An adaptive TSN offline fault-tolerant scheduling method based on resource reservation

Abstract

This invention relates to an adaptive TSN offline fault-tolerant scheduling method based on resource reservation, comprising the following steps: 1) Inputting initial conditions to initialize the network; 2) Solving the offset for all real-time flows, and performing a process of allocating feasible network resources to flows by inputting non-real-time flow attributes; 3) Determining whether there is a feasible solution for the real-time flow offset in step 2). If there is a feasible solution, then allocating feasible network resources to all real-time flows, and performing a process of solving the offset for flows by inputting non-real-time flow attributes; 4) Determining whether there is a feasible solution for solving the offset for flows in step 3). If there is a feasible solution, then outputting the gating list content based on the flow offset. This invention improves the reliability of real-time flow transmission and the network transmission service of non-real-time flows while saving network resources. It uses Network Retransmission Request (ARQ) to trigger retransmission flows and ensures that they meet the TSN deterministic transmission constraints.

Description

Technical Field

[0001] The present invention relates to a fault-tolerant scheduling technology based on a real-time network, and in particular to an adaptive TSN offline fault-tolerant scheduling method based on resource reservation. Background Art

[0002] As an important support for the development of intelligent manufacturing in my country, the Industrial Internet has been highly recognized and valued by the country. The current industrial Internet network adopts a layered heterogeneous network system. Multiple services require hybrid cross-network transmission and execution. The transmission and scheduling of industrial heterogeneous dynamic hybrid services is a bottleneck restricting the application of the Industrial Internet. Time-sensitive networking (TSN) is widely regarded as an effective way to solve the real-time scheduling problem of industrial heterogeneous networks.

[0003] Based on the time sensitivity of different tasks, TSN provides universal standards for managing the time nodes and process sequence of information interaction. TSN allows real-time and non-real-time data streams to be transmitted in the same standard network, making standard Ethernet have the characteristics of deterministic transmission while also having the advantages of low latency and high bandwidth. TSN has now become a key technology that has attracted widespread attention.

[0004] During the TSN message transmission process, electromagnetic interference and other conditions may cause temporary errors in the data transmission process. Data redundancy (CRC) checking can quickly detect such errors and notify the data sender to retransmit the erroneous data through automatic repeat request (ARQ). However, data retransmission increases the end-to-end delay of the data stream, causing the critical real-time stream to fail to meet the time constraint, thus missing the deadline and causing catastrophic consequences for the system.

[0005] Although there are already technologies related to TSN fault tolerance, including TSN protocols, most of them focus on routing fault tolerance and lack fault tolerance technology for retransmission missed deadlines.

[0006] The current technology basically directly allocates exclusive network redundant resources for real-time streams. Since not every transmission can be correctly transmitted to the receiving end, this method causes a huge waste of network resources. At present, the research on TSN has not yet considered ARQ-related streams as deterministic transmission services. Summary of the invention

[0007] In view of the shortcomings of the prior art in the TSN message transmission process, such as the lack of fault-tolerant technology for retransmission missed deadlines, low fault-tolerant capability, and waste of network resources, the problem to be solved by the present invention is to provide an adaptive TSN offline fault-tolerant scheduling method based on resource reservation that can improve the time fault-tolerant capability of a time-sensitive network scheduling algorithm.

[0008] In order to solve the above technical problems, the technical solution adopted by the present invention is:

[0009] An adaptive TSN offline fault-tolerant scheduling method based on resource reservation includes the following steps:

[0010] 1) Input initial conditions and initialize the network;

[0011] 2) Solve the offsets for all real-time flows, and perform the process of allocating feasible network resources to the flows by inputting the attributes of non-real-time flows;

[0012] 3) Determine whether there is a feasible solution for the real-time stream offset in step 2). If there is a feasible solution, allocate feasible network resources to all real-time streams, and execute the process of solving the stream offset by inputting non-real-time stream attributes;

[0013] 4) Determine whether there is a feasible solution for the offset of the flow in step 3). If there is a feasible solution, output the gating list content according to the offset of the flow.

