that all frames that appear in the scheduling period must be considered, and thus ⁇ and ⁇ multiples.
  • constraint assumes that the moment at which a switch stores a frame in a queue does not depend on the frame's transmission duration. But if the frame is stored in a queue after it has been fully received by the switch, constraint (25) can be replaced by a constraint including transmission duration as described below.
  • a frame isolation constraint with a duration is shown below.
  • Traffic Delivery Constraint This constraint models the precedence relationship, i.e. the transmission of a frame from the switch cannot begin until the frame has been fully delivered to the switch and processed.
  • optimization objective Consider the following objective function to minimize the maximum end-to-end delay associated with the flow period, also known as response time, in order to ensure the reliability of traffic transmission and ensure that frames are transmitted to the receiving end within the transmission period.
  • This variable represents the ratio of the end-to-end delay of any flow to its period. This solution requires the value of this variable to be in [0,1], which means that the scheduling is optimal.
  • the expression is as follows.
  • Step 8 CNC encapsulates the calculation result into an XML file, uses the NETCONF protocol to configure the XML-based gating schedule to the TSN switch, and sends the flow transmission calculation result to the TSN terminal device through the CUC.

Landscapes

  • Engineering & Computer Science (AREA)
  • Computer Networks & Wireless Communication (AREA)
  • Signal Processing (AREA)
  • Data Exchanges In Wide-Area Networks (AREA)

Abstract

本发明涉及一种时间敏感网络中路径选择和门控调度的联合优化方法,属于时间敏感网络技术领域,包括以下步骤:S1:CNC发现TSN网络拓扑,并抽象为网络有向图;S2:终端设备向CUC发送TSN连接请求,CUC将连接请求发送到CNC;S3:CNC选出K条最短路径作为备选路径;S4:CNC选择出m条优选路径;S5:CNC为TT流找到一条最优的传输路径,为非TT流找到合适的传输路径;S6:CNC遍历完成;S7:为每一对终端设备的TT流最优传输路径配置门控列表;S8:CNC将计算的结果封装为门控调度表,并配置到TSN交换机,再将流量传输计算结果通过CUC发送到TSN终端设备。

