Time-deterministic traffic burst shaping method for distributed networks

Abstract

The present invention discloses a time-deterministic traffic burst shaping method for distributed networks, which is used to solve the problem that the existing technology cannot guarantee the timely transmission of traffic in the specified transmission time slot when traffic micro-bursts occur. The solution is: set up two layers of packet queues at the output port of the network switching node; set the second layer queue set S i The group maintenance weight value set W in i ; Determine the queue type to be imported according to the packet type arriving at the port, prioritize the time-sensitive packets, and add the time-sensitive packets of different priorities to the shaping queue; at the beginning of the transmission time slot, the shaping queue S i,j The time-sensitive packets in are imported into the deterministic queue T of the current transmission i After the transmission time slot ends, the remaining packets in the shaping queue are imported into the deterministic queue. This invention can ensure the time determinism of services, improve the overall network service quality, and avoid the formation of traffic microbursts. It can be used in the Internet, Internet of Things, mobile communications, and satellite networks.

Description

Technical Field

[0001] The present invention belongs to the field of communication technology, and in particular relates to a time-deterministic traffic burst shaping method, which can be used in the Internet, the Internet of Things, mobile communications and satellite networks. Background Art

[0002] With the development of network technology, emerging network applications such as the Industrial Internet, AR / VR, the Internet of Vehicles, ultra-high-definition video, telemedicine, and cloud gaming continue to emerge. These applications place stringent requirements on transmission link bandwidth, latency, and jitter. Therefore, exploring time-deterministic transmission methods for these applications has become a research hotspot in the communications field. However, existing time-deterministic network technologies have limitations in the distributed architecture of satellite networks. Due to the frequent topological changes and extended link times of satellite networks, it is difficult to synchronize transmission resource reservations and routing decision information between different satellite nodes. This can cause micro-bursts in service traffic at network switching nodes, making it difficult to ensure deterministic latency and jitter for latency-sensitive services.

[0003] To address the above-mentioned problems in satellite networks, it is necessary to consider the characteristics of distributed networks and study traffic burst shaping methods that ensure time determinism on the data forwarding plane, thereby achieving deterministic delay and jitter guarantees for business traffic.

[0004] The IEEE 802.1 Time-Sensitive Networking (TSN) Working Group has released a series of standard protocol documents, proposing mechanisms such as time synchronization, credit-based shapers (CBS), round-robin time slot queues, and frame deprivation to ensure time determinism for latency-sensitive services. The CBS mechanism allocates logical bandwidth to queues of different priorities, guiding the alternating transmission of packets in different queues, thereby avoiding starvation of low-priority queues caused by transmissions in high-priority queues. While TSN technology can create deterministic flows on demand in small-scale networks such as Layer 2 local area networks, it lacks deterministic latency guarantees in the existing Layer 3 large-scale IP routing and packet forwarding of the Internet.

[0005] The Internet Engineering Task Force (IETF) aims to combine Layer 2 and Layer 3 technologies to ensure the time determinism of service flows in large-scale networks and has established the Deterministic Networking (DetNet) Working Group. DetNet proposes a time slot-specified cyclic slot queue forwarding mechanism (CSQF) combined with segment routing (SR). Under reasonable SR routing decisions, CSQF carries the specified transmission time slot in the form of a segment identifier (SID) in the packet header, using the time slot as the scheduling granularity to guide the sending scheduling of packets in the packet forwarding link of each switching node, such as Figure 8However, this mechanism cannot handle the randomness and micro-bursts of the arrival process of service packets in distributed networks due to the untimely synchronization of distributed decision-making and resource reservation information. Figure 9 The delay accumulation and delay jitter caused by the failure of business packets to be transmitted in time within the specified transmission time slot ultimately make it impossible to achieve deterministic forwarding of business packets. Summary of the Invention

[0006] The purpose of the present invention is to address the above-mentioned problems and propose a time-deterministic traffic burst shaping method for distributed networks, so as to effectively solve the problem that the existing technology cannot guarantee the timely transmission of traffic in the specified transmission time slot when traffic microbursts occur, realize the deterministic forwarding of service packets, ensure the time determinism of the service, improve the overall service quality of the network, and avoid the formation of traffic microbursts.