[0014] In step 2), the offsets of all real-time flows are solved, and the process of allocating feasible network resources to the flows is performed by inputting the attributes of non-real-time flows as follows:

[0015] 201) assigning a route to each flow using the Dijkstra method according to the input flow attributes;

[0016] 202) Based on the assigned routing, flow attributes and network attribute information, an integer linear programming solver is used to find feasible queue assignments and offsets for each flow, with the following specific constraints:

[0017] Define the following notation: Each stream Has a unique category, i.e. belongs to the live stream collection Or belongs to the non-real-time stream collection therefore f i The period is T i , the deadline is d i , in the physical link [v a , v b The execution transmission delay on ] is The propagation delay is Through the physical link [v a , v b The set of flows of ] is Also divided into real-time stream collections and non-real-time stream collections f i Relative to the super cycle In the path [v a , v b], that is, the offset on Time, turn on f i The queue Q i ;κ is the number of queues allocated to TT flows;

[0018] Determine the optimization goal: Minimize κ so that 7-Q i ≤κ;

[0019] The constraints for real-time streaming are as follows:

[0020] Schedulability Constraints: Flow f i From the starting point v s Any time the last bit in a packet is released, it reaches the destination v e The time must be less than or equal to f i Deadline d i ; If this constraint is satisfied, the flow schedulability and end-to-end delay constraints are guaranteed; this constraint contains f i The transmission time, the propagation delay on the link, and the waiting time in the queue are constrained by the following formula:

[0021]

[0022] Link constraint: Two packets passing through the same physical link cannot overlap in the time domain, that is, any flow f j Start by link [v a , v b ], recorded as No earlier than any flow f i In the link [v a , v b ]The time when the transmission is completed is obtained as follows:

[0023]

[0024] Where α = 0, 1, ..., H / T i -1, β = 0, i, ..., H / T j -1.

[0025] Streaming constraints: Stream f i The propagation of packets follows the order of the flow routing path, that is, in flow f i The route exists, f i In [v x , v b ] starts transmission no earlier than the time from [v a , v x ]arrival x , v b ] time, the constraint formula is as follows:

[0026]

[0027] Wherein, network accuracy δ is expressed as the maximum difference between the local clocks of any two devices synchronized by the IEEE 802.1AS time synchronization standard;

[0028] Flow isolation constraint: Link [v a , v x ] and [v x , v b ] is the flow i Two consecutive links in a route, [v x , v b ] is also a flow j The links in the route, and f i and f j In the network node v x are assigned to the same egress queue, f i Reaching node v x The time should not be earlier than f j In [v x , v b ]The time when the transmission is completed, the constraint formula is as follows:

[0029]

[0030] In step 4), the gated list content is output according to the offset of the stream. The steps are as follows:

[0031] 401) Select any path from all paths [v a , v b ]∈∏;

[0032] 402) Initialized at path [v a , v b ] a GCL Content

[0033] 403) for the path [v a , v b ] on any non-retransmission flow f i ,Will and queue switch status, added to In which, the queue switch state Q i Set to 1, and the rest of the queue states are set to 0;

[0034] 404) for the path [v a , v b ] on any retransmission flow f RE，1 ,Will and queue switch status, added to In which, the queue switch state Q i Set to 1, all queues containing non-real-time streams are also set to 1, and the status of other queues is set to 0;

[0035] 405) Sort by time in descending order;

[0036] 406) Change the path [v a , v b ] is removed from all paths;

[0037] 407) Determine whether UI is empty, if it is empty, then end; otherwise, execute step 401).

[0038] If there is no feasible solution for the real-time flow offset in step 2), no feasible scheduling solution can be provided; or, if there is no feasible solution for the flow solution offset in step 4), no feasible scheduling solution can be provided.

[0039] The present invention has the following beneficial effects and advantages:

[0040] 1. The present invention realizes error-tolerant retransmission of data caused by instantaneous errors at the time redundancy level, reserves network and time resources for retransmission streams, and when data failure occurs, non-real-time streams are transmitted using the reserved resources, thereby saving network resources and improving the reliability of real-time stream transmission and the network transmission service of non-real-time streams.

[0041] 2. The present invention uses a network retransmission request (ARQ) to trigger the retransmission flow and make it meet the TSN deterministic transmission constraints. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] Figure 1 This is a flow chart of the adaptive TSN offline fault-tolerant scheduling method based on resource reservation of the present invention;

[0043] Figure 2 A flow chart for solving offset for convection according to the present invention;

[0044] Figure 3 The present invention is a flow chart of outputting the contents of a gating list according to the offset of a stream. DETAILED DESCRIPTION

[0045] The present invention will be further described below in conjunction with the accompanying drawings.