Description

一种时间敏感网络中路径选择和门控调度的联合优化方法 技术领域
本发明属于时间敏感网络技术领域,涉及一种时间敏感网络中路径选择和门控调度的联合优化方法。
背景技术
随着工业互联网技术的不断发展,确定性实时通信被引入到航天和工业自动化领域,以保证时间关键流的性能和系统的安全性。时间敏感网络(TSN)是由IEEE 802.1工作组的时间敏感网络任务组(TSN TG)开发的一套标准,用于在有界延迟和抖动的以太网网络上进行实时通信。TSN为以太网带来了工业级的健壮性,以促进实时、安全关键的应用(例如,工业4.0和网络物理系统),并支持在一个公共通信基础设施上对时间关键流量和最努力流量进行传输。值得注意的是,IEEE 802.1Qbv定义了一个可编程的门控机制,即时间感知的整形器,它使用时间传输门和门控制列表(GCL)来决定哪个队列被选择用于传输。同时,所有设备的时间都应基于IEEE 802.1AS-Rev同步,以保证时间感知整形器(TAS)部署成功。然而,IEEE 802.1 Qbv虽然明确了门控机制的行为,但实现确定性端到端延迟的调度配置和相关方法还需要进一步的研究。
在时间敏感网络中存在多种数据流,其中包括TT流和非TT流(音频流,视频流,BE(Best Effort)流)。TT流常见于周期性的实时应用程序中,例如需要严格限制最大端到端延迟的时间敏感型控制应用程序。TT流的传输具有周期性特点,非TT流用于表示传输不具有周期性特点的流。针对时间敏感网络中TT流的确定性实时传输调度问题,目前最常见的建模方法主要是整数线性规划方法(Integer Linear programming,ILP)、可满足性模理论(Satisfiability Modulo Theories,SMT)方法和优化模理论(Optimization Modulo Theories,OMT)等方法,ILP方法和SMT/OMT方法都是通过构造一系列约束以实现TT流的确定性实时通信传输,两者之间的区别在于SMT方法构造的约束表达式是具有相应理论的一阶逻辑公式。但是,在复杂的网络拓扑中,传输调度的前提是需要明确各个TSN流的传输路径,而现有的研究大多提前给定各个TSN流的传输路径,这个不仅减小了可调度的解空间,使得不能获取全局的最优解,还可能会导致可调度的网络得到不可调度的解。
综上所述,在多路径多输入输出的时间敏感网络中如何提高网络的运行效率,路径规划和门控调度联合优化的设计方案在TSN标准的调度机制中还没有得到解决,且国内外对该方 面的研究也相对较少。
发明内容
有鉴于此,本发明的目的在于为了解决单路径的拥塞或者故障引起的传输性能下降的问题,提供一种时间敏感网络中路径选择和门控调度的联合优化方法,满足大规模时间敏感网络的传输需求。通过为TSN网络中TT流规划最优传输路径,同时也为非TT流规划传输路径,设置流量传输约束条件,为TT流的最优传输路径配置门控列表,减小TT流传输过程中因传输链路重叠造成的流量冲突问题。
为达到上述目的,本发明提供如下技术方案:
一种时间敏感网络中路径选择和门控调度的联合优化方法,包括以下步骤:
S1:集中式网络配置模块CNC发现TSN网络拓扑,并将TSN网络拓扑抽象为网络有向图;
S2:终端设备通过用户配置协议向集中式用户配置模块CUC发送TSN连接请求,CUC将所述连接请求通过用户网络接口UNI发送到CNC;
S3:CNC收到路径选择的需求时,执行融合路径选择与门控调度算法选出K条最短路径作为备选路径;
S4:CNC将步骤S3中选出的K条备选路径利用路径关键度η k来选择出m条优选路径;
S5:CNC将步骤S4中选出的m条优选路径作为路径选择阶段的输入,根据TSN流量类型的特征,基于链路传输的代价和信息素更新的方式为一对发送端到接收端的TT流找到一条最优的传输路径,并将该条路径存入路径信息表ω中,同时为非TT流找到合适的传输路径;
S6:CNC遍历是否存在还没有被计算发送端和接收端的路径,如果有则返回执行步骤S3-S6,并将计算的每对发送端到接收端TT流的最优路径存到路径信息表ω中,如果没有则执行步骤S7;
S7:将步骤S6中计算出的TT流量的最优传输路径信息表ω作为输入,设置流量传输约束条件,为每一对终端设备的TT流最优传输路径配置门控列表;
S8:CNC将计算的结果封装为门控调度表,并配置到TSN交换机,再将流量传输计算结果通过CUC发送到TSN终端设备。
进一步,步骤S1具体包括:CNC根据链路发现协议(LLDP)来发现TSN网络拓扑,并通过网络建模方法将TSN网络拓扑抽象为网络有向图;
TSN的网络拓扑表示为有向图G=(V,E),其中V是TSN网络中的节点集合,
Figure PCTCN2022091458-appb-000001
其中S是TSN交换机的集合,H是终端设备的集合;E是边集合,是一组二元组,表示TSN 网络中的所有链接,使得
E≡{(BR i,BR j)|BR i,BR j∈V,BR i≠BR j并且BR i和BR j之间存在联系},其中(BR i,BR j)表示交换机BR i和交换机BR j之间的链路;
与每个链路(BR i,BR j)∈E相关联的是由元组(b,ld)表示的测量值列表,其中
Figure PCTCN2022091458-appb-000002
是链路(BR i,BR j)的剩余带宽;
Figure PCTCN2022091458-appb-000003
是链路延迟,由
Figure PCTCN2022091458-appb-000004
Figure PCTCN2022091458-appb-000005
组成;
Figure PCTCN2022091458-appb-000006
是有界的;
把从发送端开始,按照一定要求传输到接收端的有序数据序列称为流,将所有的TSN流的集合记为F;对于不同类型的流,其中的主要参数包括TSN流的传输路径R i、TSN流的端到端时延D i、TSN流的传输周期T i和TSN流的大小S i;每个TSN流F i表示为四元组F i≡(R i,D i,T i,S i);
对于发送端ES i和接收端ES′ i,中间经过的交换机BR 1,BR 2,…,BR n共有n个节点的流,将其第i对发送端到接收端的路径表示为R i={ES i,BR 1,BR 2,…,BR n,ES′ i},每个帧的最大长度为以太网的最大传输单元MTU。
进一步,步骤S2具体包括:终端设备通过用户配置协议将备选路径数K、路径关键度η k选出的优选路径数量m、算法最大循环次数N cyc、蚂蚁的最大数量N ant、信息素总量Q、TSN流的周期T i、大小S i和时延D i发送给CUC,CUC将连接请求通过用户网络接口UNI发送到CNC。
进一步,步骤S3中所述选出K条最短路径作为备选路径,具体为:采用K最短路径算法KSP,对于所有ES i,ES′ i∈H,基于它们的最短路径递增排序,通过输入网络有向图G、发送端ES i、接收端ES′ i和路径的条数K,输出K条路径集p K,然后将路径集p K作为步骤S4的输入;具体执行步骤如下:
S31:输入网络有向图G、发送端ES i、接收端ES′ i和备选路径数K;
S32:CNC使用融合路径选择与门控调度算法求出ES i到ES′ i的最短路径,并记为p n(n=1):
p n=ES i→BR a→BR b→…→BR n→ES′ i
S33:判断n是否小于等于K,且仍有候选路径存在,如果存在执行S34,否则表示找到K条备选路径;
S34:在求出p n的基础上,根据偏离点和Dijkstra算法求出p n+1,把位于p n上除去接收端ES′ i的每个节点分别看作偏离点,共有x个;将每个偏离点记为BR i(i=1,2,…,x);
S35:开始遍历偏离点,从BR i(i=0)开始遍历每一个偏离点,并对每一个偏离点求BR i到接收端ES′ i的最短路径;
S36:将p n上从起点到BR i的路径加上求得的BR i到接收端ES′ i的最短路径作为求p n+1的候选路径,放到候选路径列表U中;
S37:偏离点遍历结束;
S38:判断此时候选路径列表U是否为空;
S39:如果此时候选路径列表U不为空,则遍历完偏离点后,找出B中权值最小的路径即为所求的p n+1,将该路径从U中移除,并放入备选路径列表L中,在求得p n+1的基础之上,当n+1≤K时,继续执行步骤S33-S38,否则表示找到K条备选路径;
S310:若候选路径列表U为空,则表示已经找到了K条备选路径,最后确定路径集合为P k={p 1,p 2,…,p n,p k}∈P sd,P sd表示从ES i到ES′ i的路径集合。
进一步,步骤S4具体包括:
路径p k的跳数HC为除发送端和接收端之外的路径p k的TSN交换机数量,定义如下:
p k.HC=len(p k)-2         (1)
路径p k的剩余带宽SBW为其所有组成链路中的剩余带宽的最小值,假设p k={ES i,BR 1,BR 2,…,BR n,ES′ i},令S=len(p k)-2,路径p k的剩余带宽SBW由如下公式表示:
Figure PCTCN2022091458-appb-000007
路径p k的端到端时延DL为其链路时延
Figure PCTCN2022091458-appb-000008
之和,表示为;
Figure PCTCN2022091458-appb-000009
函数Δ(HC,SBW,DL)用于映射其路径关键度η k为一个介于0和1之间的值;在所有备选路径集p k中,选择η k最大的路径作为路径选择阶段的输入路径;对于路径p k∈P sd,η k表示如下:
Figure PCTCN2022091458-appb-000010
ω 123=1         (5)
其中