[0007] The technical solution of the present invention is to set up a traffic shaping queue on the data transmission plane of the network switching node, divide the transmission priority of the arriving traffic, and give priority to the transmission of network control signaling and high-priority traffic. The implementation steps include the following:

[0008] (1) Set up two layers of packet queues at each output port of the network packet switching node:

[0009] (1a) Set the first-layer packet queue to M deterministic queues T i and 1 best-effort queue T BE , each deterministic queue T i The queue capacity is C, the best-effort queue is T BE The queue capacity is C # , 1≤i≤M;

[0010] (1b) Set the second-layer packet queue to M queue sets S i , each queue set S i Includes a high priority shaping queue S i,1 and a low-priority shaping queue S i,2 , is the shaping queue S i,j Allocate different logical bandwidths, S i,1 and S i,2 The logical bandwidth ratio is x:y, and each shaping queue S i,j The capacity is C * , 1≤j≤2;

[0011] (2) Each queue set S in the second layer i The groups in maintain a weight value set W i ;

[0012] (3) For the packet F that reaches each output port of the switching nodek , check its group type:

[0013] If group F k For network control signaling, the packet is directly imported into the deterministic queue T that is currently transmitting packets in the current transmission time slot. i middle;

[0014] If group F k For time-sensitive packets, the transmission time information of the segment identifier SID of the time-sensitive packet SR extension header at this node is obtained to determine the specific deterministic queue T to which the packet is imported. i , execute (4);

[0015] If group F k For best-effort packets, the packets are directed to the best-effort queue T. BE middle;

[0016] (4) For time-sensitive packets, determine the deterministic queue T to which they are imported. i After that, the queue priority is divided and the time-sensitive packets are added to the shaping queues of different priorities:

[0017] (4a) Extract the packet length information L from the time-sensitive packet header and perform normalization to obtain the normalized packet length W′ i,k ;

[0018] (4b) Extract the remaining hop count information W″ in the time-sensitive packet header i,k ;

[0019] (4c) Extract the transmission time information t at this node from the time-sensitive packet header send , and compare it with the system current time t current Take the difference and get the transmission time deviation W″′ i,k =t current -t send ;

[0020] (4d) Normalize the packet length W′ i,k , the remaining number of hops W″ i,k and transmission time deviation W″′ i,k Weighted summation is performed to obtain the weight value W of the time-sensitive group entry i,k =aW′ i,k +bW″ i,k +cW″′ i,k ;

[0021] (4e) Enqueue the time-sensitive group with weight W i,k and weight value set W i The statistic P i For comparison:

[0022] If W i,k ≥P i , then the time-sensitive packets are imported into the high-priority shaping queue S i,1 and enqueue it with weight W i,k Add weight value set W i ;

[0023] Otherwise, the time-sensitive packet is directed to the low-priority shaping queue S. i,2 and enqueue it with weight W i,k Add weight value set W i ;

[0024] (5) After the time-sensitive packets are imported into the shaping queues of different priorities, each transmission time slot obtains the deterministic queue T of the current transmission at the beginning. i , and within T i Under the condition of remaining transmission capacity, the corresponding shaping queue S i,j According to the allocated logical bandwidth, the time-sensitive packets in each shaping queue are imported into the current transmission deterministic queue T. i And the queue weight value W corresponding to the time-sensitive group i,k From the weight value set W i leave;

[0025] (6) After each transmission time slot ends, determine its deterministic queue T i Corresponding shaping queue S i,j Are there any groups remaining?

[0026] If yes, first queue the high priority shaping queue S i,1 The remaining packets in are imported into the deterministic queue T i , and then the low priority shaping queue S i,2 The remaining packets in are imported into the deterministic queue T i , complete the shaping of burst traffic;

[0027] Otherwise, there is no traffic conflict.

[0028] Compared with the prior art, the present invention has the following advantages

[0029] 1. It can ensure the time certainty of business.

[0030] The present invention divides transmission priorities for bursty traffic by setting a traffic shaping queue, gives network control signaling the highest priority, and can prioritize the transmission of network control signaling; at the same time, because the present invention comprehensively considers multi-dimensional packet information such as packet length, number of remaining hops of the packet, and the difference between the packet arrival time and its expected transmission time, it gives high priority to time-sensitive packets with larger packet length, more remaining hops, and larger difference between the arrival time and its expected transmission time. Therefore, when the service arrival presents a micro-burst phenomenon and the time slot transmission resources are insufficient, it can ensure that some high-priority time-sensitive services are transmitted in a timely manner in the specified transmission time slot, eliminate the delay accumulation and delay jitter of the service during the transmission process, and ensure the deterministic delay and delay jitter of the high-priority time-sensitive services.