[0046] like Figure 1 As shown, the present invention provides an adaptive TSN offline fault-tolerant scheduling method based on resource reservation, comprising the following steps:

[0047] 1) Input initial conditions and initialize the network;

[0048] 2) Solve the offsets for all real-time flows, and perform the process of allocating feasible network resources to the flows by inputting the attributes of non-real-time flows;

[0049] 3) Determine whether there is a feasible solution for the real-time stream offset in step 2). If there is a feasible solution, allocate feasible network resources to all real-time streams, and execute the process of solving the stream offset by inputting non-real-time stream attributes;

[0050] 4) Determine whether there is a feasible solution for the offset of the flow in step 3). If there is a feasible solution, output the gating list content according to the offset of the flow.

[0051] When an instantaneous transmission failure occurs during the flow transmission process, the method of the present invention ensures that the real-time flow is still transmitted before the flow deadline under the premise of ensuring one retransmission, thereby improving the time fault tolerance capability of the time-sensitive network scheduling algorithm.

[0052] In step 1), input the initial conditions: initialize the flow attributes (including flow attributes, cycle, deadline, total length of all data frames released each time, start point and end point, etc.), network attributes (such as topology, link bandwidth, link length, link medium transmission rate, etc.);

[0053] In step 2), the offsets of all real-time flows are solved, and the process of allocating feasible network resources to the flows is performed by inputting the attributes of non-real-time flows. The specific process is as follows: Figure 2 As shown;

[0054] 201) assigning a route to each flow using the Dijkstra method according to the input flow attributes;

[0055] 202) Based on the assigned routing, flow attributes, and network attribute information, an integer linear programming (ILP) solver is used to find feasible queue assignments and offsets for each flow, with the following specific constraints:

[0056] Define the following notation: Each stream Has a unique category, i.e. belongs to the live stream collection Or belongs to the non-real-time stream collection f i The period is T i , the deadline is d i , in the physical link [v a , v b The execution transmission delay on ] is The propagation delay is Through the physical link [v a , v b The set of flows of ] is Also divided into real-time stream collections and non-real-time stream collections f i Relative to the super cycle (i.e. the least common multiple of all flow cycles) on the path [v a , v b ], that is, the offset on Time, turn on f i The queue Q i ;κ is the number of queues allocated to TT flows;

[0057] Determine the optimization goal: Minimize κ so that 7-Q i ≤κ;

[0058] The constraints for real-time streaming are as follows:

[0059] Schedulability Constraints: Flow f i From the starting point v s Any time the last bit in a packet is released, it reaches the destination v e The time must be less than or equal to f i Deadline d i ; If this constraint is satisfied, the flow schedulability and end-to-end delay constraints are guaranteed; this constraint contains f i The transmission time, the propagation delay on the link, and the waiting time in the queue are constrained by the following formula:

[0060]

[0061] Link constraint: Two packets passing through the same physical link cannot overlap in the time domain, that is, any flow f i Start by link [v a , v b ], recorded as No earlier than any flow f i In the link [v a , v b ]The time when the transmission is completed is obtained as follows:

[0062]

[0063] Where α = 0, 1, ..., H / T i -1, β = 0, i, ..., H / T j -1.

[0064] Streaming constraints: Stream f i The propagation of packets must follow the order of the flow routing path, that is, in flow f i The route exists, f i In [vx , v b ] starts transmission no earlier than the time at which the a , v x ]arrival x , v b ] time, the constraint formula is as follows:

[0065]

[0066] The network accuracy δ is expressed as the maximum difference between the local clocks of any two devices synchronized by the IEEE 802.1AS time synchronization standard. Using δ here can avoid transmission jitter caused by inconsistent global clocks.