Figure PCTCN2022091458-appb-000011
SBW max表示所有路径p k∈P sd中的最大值SBW;HC min是所有p k∈P sd中最小的HC,DL min是所有p k∈P sd中最小的DL;η k的值越大则表示网络的跳数越小,剩余带宽越大,网络延迟越小;
通过以上方式计算每一条路径p k的路径关键度η k的值,选取前m条η k值最大的路径作为优选路径,并存入优选路径集R ant中。
进一步,所述步骤S5中,基于蚁群算法找到最优传输路径,具体包括以下步骤:
S51:先设置初始化参数,从预选路径阶段输出融合路径选择和门控调度算法的初始路径表R ant;设置蚂蚁禁忌表R b,算法最大循环次数N cyc,蚂蚁的最大数量N ant,信息素总量Q;
S52:判断流量类型,根据TT流和非TT流分别设置影响因子α和β,链路权重因子δ和ε,信息素挥发系数ρ,信息素增量Δτ。蚂蚁置于发送端,并将发送端存入R b中;
S53:循环次数n cyc=n cyc+1;
S54:蚂蚁数量n ant=n ant+1;
S55:蚂蚁根据公式(6)选择下一个节点:
Figure PCTCN2022091458-appb-000012
Figure PCTCN2022091458-appb-000013
其中,P(BR i,BR j)表示蚂蚁a从节点BR i到BR j的传输概率,allow a表示从节点到下一个节点的集合,α越大,信息素的指导作用越强,β越大,蚂蚁决策受路径距离信息影响越大,贪婪当前效果;δ和ε为权重因子且δ+ε=1,0<δ<1,0<ε<1,δ和ε的取值根据当前流量类型而定,当数据流为TT流时,δ的取值小于ε;τ(BR i,BR j)表示链路(BR i,BR j)的信息素数量;μ(BR i,BR j)表示节点选择的启发因子;
Figure PCTCN2022091458-appb-000014
表示链路(BR i,BR j)的剩余带宽;
Figure PCTCN2022091458-appb-000015
表示链路(BR i,BR j)的延迟;
S56:判断是否到达接收端,如果未达到接收端,返回S55;如果到达接收端,将该蚂蚁的走过的路径存入路径表R te中,将该路径节点存入禁忌表R b中,避免与下一只蚂蚁的路径产生交叉,然后跳转至S57;
S57:判断蚂蚁数量n ant是否等于N ant,如果n ant<N ant,则返回S54,如果n ant=N ant,从R te中选取传输代价TV最小的路径作为该次循环得出的最优路径存入最佳路径集合R中,跳转至S58;
S58:判断n cyc是否等于N cyc,如果n cyc≠N cyc,返回S53,根据公式(8)更新链路中的信息素,清空R b;如果n cyc=N cyc,则跳转到S59;
Figure PCTCN2022091458-appb-000016
ρ表示信息素的挥发系数,0<ρ<1,τ(BR i,BR j)表示链路(BR i,BR j)的信息素数量,Δτ(BR i,BR j)表示链路(BR i,BR j)的信息素增量;
Δτ(BR i,BR j)取值根据TSN数据流类型而确定,其表达式如下所示:
Figure PCTCN2022091458-appb-000017
Q表示信息素的总数量,
Figure PCTCN2022091458-appb-000018
表示路径(BR i,BR j)的传输延迟,路径选择系数
Figure PCTCN2022091458-appb-000019
的表达式如下所示:
Figure PCTCN2022091458-appb-000020
S59:根据TSN流量类型,分别输出TT流和非TT流的最优传输路径;如果是TT流f i,从路径集合R中选择延迟最小的路径作为最优路径,可表示为
Figure PCTCN2022091458-appb-000021
如果是非TT流f i′,从路径集合R中选择N 非TT流条路径作为输出最优解,并且进行转发。
进一步,步骤S7中,门控列表的循环周期GC表示为:
GC=lcm(T)          (11)
其中,lcm表示最小公倍数,T为所有数据流的周期,
T={f 0.T 0,f 1.T 1,…,f n.T n}
其中f n.T n表示流f n的周期T n
假设流f i∈F的每个后续帧出现的时间距离始终等于T i,TT流f i在(BR i,BR j)传输偏移量
Figure PCTCN2022091458-appb-000022
TT流
Figure PCTCN2022091458-appb-000023
存储的队列ID为
Figure PCTCN2022091458-appb-000024
(BR i,BR j)∈E,则流f i在链路(BR i,BR j)上的传输任意数据帧所用的时间
Figure PCTCN2022091458-appb-000025
表示为:
Figure PCTCN2022091458-appb-000026
根据GCL的传输规则,对时隙长度LOS进行分析,最大的时隙长度为:
MAX(LOS)=GCD(T)           (13)
即LOS的最大值为数据流周期的最大公约数;
最小的时隙长度为:
Figure PCTCN2022091458-appb-000027
其中,LOG是传输队列的长度,
Figure PCTCN2022091458-appb-000028
即队列资源全部被占用时,最后1byte发送到链路上所花费的时间;
传输路径为{ES i,BR 1,BR 2,…,BR n,ES′ i},交换机BR n的GCL循环开始时间
Figure PCTCN2022091458-appb-000029
的计算公式如下:
Figure PCTCN2022091458-appb-000030
Figure PCTCN2022091458-appb-000031
表示发送端发送第一个数据帧的发送时间,
Figure PCTCN2022091458-appb-000032
表示流f i在链路(ES i,BR 1)上的传输任意数据帧所用的时间,
Figure PCTCN2022091458-appb-000033
表示BR i中的处理延迟,
Figure PCTCN2022091458-appb-000034
表示BR i中的传输延迟,
Figure PCTCN2022091458-appb-000035
表示流f i在链路(BR j-1,BR j)上的传输任意数据帧所用的时间,synPre表示两个节点设备的时间同步最大差值;
一个交换机到下一个交换机之间的时延D的表达式如下:
Figure PCTCN2022091458-appb-000036
Figure PCTCN2022091458-appb-000037
表示BR n中的处理延迟,
Figure PCTCN2022091458-appb-000038
表示BR n-1中的传输延迟,
Figure PCTCN2022091458-appb-000039
表示(BR n-1,BR n)中的传播延迟;
端到端的时延约束表示为:
Figure PCTCN2022091458-appb-000040
帧约束表示为:
Figure PCTCN2022091458-appb-000041
Figure PCTCN2022091458-appb-000042
链接约束表示为:
Figure PCTCN2022091458-appb-000043
无传输持续时间的帧隔离约束表示为:
Figure PCTCN2022091458-appb-000044
其中,
Figure PCTCN2022091458-appb-000045
Figure PCTCN2022091458-appb-000046
Figure PCTCN2022091458-appb-000047
Figure PCTCN2022091458-appb-000048
带有传输持续时间的帧隔离约束表示为:
Figure PCTCN2022091458-appb-000049
流量传输约束表示为:
Figure PCTCN2022091458-appb-000050
优化目标为:
Figure PCTCN2022091458-appb-000051
引入辅助变量
Figure PCTCN2022091458-appb-000052
表示任何流的端到端延迟与其周期之比,要求该变量的值在[0,1]中即表示调度为最优;
通过将以上的约束条件进行建模并使用求解器求解,根据每一对终端设备中TT流的最优传输路径
Figure PCTCN2022091458-appb-000053
计算出TT流量传输路径上交换机的门控列表。
本发明的有益效果在于:本发明针对TSN中TT流和非TT流的传输特征,在网络中选择一条最优路径优先保证TT流量的传输,同时也为非TT流量规划传输路径,从而可以实现更好的网络传输性能。
本发明的其他优点、目标和特征在某种程度上将在随后的说明书中进行阐述,并且在某种程度上,基于对下文的考察研究对本领域技术人员而言将是显而易见的,或者可以从本发明的实践中得到教导。本发明的目标和其他优点可以通过下面的说明书来实现和获得。
附图说明
为了使本发明的目的、技术方案和优点更加清楚,下面将结合附图对本发明作优选的详细描述,其中:
图1为TSN网络的体系架构图;
图2为CNC执行融合路径选择与门控调度算法流程图;
图3为最短路径树;
图4为最短路径算法流程图;
图5为时延分析图;
图6为帧隔离约束示例图;
具体实施方式
以下通过特定的具体实例说明本发明的实施方式,本领域技术人员可由本说明书所揭露的内容轻易地了解本发明的其他优点与功效。本发明还可以通过另外不同的具体实施方式加以实施或应用,本说明书中的各项细节也可以基于不同观点与应用,在没有背离本发明的精神下进行各种修饰或改变。需要说明的是,以下实施例中所提供的图示仅以示意方式说明本发明的基本构想,在不冲突的情况下,以下实施例及实施例中的特征可以相互组合。
其中,附图仅用于示例性说明,表示的仅是示意图,而非实物图,不能理解为对本发明的限制;为了更好地说明本发明的实施例,附图某些部件会有省略、放大或缩小,并不代表实际产品的尺寸;对本领域技术人员来说,附图中某些公知结构及其说明可能省略是可以理解的。
本发明实施例的附图中相同或相似的标号对应相同或相似的部件;在本发明的描述中,需要理解的是,若有术语“上”、“下”、“左”、“右”、“前”、“后”等指示的方位或位置关系为基于附图所示的方位或位置关系,仅是为了便于描述本发明和简化描述,而不是指示或暗示所指的装置或元件必须具有特定的方位、以特定的方位构造和操作,因此附图中描述位置关系的用语仅用于示例性说明,不能理解为对本发明的限制,对于本领域的普通技术人员而言,可以根据具体情况理解上述术语的具体含义。
图1描述了本方案提出的TSN网络的体系结构。其工作流程如下,当发送端要向接收端发送TSN流时。
(1)发送端使用用户配置协议(例如OPC UA)向集中式用户配置模块(CUC)发送TSN连接请求,该请求信息包括了TSN网络拓扑、流量传输要求等信息;
(2)CUC将TSN连接请求转换为TSN连接需求,并通过用户网络接口(UNI)转发给集中式网络配置模块(CNC);
(3)CNC在收到TSN连接要求时,它使用本方案提出的融合路径选择与门控调度算法来计算满足连接要求的最优传输路径和可行的传输调度;
(4)如果存在这样的路径和调度表,则CNC用所计算的调度表配置沿所计算的路径的所有TSN交换机;
(5)CNC将传输时间表发送给CUC,将其转发到网络发送端;
(6)发送端开始以所计算的传输调度沿着所选择的路径向接收端传输TSN流。
本方案所提方法的创新在于对上面步骤(3)中CNC路径计算和流量调度提出了融合路径选择与门控调度算法,该方法的执行分为四个阶段,依次分别是网络建模阶段、备选路径 阶段、路径选择阶段和门控设计阶段。所提方法概述如下:
在网络建模阶段,CNC根据网络拓扑和TSN流量特征进行网络建模,是后续阶段能够正常执行的基础;
在备选路径阶段,当CNC从CUC收到连接要求时,CNC首先根据网络拓扑使用融合路径选择与门控调度算法选出K条备选路径,再根据路径关键度选出m(m<K)条链路,将这m条路径作为路径选择阶段的输入;
在路径选择阶段,CNC根据TSN流量中TT流量和非TT流的特点,基于链路传输的代价和信息素更新的方式,计算出每一对发送端到接收端的TT流量的最优传输路径,并将其他的有效路径作为非TT流量的传输路径;
在门控设计阶段,根据路径选择阶段计算出的TT流的最优路径,CNC执行融合路径选择与门控调度算法为TT流量的传输计算出门控调度列表,并将计算结果下发到TSN交换机。
集中式用户配置管理系统包括了CNC和CUC两个部分,CNC中包括了TSN网络拓扑发现模块、TSN网络配置模块、TSN网络更新模块、路径选择模块和门控计算模块,其中路径选择模块和门控计算模块联合执行融合路径选择与门控调度算法,TSN网络拓扑发现模块主要是生成TSN网络的网络拓扑,TSN网络配置模块的主要功能是配置TSN域设备数据,并将配置数据添加到数据库,TSN网络更新模块的主要功能是更新TSN域的所有设备配置数据,并将更新操作在数据库中保存。CNC通过NETCONF/YANG模型来感知TSN网络的资源信息,并下发配置信息。CUC通过用户配置协议将TSN流配置信息下发到TSN网络的终端设备(发送端和接收端)。
CNC利用链路发现协议(LLDP)发现TSN网络拓扑,为了保证网络的实时传输需求,终端设备会将TSN网络连接请求通过用户配置协议将TSN连接请求发送到CUC;CUC将TSN连接请求转换为TSN连接需求,并通过用户网络接口(UNI)转发给集中式网络配置模块(CNC);此时,CNC根据TSN网络拓扑信息和连接需求来计算每一对发送端到接收端TT流的最优传输路径和非TT流的传输路径,为了保证TT流传输过程中因传输链路重叠造成的流量冲突问题,本方案设置了流量传输约束条件,为TT流的最优传输路径配置门控列表,以减小流量冲突的问题,然后CNC利用NETCONF协议将调度表下发到TSN交换机,完成对TSN网络的配置管理。
融合路径选择与门控调度算法的执行步骤如下所示:
第一步:CNC根据链路发现协议(LLDP)来发现TSN网络拓扑,并通过网络建模方法将TSN网络拓扑抽象为网络有向图;
第二步:终端设备通过用户配置协议将备选路径数K、路径关键度η k选出的优选路径数量m、算法最大循环次数N cyc、蚂蚁的最大数量N ant和信息素总量Q、TSN流的周期、大小和时延D i发送给CUC,CUC将该连接请求通过用户网络接口UNI发送到CNC;
第三步:CNC收到路径选择的需求时,为了增加路径选择样本的多样性,首先CNC执行融合路径选择与门控调度算法选出K条最短路径作为备选路径;
第四步:CNC将第三步中选出的K条备选路径利用路径关键度η k来选择出m(m<K)条优选路径;
第五步:CNC将第四步中求出的m条优选路径作为路径选择阶段的输入,根据TSN流量类型的特征,基于链路传输的代价和信息素更新的方式为一对发送端到接收端的TT流找到一条最优的传输路径,并将该条路径存入路径信息表ω中,同时为非TT流找到合适的传输路径;
第六步:CNC遍历是否存在还没有被计算发送端和接收端的路径,如果有则返回执行第三步到第六步,并将计算的每对发送端到接收端TT流的最优路径存到路径信息表ω中,如果没有则执行第七步;
第七步:在门控设计阶段主要为了避免TT流传输过程中因传输链路重叠造成的流量冲突问题,将第六步中计算出的TT流量的最优传输路径信息表ω作为门控设计阶段的输入,设置流量传输约束条件,为每一对终端设备的TT流最优传输路径配置门控列表,保证TT流的可靠性传输;
第八步:CNC将计算的结果封装为XML文件,利用NETCONF协议将基于XML的门控调度表配置到TSN交换机,并将流量传输计算结果通过CUC发送到TSN终端设备。
算法的流程图如图2所示,下面详细说明融合路径选择与门控调度算法。
网络建模阶段:
第一步:CNC根据链路发现协议(LLDP)来发现TSN网络拓扑,并通过网络建模方法将TSN网络拓扑抽象为网络有向图;
TSN路径选择的主要任务是TSN流找到合适的传输路径,并通过门控列表(GCL)来调度TT流的传输。在这个过程中需要考虑的因素有网络拓扑和TSN流量模型。
对于本方案中的术语的定义如表1所示:
表1 TSN网络拓扑参数定义
Figure PCTCN2022091458-appb-000054
Figure PCTCN2022091458-appb-000055
Figure PCTCN2022091458-appb-000056
Figure PCTCN2022091458-appb-000057
TSN的网络拓扑表示为有向图G=(V,E),其中V是节点集合,E是边集合。
Figure PCTCN2022091458-appb-000058
其中S是TSN交换机的集合,H是终端设备的集合。E是一组二元组,代表网络中的链路,使得E≡{(BR i,BR j)|BR i,BR j∈V,BR i≠BR j并且BR i和BR j之间存在联系}。
与每个链路(BR i,BR j)∈E相关联的是由元组(b,ld)表示的测量值列表,其中
Figure PCTCN2022091458-appb-000059
是链路(BR i,BR j)的剩余带宽,单位为Mbps,
Figure PCTCN2022091458-appb-000060
是链路延迟,以微秒为单位,它由
Figure PCTCN2022091458-appb-000061
Figure PCTCN2022091458-appb-000062
组成。需要说明的是
Figure PCTCN2022091458-appb-000063
是有界的。
本方案规定把从发送端开始,按照一定要求传输到接收端的有序数据序列称为流,将所有的TSN流的集合记为F。对于不同类型的流量,其中的主要参数包括了R i、D i、T i和S i。每个TSN流F i表示为四元组F i≡(R i,D i,T i,S i)。
对于发送端ES i和接收端ES′ i,中间经过的BR 1,BR 2,…,BR n共有n个节点的流,可以将其第i对发送端到接收端的路径表示为R i={ES i,BR 1,BR 2,…,BR n,ES′ i},流大小S i的单位是字节,每个帧的最大长度为以太网的MTU(Maximum Transmission Unit)。
第二步:终端设备通过用户配置协议将备选路径数K、路径关键度η k选出的优选路径数量m、算法最大循环次数N cyc、蚂蚁的最大数量N ant、信息素总量Q、TSN流的周期T i、大小S i和时延D i发送给CUC,CUC将该连接请求通过用户网络接口UNI发送到CNC。
备选路径阶段:
第三步:CNC收到路径选择的需求时,为了增加路径选择样本的多样性,首先CNC执行融合路径选择与门控调度算法选出K条最短路径作为备选路径。
本阶段通过融合路径选择与门控调度算法求出一对ES i和ES′ i之间的K条备选路径P k,其原理是采用K最短路径算法(KSP),对于所有ES i,ES′ i∈H,基于它们的最短路径递增排序,要找到这些路径,通过输入网络有向图G、发送端ES i、接收端ES′ i和路径的条数K,输出K条路径集p K,然后将路径集p K作为第四步的输入。