[0067] Flow isolation constraint: In order to avoid the interweaving of two flows assigned to the same queue and ensure the determinism of flow transmission, it is necessary to isolate different flows assigned to the same queue. a , v x ] and [v x , v b ] is the flow i Two consecutive links in a route, [v x , v b ] is also a flow j The links in the route, and f i and f j In the network node v x are assigned to the same egress queue. Therefore, f i Reaching node v x The time should not be earlier than f j In [v x , v b ]The time when the transmission is completed, the constraint formula is as follows:

[0068]

[0069] like Figure 3 As shown, in step 4), the gated list content is output according to the offset of the stream, and the steps are as follows:

[0070] 401) Select any path from all paths [v a , v b ]∈Π;

[0071] 402) Initialized at path [v a , v b ] a GCL Content

[0072] 403) for the path [v a , vb ] on any non-retransmission flow f i , solve and queue switch status, and add it to Among them, the queue switch state Q i Set to 1, and the rest of the queue states are set to 0;

[0073] 404) for the path [v a , v b ] on any retransmission flow f RE，1 , solve and queue switch status, and add it to Among them, the queue switch state Q i Set to 1, all queues containing non-real-time streams are also set to 1, and the status of other queues is set to 0;

[0074] 405) Sort by time in descending order;

[0075] 406) The path [v a , v b ] is removed from all paths;

[0076] 407) Determine whether UI is empty, if it is empty, then end; otherwise, execute step 401).

[0077] All the following constraints are solved in step 403) and step 404) And the process of queue switch status

[0078] ARQ Constraint: f i The instance is transmitted to the next hop node. The next hop node passes the CRC. If the received data is correct, it returns an ACK message, otherwise it returns a NAK message. It should be noted that no matter what kind of message it is, it needs to comply with the TSN deterministic transmission requirements. Therefore, according to the above constraints, we can get ARQ relative to f i The offset

[0079]

[0080] Since the offset of the instance transmission on each link is periodic, the ARQ message can be considered as f i flows with the same period. For any f ARQ , in any link [v b , v a ], we can get the following constraints:

[0081]

[0082]

[0083] Reservation constraint: TT flow f i The instance from node v a By [v a , v b ]Transmit to node v b After that, there is a retransmission reservation time interval. During this interval, all flows in the same transmission direction cannot be transmitted. The following constraints can be obtained:

[0084]

[0085]

[0086]

[0087] For non-real-time streams, the present invention assumes that they also meet deterministic transmission. Therefore, the following constraints need to be met:

[0088] Streaming constraints: non-real-time streams i Each instance of should be in the path [v a , v b ] before starting the subsequent path [v b , v x ] on the transmission.

[0089]

[0090] Flow isolation constraint: To prevent instances of different non-TT flows from leaving the queue at the same time, an instance isolation constraint is introduced. This constraint specifies that as long as f i Examples of i If Q i =Q j , then another flow f j Instances of f cannot be added to this queue. j An instance of f can only be i The device can only be reached after the master instance of has left the queue and its transmission on the associated link has started, resulting in the following constraints:

[0091]

[0092] This embodiment takes an industrial field environment where the transmission of key data (ie, real-time streaming data) is affected by factors such as electromagnetic interference, resulting in the data being unable to be correctly received by the receiving end and requiring retransmission as an example.

[0093] First, the Dijkstra algorithm is used to generate a routing path for each flow. Then, TSN constraints are formulated so that all flows, including ARQ-related flows and retransmissions, are in deterministic transmission. According to the proposed constraints, the feasible offset of each flow is found through integer linear programming (ILP) to generate the offset of each export. Finally, in order to improve the throughput of the network, in the time slice reserved for retransmission, a queue is also opened for non-real-time flows so that the data of non-real-time flows can be sent when no real-time flow failure is found, realizing TSN fault-tolerant scheduling.

[0094] According to the properties of flows (including real-time flows, reserved retransmission flows, and ARQ-related flows), network resources, etc., under the constraints of deterministic transmission and IEEE 802.1Qbv standards, the queue open offset of each flow relative to the super period at each switch exit is solved in the queue assigned to the flow;

[0095] According to the attributes of the non-real-time flow and the remaining queue information in step 1), the queue open offset of each non-real-time flow at each switch exit relative to the super period is solved;

[0096] Generate a gate control list according to the offsets in step 1) and step 2);

[0097] When the queue for the reserved retransmission flow is opened, the queue for the non-real-time flow is also opened.