从ES i到ES′ i的路径集合用P sd表示,最短路径问题就是通过有向图G中从ES i到ES′ i的具有最短长度的路径P s,即确定P s∈P sd,使得对于任何其他路径P(P∈P sd,p≠P s)都有c(P s)≤c(P),KSP问题除了要找到最短路径之外,还要确定次短路径、第三短路径,…,直到找到第K短路径为止,用p i表示从ES i到ES′ i的第i短路径,KSP问题是确定路径集合P k={p 1,p 2,…,p K}∈P sd,使得满足以下4个条件:
(1)K条路径是按次序产生的,即对于所有的i(i=1,2,…,K-1),p i是在p i+1之前确定的;
(2)K条路径是按长度由小到大排序的,即对于所有的i(i=1,2,…,K-1),都有c(p i)<c(p i+1);
(3)K条路径是最短的,即对于所有的p∈P sd-P k,都有c(p k)<c(p);
(4)所求得的K条路径都是无环的。
最短路径树T k的构建基于一个重要的概念—偏离路径。假定存在从ES i和ES′ i的两条路径p i=(v 1,v 2,…,v l)和p j=(u 1,u 2,…,u w),如果存在一个整数x满足以下4个条件:
(1)x<l,并且x<w;
(2)v i=u i(0≤i≤x);
(3)v x+1≠u x+1
(4)(u x+1,u x+2,…,u w=t)是从u x+1到t的最短路径。
则称(u x,u x+1)为p j相对于p i的偏离边,u x为p j相对于p i的偏离节点,路径(u x+1,u x+2,…,u w=t)为p j相对于p i的最短偏离路径。如图3所示,p 2相对于p 1的偏离节点为节点1,偏离边为边(1,3),偏离路径为(1,3,4,5)。
备选路径阶段需要输入网络有向图G、发送端ES i、接收端ES′ i和备选路径数K,输出K条备选路径集P k={p 1,p 2,…,p n}∈P sd(1≤n≤K),本方案以链路延时
Figure PCTCN2022091458-appb-000064
的值表示权值,一条路径的权值等于该路径中各个链路的延迟之和。如图4所示,具体执行步骤如下:
3.1:输入网络有向图G、发送端ES i、接收端ES′ i和备选路径数K;
3.2:CNC使用融合路径选择与门控调度算法求出ES i到ES′ i的最短路径,并记为p n(n=1):
p n=ES i→BR a→BR b→…→BR n→ES′ i
3.3:判断n是否小于等于K,且仍有候选路径存在,如果存在执行3.4,否则表示找到K条备选路径;
3.4:在根据第一步求出p n的基础上,根据偏离点和Dijkstra算法求出p n+1,把位于p n上的每个节点(除去接收端ES′ i)分别看作偏离点(设共有x个)。将每个偏离点记为BR i(i=1,2,…,x);
3.5:开始遍历偏离点,从BR i(i=0)开始遍历每一个偏离点,并对每一个偏离点求BR i到接收端ES′ i的最短路径;
3.6:将p n上从起点到BR i的路径加上上一步求得的BR i到接收端ES′ i的最短路径作为求p n+1的候选路径,放到候选路径列表U中;
表2 候选路径列表U和备选路径列表L
备选路径列表L 候选路径列表U
p 1 ES i→BR a→…→BRe→ES′ i
p 2 ES i→BR b→…→BRf→ES′ i
p n ES i→BR c→…→BRg→ES′ i
3.7:偏离点遍历结束;
3.8:判断此时候选路径列表U是否为空;
3.9:如果此时候选路径列表U不为空,则遍历完偏离点后,找出B中权值最小的路径即为所求的p n+1,将该路径从U中移除,并放入备选路径列表L中,在求得p n+1的基础之上,当n+1≤K时,继续执行上述步骤3.3-3.8,否则表示找到K条备选路径;
3.10:若候选路径列表U为空,则表示已经找到了K条备选路径,最后确定路径集合为P k={p 1,p 2,…,p n,p k}∈P sd
第四步:CNC将第三步中选出的K条备选路径利用路径关键度η k来选择出m(m<K)条优选路径;
为了降低算法执行的开销,提高网络的运行效率,本方案采用路径关键度η k对路径表中的各路径进行量化计算,此阶段通过输入K条备选路径集P k,输出m条优选路径集R ant(m<K)。
每个路径p k都有其自己的路径关键度指标Δ(HC,SBW,DL),本部分中“.”表示“的”的意思,路径p k的跳数HC(用p k.HC表示)等于TSN交换机的数目,p k.HC等于除发送端和接收端之外的路径p k的TSN交换机数量,可以定义如下:
p k.HC=len(p k)-2          (1)
路径p k的剩余带宽SBW(用p k.SBW表示)等于其所有组成链路中的剩余带宽的最小值,为 了方便描述,假设p k={ES i,BR 1,BR 2,…,BR n,ES′ i},令S=len(p k)-2,并且可由如下公式表示:
Figure PCTCN2022091458-appb-000065
路径p k的端到端时延DL(用p k.DL表示)等于其链路时延
Figure PCTCN2022091458-appb-000066
之和,可以表示为;
Figure PCTCN2022091458-appb-000067
函数Δ(HC,SBW,DL)用于映射其路径关键度η k(用p k.Δ(HC,SBW,DL)表示)为一个介于0和1之间的值。在所有备选路径集p k中,选择η k最大的路径作为路径选择阶段的输入路径。对于路径p k∈P sd,η k表示如下。
Figure PCTCN2022091458-appb-000068
ω 123=1         (5)
其中
Figure PCTCN2022091458-appb-000069
SBW max表示所有路径p k∈P sd中的最大值SBW;HC min是所有p k∈P sd中最小的HC,DL min是所有p k∈P sd中最小的DL。η k的值越大则表示网络的跳数越小,剩余带宽越大,网络延迟越小。本方案综合考虑路径的跳数和剩余带宽,故中ω 1=0.3,ω 2=0.3,ω 3=0.4。
通过以上方式计算每一条路径p k的路径关键度η k的值,选取前m条η k值最大的路径作为优选路径,并存入路径集R ant中。
路径选择阶段:
第五步:CNC将第四步中求出的m条优选路径作为路径选择阶段的输入,根据TSN流量类型的特征,基于链路传输的代价和信息素更新的方式为一对发送端到接收端的TT流找到一条最优的传输路径,并将该条路径存入路径信息表ω中,同时为非TT流找到合适的传输路径。
融合路径选择与门控调度算法中执行路径选择功能的原理是基于蚁群算法,蚁群算法(ACO)是一种群智能算法,它是由一群无智能或有轻微智能的个体(Agent)通过相互协作而表现出智能行为,从而为求解复杂问题提供了一个新的可能性。
蚂蚁a从节点BR i到BR j的传输概率为:
Figure PCTCN2022091458-appb-000070
Figure PCTCN2022091458-appb-000071
allow a表示从节点到下一个节点的集合,α越大,信息素的指导作用越强,β越大,蚂蚁决策受路径距离信息影响越大,贪婪当前效果。δ和ε为权重因子且δ+ε=1,0<δ<1,0<ε<1,δ和ε的取值根据当前流量类型而定,当数据流为TT流时,由于TT流对延迟较为敏感,此时δ的取值要小于ε,本方案取δ=0.3,ε=0.7。
当蚂蚁在完成第一次寻找路径之后,需要对信息素进行更新,其更新的公式如下:
Figure PCTCN2022091458-appb-000072
Δτ(BR i,BR j)取值根据TSN数据流类型而确定,其表达式如下所示。
Figure PCTCN2022091458-appb-000073
Figure PCTCN2022091458-appb-000074
的表达式如下所示。
Figure PCTCN2022091458-appb-000075
由Δτ(BR i,BR j)的表达式可知,本方案根据TT流对端到端的延迟和非TT流对带宽的不同要求,对信息素进行分类更新,从而保证TT流和非TT流的最优路径都满足各自的需求,提升网络的利用率。
假设融合路径选择与门控调度算法在完成执行后得到路径集合为R={path 1,path 2,path 3,…,path N},当数据流为TT流时,选取R中延迟最小的路径为TT流传输的最优路径,当数据流为非TT流时,需要从R中选择N 非TT流条剩余带宽较大的路径进行传输,其中1<N 非TT流<N。
本方案根据TSN流量的类型,分别设置融合路径选择与门控调度算法中链路传输的代价和信息素更新的方式,通过计算TT流和非TT流的最佳传输路径,降低丢包率,提高网络的利用率。本算法通过输入R ant、R b、N cyc、N ant和Q,输出R TT流和非TT流f i′的传输路径。具体的算法流程如下所示。
5.1:先设置初始化参数,从预选路径阶段输出融合路径选择和门控调度算法的初始路径表R ant。设置蚂蚁禁忌表R b,算法最大循环次数N cyc,蚂蚁的最大数量N ant,信息素总量Q;
5.2:判断流量类型,根据TT流和非TT流分别设置影响因子α和β,链路权重因子δ和ε,信息素挥发系数ρ,信息素增量Δτ。蚂蚁置于发送端,并将发送端存入R b中;
5.3:循环次数n cyc=n cyc+1;
5.4:蚂蚁数量n ant=n ant+1;
5.5:蚂蚁根据公式(6)选择下一个节点;
5.6:判断是否到达接收端,如果未达到接收端,返回5.5;如果到达接收端,将该蚂蚁的走过的路径存入路径表R te中,将该路径节点存入禁忌表R b中,避免与下一只蚂蚁的路径产生交叉,然后跳转至5.7。
5.7:判断n ant是否等于N ant,如果n ant<N ant,则返回5.4,如果n ant=N ant,从R te中选取传输代价TV最小的路径作为该次循环得出的最优路径存入最佳路径集合R中,跳转至5.8;
5.8:判断n cyc是否等于N cyc,如果n cyc≠N cyc,返回5.3,根据公式(8)更新链路中的信息素,清空R b;如果n cyc=N cyc,则跳转到5.9;
5.9:根据TSN流量类型,分别输出TT流和非TT流的最优传输路径。如果是TT流f i,从路径集合R中选择延迟最小的路径作为最优路径,可表示为
Figure PCTCN2022091458-appb-000076
如果是非TT流f i′,从路径集合R中选择N 非TT流条路径作为输出最优解,并且进行转发。
第六步:CNC遍历是否存在还没有被计算的发送端和接收端,如果有则执行第三步到第六步,并将计算的每对发送端到接收端TT流的最优路径存到路径信息表ω中,如果没有则执行第七步。
由于本方案适用的网络拓扑是一个多输入多输出的网络,故对于每一对TSN发送端和接收端都存在路径选择的问题,因后续的门控调度机制只针对TT流,因此求出的最优路径也是TT流的最优路径。需要重复第三步到第六步,计算出每对接收端和发送端的最优路径都存储在路径信息表ω中。门控设计阶段会将路径信息表ω中每一条TT流的最优路径计算门控列表并调度。表3说明了路径选择阶段的计算结果。
表3 每对终端设备路径计算信息表ω
Figure PCTCN2022091458-appb-000077
门控设计阶段:
IEEE 802.1Qbv标准指定了一种称为时间感知整形器(TAS)的选通机制,该机制基于静态生成的称为门控控制列表的周期性计划,启用和禁用要连接到关联的出口端口的队列的选择。网络中的通信是通过定期将数据流从发送端发送到接收端来实现的。本阶段网络中数据流的通信路径已经在路径选择阶段中算出。TT流量是最具有延迟敏感的TSN流类别,本方案为了保证TT流传输过程中因传输链路重叠造成的流量冲突问题,通过截止时间、流量传输约束条件和TT流的最优传输路径信息表ω,为TT流构建周期性的GCL调度,以减小流量冲突的带来的延迟影响。
链路(BR i,BR j)∈E表示一个通信的方向,因此一对链路(BR i,BR j),(BR j,BR i)∈E表示节点BR i和节点BR j之间的全双工物理链路,从调度的角度上来看,这两条链路是两种不同的资源。假设网络中的所有设备都是时间同步的。最坏情况下的同步误差,即网络中任意两个设备的本地时钟之间的最大差异,表示为synPre。TSN流可以有不同的周期,将所有数据流的周期记作T,T={f 0.T 0,f 1.T 1,…,f n.T n},其中f n.T n表示流f n的周期T n,而门控列表的循环周期GC等于所有流的最小公倍数(lcm),可以表示为:
GC=lcm(T)          (11)
因此,TSN的TT流在一个门控列表的周期内可能出现多次。
本方案采取的融合路径选择与门控调度算法来搜索计算流的传输路径,该算法可以计算出TT流需要传输的最优路径,本节的重点是在TT流量的调度。如果流f i通过链路(BR i,BR j),则
Figure PCTCN2022091458-appb-000078
其中
Figure PCTCN2022091458-appb-000079
可以是空集。
假设需要零抖动(严格周期性调度),即流f i∈F的每个后续帧出现的时间距离始终等于T i。整型变量
Figure PCTCN2022091458-appb-000080
的计算方法可以表示为:
Figure PCTCN2022091458-appb-000081
第七步:在门控设计阶段主要为了避免TT流传输过程中因传输链路重叠造成的流量冲突问题,将第六步中计算出的TT流量的最优传输路径信息表ω作为门控设计阶段的输入,设置流量传输约束条件,为每一对终端设备的TT流最优传输路径配置门控列表,保证TT流的可靠性传输。
在时间敏感网络中,为了保证时间敏感的流量能够及时准确的传输,需要对网络的传输行为添加一定的约束,本方案的研究目标是求满足这些约束条件的可行解,并通过相应的算法工具求出最优解。
根据GCL的传输规则,对时隙长度LOS进行分析。由于以时隙为基本单位对数据进行偏移,假定LOS能够整除所有数据流的周期。那么最大的时隙长度为:
MAX(LOS)=GCD(T)          (13)
即LOS的最大值为数据流周期的最大公约数。TSN标准规定,报文在相邻两个节点的发送时隙与接收时隙相同。那么,时隙的长度至少需要保证队列中的最后一个报文在相邻节点之间的发送时隙与接收时隙相同,因此最小的时隙长度为:
Figure PCTCN2022091458-appb-000082
其中,LOG是传输队列的长度,
Figure PCTCN2022091458-appb-000083
即队列资源全部被占用时,最后1byte发送到链路上所花费的时间;本方案取synPre=0.1μs。
假设传输路径为{ES i,BR 1,BR 2,…,BR n,ES′ i},
Figure PCTCN2022091458-appb-000084
的计算公式如下所示。
Figure PCTCN2022091458-appb-000085
TT流从发送端到接收端沿着计算出的最优路径进行传输,其端到端的时延包括了传播时延、处理时延和传输时延,时延分析如图5所示,其中,t n表示交换机BR i的发送时间,D的表达式如下所示。
Figure PCTCN2022091458-appb-000086
端到端的时延约束:流f i从发送端经过j个交换机到达接收端的端到端时延应小于等于D i,设(ES i,BR j)和(BR t,ES′ i)分别表示流f i的传输路径的第一个和最后一个链路,可表示为:
Figure PCTCN2022091458-appb-000087
帧约束:此约束确保在传输路径的第一条链路上传输帧不能在发送时间之前开始,必须在截止日期之前的最后一条链路上完成。
Figure PCTCN2022091458-appb-000088
Figure PCTCN2022091458-appb-000089
链接约束:链路上两个不同帧的传输不能在时间上重叠。也就是说,对于同一链路上的每对不同帧,一帧的传输必须在另一帧的传输开始之前完成,反之亦然。该约束必须考虑循环周期GC中的所有帧出现。由于零抖动要求,链路(BR i,BR j)∈E上的流f i的第k个帧出现的传输偏移等于
Figure PCTCN2022091458-appb-000090
因此,链接约束可以写成如下形式:
Figure PCTCN2022091458-appb-000091
Figure PCTCN2022091458-appb-000092
帧隔离约束:为了避免流传输过程中帧丢失的问题,本方案提出的解决方法是增加帧的隔离约束。
a)无传输持续时间的帧隔离约束
首先对图6所示的符号进行解释说明。
Figure PCTCN2022091458-appb-000093
Figure PCTCN2022091458-appb-000094
Figure PCTCN2022091458-appb-000095
Figure PCTCN2022091458-appb-000096
无传输持续时间的帧隔离约束的目的是隔离两个不相同的帧,使得一个帧只能在另一个帧从队列中被调度之后才发送到共享队列中。有三种方法可以满足这个约束,第一种方法是交换机BR a中的f j到达之前在链路(BR a,BR b)上调度f i,第二种方法是在交换机BR a中的f i到达之前在链路(BR a,BR b)上调度f j,第三种方法就是帧存储在不同的队列中,需要注意的是,必须考虑调度周期中所有出现的帧,由此得出Γ和ξ倍数。
Figure PCTCN2022091458-appb-000097
该约束假设交换机将帧存储在队列中的时刻不取决于帧的传输持续时间。但是如果该帧在被交换机完全接收之后被存储在队列中,则约束(25)可以被如下所述的包括传输持续时间的约束所代替。
b)带有传输持续时间的帧隔离约束:
带有持续时间的帧隔离约束如下所示。
Figure PCTCN2022091458-appb-000098
流量传输约束:该约束对优先关系进行建模,即只有在帧完全传送到交换机并经过处理 后,才能开始从交换机传输帧。
Figure PCTCN2022091458-appb-000099
优化目标:考虑以下目标函数,最小化与流周期相关的最大端到端延迟,也称为响应时间,目的是为了保证流量传输的可靠性,确保帧在传输周期之内传输到接收端。现在引入辅助变量
Figure PCTCN2022091458-appb-000100
该变量表示任何流的端到端延迟与其周期之比,本方案要求该变量的值在[0,1]中即表示调度为最优,表达式如下所示。
Figure PCTCN2022091458-appb-000101
通过将以上的约束条件进行建模并使用求解器求解,根据每一对终端设备中TT流的最优传输路径
Figure PCTCN2022091458-appb-000102
计算出TT流量传输路径上交换机的门控列表,输出结果如表4所示。
表4 计算调度成功结果
Figure PCTCN2022091458-appb-000103
第八步:CNC将计算的结果封装为XML文件,利用NETCONF协议将基于XML的门控调度表配置到TSN交换机,并将流量传输计算结果通过CUC发送到TSN终端设备。
最后说明的是,以上实施例仅用以说明本发明的技术方案而非限制,尽管参照较佳实施例对本发明进行了详细说明,本领域的普通技术人员应当理解,可以对本发明的技术方案进行修改或者等同替换,而不脱离本技术方案的宗旨和范围,其均应涵盖在本发明的权利要求范围当中。