Claims

1. An adaptive TSN offline fault-tolerant scheduling method based on resource reservation, characterized in that: The following steps are involved: 1) Input initial conditions and initialize the network; 2) Solve the offsets for all real-time flows, and perform the process of allocating feasible network resources to the flows by inputting the attributes of non-real-time flows; 3) Determine whether there is a feasible solution for the real-time stream offset in step 2). If there is a feasible solution, allocate feasible network resources to all real-time streams, and execute the process of solving the stream offset by inputting non-real-time stream attributes; 4) Determine whether there is a feasible solution for the offset of the flow in step 3). If there is a feasible solution, output the gating list content according to the offset of the flow; In step 2), the offsets of all real-time flows are solved, and the process of allocating feasible network resources to the flows is performed by inputting the attributes of non-real-time flows as follows: 201) assigning a route to each flow using the Dijkstra method according to the input flow attributes; 202) Based on the assigned routing, flow attributes and network attribute information, an integer linear programming solver is used to find feasible queue assignments and offsets for each flow, with the following specific constraints: Define the following notation: Each stream Has a unique category, i.e. belongs to the live stream collection Or belongs to the non-real-time stream collection therefore f i The period is T i , the deadline is d i , in the physical link [v a , v b The execution transmission delay on ] is The propagation delay is Through the physical link [v a , v b The set of flows of ] is Also divided into real-time stream collections and non-real-time stream collections f i Relative to the superperiod H on the path [v a , v b ], that is, the offset on Time, turn on f i The queue Q i ;κ is the number of queues allocated to TT flows, TT flows are time-sensitive flows; The superperiod H is the least common multiple of all flow periods; Determine the optimization goal: Minimize κ so that 7-Q i ≤κ; The constraints for real-time streaming are as follows: Schedulability Constraints: Flow f i From the starting point v s Any time the last bit in a packet is released, it reaches the destination v e The time must be less than or equal to f i Deadline d i ; If this constraint is satisfied, the flow schedulability and end-to-end delay constraints are guaranteed; this constraint contains f i The transmission time, the propagation delay on the link, and the waiting time in the queue are constrained by the following formula: Link constraint: Two packets passing through the same physical link cannot overlap in the time domain, that is, any flow f j Start by link [v a , v b ], recorded as No earlier than any flow f i In the link [v a , v b ]The time when the transmission is completed is obtained as follows: Where α = 0, 1, ..., H / T i -1, β = 0, 1, ..., H / T j -1, T i For flow i The period, T j For flow j The period of , H is the super period; Flow transmission constraint: The propagation of packets of flow fi follows the order on the routing path of the flow, that is, i The route exists, f i In [v x , v b ] starts transmission no earlier than the time from [v a , v x ]arrival x , v b ] time, the constraint formula is as follows: Wherein, network accuracy δ is expressed as the maximum difference between the local clocks of any two devices synchronized by the IEEE 802.1AS time synchronization standard; Flow isolation constraint: Link [v a , v x ] and [v x , v b ] is the flow i Two consecutive links in a route, [v x , v b ] is also a flow j The links in the route, and f i and f j In the network node v x are assigned to the same egress queue, f i Reaching node v x The time must not be earlier than f j In [v x , v b ]The time when the transmission is completed, the constraint formula is as follows: In step 4), the gated list content is output according to the offset of the stream, and the steps are as follows: 401) Select any path [v a , v b ]∈Π; 402) Initialized at path [v a , v b ] a Gated list GCL content 403) for the path [v a , v b ] on any non-retransmission flow f i ,Will and queue switch status, added to In which, the queue switch state Q i Set to 1, and the rest of the queue states are set to 0; 404) for the path [v a , v b ] on any retransmission flow f RE ,Will and queue switch status, added to In which, the queue switch state Qi is set to 1, the queues where all non-real-time flows are located are also set to 1, and the states of the remaining queues are set to 0; For flow i In the link [v a , v b ]Offset during retransmission; 405) Sort by time in descending order; 406) Change the path [v a , v b ] is removed from all paths Π; 407) Determine whether UI is empty, if it is empty, then end; otherwise, execute step 401).

2. The adaptive TSN offline fault-tolerant scheduling method based on resource reservation according to claim 1 is characterized in that: If there is no feasible solution for the real-time flow offset in step 2), no feasible scheduling solution can be provided; or, if there is no feasible solution for the flow solution offset in step 4), no feasible scheduling solution can be provided.

Images