Claims (7)

  1. 一种时间敏感网络中路径选择和门控调度的联合优化方法,其特征在于:包括以下步骤:
    S1:集中式网络配置模块CNC发现TSN网络拓扑,并将TSN网络拓扑抽象为网络有向图;
    S2:终端设备通过用户配置协议向集中式用户配置模块CUC发送TSN连接请求,CUC将所述连接请求通过用户网络接口UNI发送到CNC;
    S3:CNC收到路径选择的需求时,执行融合路径选择与门控调度算法选出K条最短路径作为备选路径;
    S4:CNC将步骤S3中选出的K条备选路径利用路径关键度η k来选择出m条优选路径;
    S5:CNC将步骤S4中选出的m条优选路径作为路径选择阶段的输入,根据TSN流量类型的特征,基于链路传输的代价和信息素更新的方式为一对发送端到接收端的TT流找到一条最优的传输路径,并将该条路径存入路径信息表ω中,同时为非TT流找到合适的传输路径;
    S6:CNC遍历是否存在还没有被计算发送端和接收端的路径,如果有则返回执行步骤S3-S6,并将计算的每对发送端到接收端TT流的最优路径存到路径信息表ω中,如果没有则执行步骤S7;
    S7:将步骤S6中计算出的TT流量的最优传输路径信息表ω作为输入,设置流量传输约束条件,为每一对终端设备的TT流最优传输路径配置门控列表;
    S8:CNC将计算的结果封装为门控调度表,并配置到TSN交换机,再将流量传输计算结果通过CUC发送到TSN终端设备。
  2. 根据权利要求1所述的时间敏感网络中路径选择和门控调度的联合优化方法,其特征在于:步骤S1具体包括:CNC根据链路发现协议LLDP来发现TSN网络拓扑,并通过网络建模方法将TSN网络拓扑抽象为网络有向图;
    TSN的网络拓扑表示为有向图G=(V,E),其中V是TSN网络中的节点集合,
    Figure PCTCN2022091458-appb-100001
    其中S是TSN交换机的集合,H是终端设备的集合;E是边集合,是一组二元组,表示TSN网络中的所有链接,使得
    E≡{(BR i,BR j)|BR i,BR j∈V,BR i≠BR j并且BR i和BR j之间存在联系},其中(BR i,BR j)表示交换机BR i和交换机BR j之间的链路;
    与每个链路(BR i,BR j)∈E相关联的是由元组(b,ld)表示的测量值列表,其中
    Figure PCTCN2022091458-appb-100002
    是链路(BR i,BR j)的剩余带宽;
    Figure PCTCN2022091458-appb-100003
    是链路延迟,由
    Figure PCTCN2022091458-appb-100004
    Figure PCTCN2022091458-appb-100005
    组成;
    Figure PCTCN2022091458-appb-100006
    是有界的;
    把从发送端开始,按照一定要求传输到接收端的有序数据序列称为流,将所有的TSN流的集合记为F;对于不同类型的流,其中的主要参数包括TSN流的传输路径R i、TSN流的端到端时延D i、TSN流的传输周期T i和TSN流的大小S i;每个TSN流F i表示为四元组F i≡(R i,D i,T i,S i);
    对于发送端ES i和接收端ES′ i,中间经过的交换机BR 1,BR 2,…,BR n共有n个节点,将其第i对发送端到接收端的路径表示为R i={ES i,BR 1,BR 2,…,BR n,ES′ i},每个帧的最大长度为以太网的最大传输单元MTU。
  3. 根据权利要求1所述的时间敏感网络中路径选择和门控调度的联合优化方法,其特征在于:步骤S2具体包括:终端设备通过用户配置协议将备选路径数K、路径关键度η k选出的优选路径数量m、算法最大循环次数N cyc、蚂蚁的最大数量N ant、信息素总量Q、TSN流的周期T i、大小S i和时延D i发送给CUC,CUC将连接请求通过用户网络接口UNI发送到CNC。
  4. 根据权利要求1所述的时间敏感网络中路径选择和门控调度的联合优化方法,其特征在于:步骤S3中所述选出K条最短路径作为备选路径,具体为:采用K最短路径算法KSP,对于所有ES i,ES′ i∈H,基于它们的最短路径递增排序,通过输入网络有向图G、发送端ES i、接收端ES′ i和路径的条数K,输出K条路径集p K,然后将路径集p K作为步骤S4的输入;具体执行步骤如下:
    S31:输入网络有向图G、发送端ES i、接收端ES′ i和备选路径数K;
    S32:CNC使用融合路径选择与门控调度算法求出ES i到ES′ i的最短路径,并记为p n(n=1):
    p n=ES i→BR a→BR b→…→BR n→ES′ i
    S33:判断n是否小于等于K,且仍有候选路径存在,如果存在执行S34,否则表示找到K条备选路径;
    S34:在求出p n的基础上,根据偏离点和Dijkstra算法求出p n+1,把位于p n上除去接收端ES′ i的每个节点分别看作偏离点,共有x个;将每个偏离点记为BR i(i=1,2,…,x);
    S35:开始遍历偏离点,从BR i(i=0)开始遍历每一个偏离点,并对每一个偏离点求BR i到接收端ES′ i的最短路径;
    S36:将p n上从起点到BR i的路径加上求得的BR i到接收端ES′ i的最短路径作为求p n+1的候选路径,放到候选路径列表U中;
    S37:偏离点遍历结束;
    S38:判断此时候选路径列表U是否为空;
    S39:如果此时候选路径列表U不为空,则遍历完偏离点后,找出B中权值最小的路径 即为所求的p n+1,将该路径从U中移除,并放入备选路径列表L中,在求得p n+1的基础之上,当n+1≤K时,继续执行步骤S33-S38,否则表示找到K条备选路径;
    S310:若候选路径列表U为空,则表示已经找到了K条备选路径,最后确定路径集合为P k={p 1,p 2,…,p n,p k}∈P sd,P sd表示从ES i到ES′ i的路径集合。
  5. 根据权利要求1所述的时间敏感网络中路径选择和门控调度的联合优化方法,其特征在于:步骤S4具体包括:
    路径p k的跳数HC为除发送端和接收端之外的路径p k的TSN交换机数量(用p k.HC表示),定义如下:
    p k.HC=len(p k)-2       (1)
    路径p k的剩余带宽SBW为其所有组成链路中的剩余带宽的最小值,假设p k={ES i,BR 1,BR 2,…,BR n,ES′ i},令S=len(p k)-2,路径p k的剩余带宽SBW由如下公式表示:
    Figure PCTCN2022091458-appb-100007
    路径p k的端到端时延DL为其链路时延
    Figure PCTCN2022091458-appb-100008
    之和,表示为;
    Figure PCTCN2022091458-appb-100009
    函数Δ(HC,SBW,DL)用于映射其路径关键度η k为一个介于0和1之间的值;在所有备选路径集p k中,选择η k最大的路径作为路径选择阶段的输入路径;对于路径p k∈P sd,η k表示如下:
    Figure PCTCN2022091458-appb-100010
    ω 123=1       (5)
    其中
    Figure PCTCN2022091458-appb-100011
    SBW max表示所有路径p k∈P sd中的最大值SBW;HC min是所有p k∈P sd中最小的HC,DL min是所有p k∈P sd中最小的DL;η k的值越大则表示网络的跳数越小,剩余带宽越大,网络延迟越小;
    通过以上方式计算每一条路径p k的路径关键度η k的值,选取前m条η k值最大的路径作为优选路径,并存入优选路径集R ant中。
  6. 根据权利要求1所述的时间敏感网络中路径选择和门控调度的联合优化方法,其特征在于:所述步骤S5中,基于蚁群算法找到最优传输路径,具体包括以下步骤:
    S51:先设置初始化参数,从预选路径阶段输出融合路径选择和门控调度算法的初始路径 表R ant;设置蚂蚁禁忌表R b,算法最大循环次数N cyc,蚂蚁的最大数量N ant,信息素总量Q;
    S52:判断流量类型,根据TT流和非TT流分别设置影响因子α和β,链路权重因子δ和ε,信息素挥发系数ρ,信息素增量Δτ。蚂蚁置于发送端,并将发送端存入R b中;
    S53:循环次数n cyc=n cyc+1;
    S54:蚂蚁数量n ant=n ant+1;
    S55:蚂蚁根据公式(6)选择下一个节点:
    Figure PCTCN2022091458-appb-100012
    Figure PCTCN2022091458-appb-100013
    其中,P(BR i,BR j)表示蚂蚁a从节点BR i到BR j的传输概率,allow a表示从节点到下一个节点的集合,α越大,信息素的指导作用越强,β越大,蚂蚁决策受路径距离信息影响越大,贪婪当前效果;δ和ε为权重因子且δ+ε=1,0<δ<1,0<ε<1,δ和ε的取值根据当前流量类型而定,当数据流为TT流时,δ的取值小于ε;τ(BR i,BR j)表示链路(BR i,BR j)的信息素数量;μ(BR i,BR j)表示节点选择的启发因子;
    Figure PCTCN2022091458-appb-100014
    表示链路(BR i,BR j)的剩余带宽;
    Figure PCTCN2022091458-appb-100015
    表示链路(BR i,BR j)的延迟;
    S56:判断是否到达接收端,如果未达到接收端,返回S55;如果到达接收端,将该蚂蚁的走过的路径存入路径表R te中,将该路径节点存入禁忌表R b中,避免与下一只蚂蚁的路径产生交叉,然后跳转至S57;
    S57:判断n ant是否等于N ant,如果n ant<N ant,则返回S54,如果n ant=N ant,从R te中选取传输代价TV最小的路径作为该次循环得出的最优路径存入最佳路径集合R中,跳转至S58;
    S58:判断n cyc是否等于N cyc,如果n cyc≠N cyc,返回S53,根据公式(8)更新链路中的信息素,清空R b;如果n cyc=N cyc,则跳转到S59;
    Figure PCTCN2022091458-appb-100016
    ρ表示信息素的挥发系数,0<ρ<1,τ(BR i,BR j)表示链路(BR i,BR j)的信息素数量,Δτ(BR i,BR j)表示链路(BR i,BR j)的信息素增量;
    Δτ(BR i,BR j)取值根据TSN数据流类型而确定,其表达式如下所示:
    Figure PCTCN2022091458-appb-100017
    Q表示信息素的总数量,
    Figure PCTCN2022091458-appb-100018
    表示链路(BR i,BR j)的传输延迟,路径选择系数
    Figure PCTCN2022091458-appb-100019
    的表达式如下所示:
    Figure PCTCN2022091458-appb-100020
    S59:根据TSN流量类型,分别输出TT流和非TT流的最优传输路径;如果是TT流f i,从路径集合R中选择延迟最小的路径作为最优路径,可表示为
    Figure PCTCN2022091458-appb-100021
    如果是非TT流f i′,从路径集合R中选择N 非TT流条路径作为输出最优解,并且进行转发。
  7. 根据权利要求1所述的时间敏感网络中路径选择和门控调度的联合优化方法,其特征在于:步骤S7中,门控列表的循环周期GC表示为:
    GC=lcm(T)       (11)
    其中,lcm表示最小公倍数,T为所有数据流的周期,
    T={f 0.T 0,f 1.T 1,…,f n.T n}
    其中f n.T n表示流f n的周期T n
    假设流f i∈F的每个后续帧出现的时间距离始终等于T i,TT流f i在(BR i,BR j)上传输偏移量为
    Figure PCTCN2022091458-appb-100022
    存储的队列ID为
    Figure PCTCN2022091458-appb-100023
    (BR i,BR j)∈E,,则流f i在链路(BR i,BR j)上的传输任意数据帧所用的时间
    Figure PCTCN2022091458-appb-100024
    表示为:
    Figure PCTCN2022091458-appb-100025
    根据GCL的传输规则,对时隙长度LOS进行分析,最大的时隙长度为:
    MAX(LOS)=GCD(T)        (13)
    即LOS的最大值为数据流周期的最大公约数;
    最小的时隙长度为:
    Figure PCTCN2022091458-appb-100026
    其中,LOG是传输队列的长度,
    Figure PCTCN2022091458-appb-100027
    即队列资源全部被占用时,最后1byte发送到链路上所花费的时间;
    传输路径为{ES i,BR 1,BR 2,…,BR n,ES′ i},交换机BR n的GCL循环开始时间
    Figure PCTCN2022091458-appb-100028
    的计算公式如下:
    Figure PCTCN2022091458-appb-100029
    Figure PCTCN2022091458-appb-100030
    表示发送端发送第一个数据帧的发送时间,
    Figure PCTCN2022091458-appb-100031
    表示流f i在链路(ES i,BR 1)上的传输任意数据帧所用的时间,
    Figure PCTCN2022091458-appb-100032
    表示BR i中的处理延迟,
    Figure PCTCN2022091458-appb-100033
    表示BR i中的传输延迟,
    Figure PCTCN2022091458-appb-100034
    表示流f i在链路(BR j-1,BR j)上的传输任意数据帧所用的时间,synPre表示两个节点设备的时间同步最大差值;
    一个交换机到下一个交换机之间的时延D的表达式如下:
    Figure PCTCN2022091458-appb-100035
    Figure PCTCN2022091458-appb-100036
    表示BR n中的处理延迟,
    Figure PCTCN2022091458-appb-100037
    表示BR n-1中的传输延迟,
    Figure PCTCN2022091458-appb-100038
    表示(BR n-1,BR n)中的传播延迟;
    端到端的时延约束表示为:
    Figure PCTCN2022091458-appb-100039
    帧约束表示为:
    Figure PCTCN2022091458-appb-100040
    Figure PCTCN2022091458-appb-100041
    链接约束表示为:
    Figure PCTCN2022091458-appb-100042
    Figure PCTCN2022091458-appb-100043
    Figure PCTCN2022091458-appb-100044
    无传输持续时间的帧隔离约束表示为:
    Figure PCTCN2022091458-appb-100045
    其中,
    Figure PCTCN2022091458-appb-100046
    Figure PCTCN2022091458-appb-100047
    Figure PCTCN2022091458-appb-100048
    Figure PCTCN2022091458-appb-100049
    带有传输持续时间的帧隔离约束表示为:
    Figure PCTCN2022091458-appb-100050
    Figure PCTCN2022091458-appb-100051
    Figure PCTCN2022091458-appb-100052
    流量传输约束表示为:
    Figure PCTCN2022091458-appb-100053
    Figure PCTCN2022091458-appb-100054
    优化目标为:
    Figure PCTCN2022091458-appb-100055
    Figure PCTCN2022091458-appb-100056
    引入辅助变量
    Figure PCTCN2022091458-appb-100057
    表示任何流的端到端延迟与其周期之比,要求该变量的值在[0,1]中即表示调度为最优;
    通过将以上的约束条件进行建模并使用求解器求解,根据每一对终端设备中TT流的最优传输路径
    Figure PCTCN2022091458-appb-100058
    计算出TT流量传输路径上交换机的门控列表。
PCT/CN2022/091458 2022-01-17 2022-05-07 一种时间敏感网络中路径选择和门控调度的联合优化方法 WO2023134091A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202210058869.6 2022-01-17
CN202210058869.6A CN114172843B (zh) 2022-01-17 2022-01-17 一种时间敏感网络中路径选择和门控调度的联合优化方法

Publications (1)

Publication Number Publication Date
WO2023134091A1 true WO2023134091A1 (zh) 2023-07-20

Family

ID=80489403

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2022/091458 WO2023134091A1 (zh) 2022-01-17 2022-05-07 一种时间敏感网络中路径选择和门控调度的联合优化方法

Country Status (2)

Country Link
CN (1) CN114172843B (zh)
WO (1) WO2023134091A1 (zh)

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117196130A (zh) * 2023-09-20 2023-12-08 中国人民解放军国防科技大学 一种基于复杂网络的杀伤链路径快速查找方法
CN117896315A (zh) * 2024-03-12 2024-04-16 南京邮电大学 一种基于时间敏感网络的智能变电站混合流调度方法及系统

Families Citing this family (15)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114172843B (zh) * 2022-01-17 2023-05-16 重庆邮电大学 一种时间敏感网络中路径选择和门控调度的联合优化方法
CN114615205B (zh) * 2022-03-23 2024-03-08 南京航空航天大学 基于时间效益函数的时间敏感网络下的混合流量调度方法
CN115333998B (zh) * 2022-05-05 2023-09-08 国网宁夏电力有限公司信息通信公司 一种适用于电力通信网络的时间敏感网络流量调度方法
CN115022182B (zh) * 2022-05-31 2023-07-28 广西大学 一种基于qsilp算法的列车通信网络实时流调度优化方法
CN114785738B (zh) * 2022-06-16 2022-09-20 北京邮电大学 时间敏感流的调度方法、装置和电子设备
CN115277306B (zh) * 2022-06-27 2023-05-26 重庆邮电大学 基于逐步逼近的面向时间敏感网络与工业无线网络的联合调度方法
CN115378865A (zh) * 2022-08-12 2022-11-22 北京智芯微电子科技有限公司 Avb流时延的路由调度方法和系统、存储介质、终端设备
CN115150274B (zh) * 2022-09-06 2022-11-25 国网湖北省电力有限公司电力科学研究院 时间敏感网络设备的统一配置方法、系统及存储介质
CN115460130B (zh) * 2022-09-20 2023-10-13 重庆邮电大学 一种时间敏感网络中的多路径联合调度方法
CN115883438A (zh) * 2022-11-16 2023-03-31 重庆邮电大学 时间敏感网络中时间触发流量的路由与调度方法、装置及可读存储介质
CN116032744A (zh) * 2023-01-03 2023-04-28 重庆邮电大学 一种工业无线网络融合时间敏感网络的统一配置方法
CN116132354B (zh) * 2023-02-23 2024-03-22 中国科学院软件研究所 一种无人机集群网络化传输路径优化方法及系统
CN116471177A (zh) * 2023-06-14 2023-07-21 国网湖北省电力有限公司电力科学研究院 时间敏感网络动态配置管理系统、方法及介质
CN117768372B (zh) * 2023-12-08 2024-06-25 北京科技大学 一种时间敏感网络中路由与调度联合优化方法及装置
CN117596200B (zh) * 2024-01-15 2024-05-07 之江实验室 一种时间敏感网络路由调度方法、电子设备、介质

Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111740924A (zh) * 2020-07-29 2020-10-02 上海交通大学 一种时间敏感网络门控机制流量整形与路由规划调度方法
CN113068263A (zh) * 2021-03-26 2021-07-02 鹏城实验室 一种时间敏感网络时隙调度方法、终端及存储介质
CN113300956A (zh) * 2020-02-24 2021-08-24 四零四科技股份有限公司 处理用于时间敏感网络中的数据流的路由路径的装置及方法
WO2021192603A1 (en) * 2020-03-27 2021-09-30 Mitsubishi Electric Corporation Scheduling data traffic in wireless time sensitive networks
WO2021227245A1 (zh) * 2020-05-11 2021-11-18 重庆邮电大学 一种面向tsn和非tsn互联的工业异构网络调度方法
CN114172843A (zh) * 2022-01-17 2022-03-11 重庆邮电大学 一种时间敏感网络中路径选择和门控调度的联合优化方法

Family Cites Families (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112953830B (zh) * 2021-01-28 2022-12-23 北京邮电大学 时间敏感网络中流量帧的路由规划调度方法及装置
CN113032938B (zh) * 2021-03-26 2022-07-08 北京邮电大学 时间敏感流的路由调度方法、装置、电子设备及介质
CN113300960B (zh) * 2021-07-27 2021-11-23 南京中网卫星通信股份有限公司 一种基于路由调度与联合优化的时延确定性传输方法
CN113645149B (zh) * 2021-10-15 2022-02-08 鹏城实验室 时间敏感网络门控列表实现方法

Patent Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113300956A (zh) * 2020-02-24 2021-08-24 四零四科技股份有限公司 处理用于时间敏感网络中的数据流的路由路径的装置及方法
WO2021192603A1 (en) * 2020-03-27 2021-09-30 Mitsubishi Electric Corporation Scheduling data traffic in wireless time sensitive networks
WO2021227245A1 (zh) * 2020-05-11 2021-11-18 重庆邮电大学 一种面向tsn和非tsn互联的工业异构网络调度方法
CN111740924A (zh) * 2020-07-29 2020-10-02 上海交通大学 一种时间敏感网络门控机制流量整形与路由规划调度方法
CN113068263A (zh) * 2021-03-26 2021-07-02 鹏城实验室 一种时间敏感网络时隙调度方法、终端及存储介质
CN114172843A (zh) * 2022-01-17 2022-03-11 重庆邮电大学 一种时间敏感网络中路径选择和门控调度的联合优化方法

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117196130A (zh) * 2023-09-20 2023-12-08 中国人民解放军国防科技大学 一种基于复杂网络的杀伤链路径快速查找方法
CN117196130B (zh) * 2023-09-20 2024-05-24 中国人民解放军国防科技大学 一种基于复杂网络的杀伤链路径快速查找方法
CN117896315A (zh) * 2024-03-12 2024-04-16 南京邮电大学 一种基于时间敏感网络的智能变电站混合流调度方法及系统
CN117896315B (zh) * 2024-03-12 2024-05-24 南京邮电大学 一种基于时间敏感网络的智能变电站混合流调度方法及系统

Also Published As

Publication number Publication date
CN114172843A (zh) 2022-03-11
CN114172843B (zh) 2023-05-16

Similar Documents

Publication Publication Date Title
WO2023134091A1 (zh) 一种时间敏感网络中路径选择和门控调度的联合优化方法
Yu et al. Traffic statistics and performance evaluation in optical burst switched networks
CN115460130B (zh) 一种时间敏感网络中的多路径联合调度方法
CN114553697B (zh) 一种面向工业无线与tsn融合的网络调度方法
CN108833279A (zh) 软件定义网络中基于业务分类的多约束QoS路由的方法
US20220150159A1 (en) Control device, switch device and methods
Huang et al. TACQ: Enabling zero-jitter for cyclic-queuing and forwarding in time-sensitive networks
CN115883438A (zh) 时间敏感网络中时间触发流量的路由与调度方法、装置及可读存储介质
Huang et al. Towards network-wide scheduling for cyclic traffic in ip-based deterministic networks
Yu et al. Deep reinforcement learning-based deterministic routing and scheduling for mixed-criticality flows
CN114884557B (zh) 一种基于网络演算的卫星时间敏感网络路径选择方法
Peng et al. Traffic shaping at the edge: Enabling bounded latency for large-scale deterministic networks
Zhao et al. DRL-M4MR: An intelligent multicast routing approach based on DQN deep reinforcement learning in SDN
Li et al. Reliable routing and scheduling in time-sensitive networks
Shahri et al. Enhancing mqtt with real-time and reliable communication services
Singh Routing algorithms for time sensitive networks
Zheng et al. Mix-flow scheduling for concurrent multipath transmission in time-sensitive networking
Li et al. Joint routing and scheduling for deterministic networking: A segment routing approach
Angilella et al. High capacity and resilient large-scale deterministic ip networks
Liu et al. PPO-based reliable concurrent transmission control for telemedicine real-time services
Cappanera et al. Efficient link scheduling for online admission control of real-time traffic in wireless mesh networks
CN115333998B (zh) 一种适用于电力通信网络的时间敏感网络流量调度方法
Zang et al. Intra-domain heuristic traffic scheduling algorithm for time-sensitive networks
Li et al. DRNet: QoS-aware Routing for SDN using Deep Reinforcement Learning
Sedaghat et al. FRT-SDN: an effective firm real time routing for SDN by early removal of late packets

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 22919731

Country of ref document: EP

Kind code of ref document: A1