[0031] 2. Improve the overall network service quality (QoS).

[0032] Due to the priority definition and division of time-sensitive services in the present invention, the deterministic delay and delay jitter guarantee capabilities of low-priority time-sensitive services are improved, compared with the existing mechanism, the overall network service quality QoS is improved, and the delivery rate of services in the network is increased.

[0033] 3. Avoid the formation of traffic micro-bursts.

[0034] The present invention sets a traffic shaping queue before the time-slot transmission of traffic and performs traffic shaping processing at the current node, so that the traffic no longer forms a micro-burst phenomenon at the next node, thereby avoiding the formation of traffic micro-bursts. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] Figure 1 It is a usage scenario diagram of the present invention;

[0036] Figure 2 This is a schematic diagram of the packet arrival at port C3 of switching node C in the usage scenario diagram of the present invention;

[0037] Figure 3 It is an implementation flow chart of the present invention;

[0038] Figure 4 It is a schematic diagram of the two-layer queue structure in the present invention;

[0039] Figure 5 This is a schematic diagram of the time-sensitive packet enqueue shaping queue in the present invention;

[0040] Figure 6 Schematic diagram of a deterministic queue for time-sensitive group enqueueing in the present invention;

[0041] Figure 7 This is a schematic diagram of grouping in a queue after the service arrival time slot ends in the present invention;

[0042] Figure 8 It is a schematic diagram of the existing mechanism for enqueuing;

[0043] Figure 9 This is a schematic diagram of grouping within a queue after the service arrival time slot of the existing mechanism ends. DETAILED DESCRIPTION

[0044] The embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0045] Reference Figure 1 The usage scenario of the present invention includes a first service source 1, a second service source 2, a first target network 1, a second target network 2, and four switching nodes A, B, C, and D; wherein the first service source 1 accesses the network through the first switching node A, and the second service source 2 accesses the network through the second switching node B, and the propagation delay between the switching nodes AC, BC, and CD is all 10ms; assuming but not limited to the transmission time slot width τ = 60us, the node port transmission bandwidth B = 1000MBps.

[0046] The packet length of the first service source 1 is 900 bytes, the end-to-end delay requirement is 25ms, and the jitter upper bound is 60us. The first service flow 1 is sent from the first service source 1, encapsulated in the routing packet header at the first switching node A, and the packet header instructs the first service flow 1 to be forwarded in the first deterministic queue of the first switching node A, forwarded in the third deterministic queue of the third switching node C, and forwarded in the second deterministic queue of the fourth switching node D, and then reaches the first target network 1, as shown in FIG. Figure 1 Indicated by the dotted line.

[0047] The packet length of the second service source 2 is 400 bytes, the end-to-end delay requirement is 15ms, and the jitter upper bound is 120us. The second service flow 2 is sent from the second service source 2, encapsulated in the routing header at the second switching node B, and the header instructs the second service flow 2 to be forwarded in the second deterministic queue of the second switching node B, and then forwarded in the third deterministic queue of the third switching node C, and then reaches the second target network 2. Figure 1 Indicated by the dotted line.

[0048] The first switching node A and the fourth switching node D interact with each other in network control signaling, such as Figure 1 Shown by the solid line.

[0049] Reference Figure 2 ,exist Figure 1In the scenario, different switching nodes have different packet arrival situations at different time points. In this example, assume that at the system time of the third switching node C at 21:32:22:233 milliseconds (640 microseconds), a burst of packets arrives at port C3. At this time, 13 packets arrive at port C3. The current time slot is transmitted by the deterministic queue T3, and T3 has only 3200 bytes of transmission capacity left in the current time slot. The 1st and 12th packets are network control signaling between the first switching node A and the fourth switching node D. The 2nd, 5th, and 9th packets arrive at port C3. 3 port and are arranged in other deterministic queues; the 3rd, 8th and 13th packets are time-sensitive packets for the first service flow 1, their transmission time at this node is 21:32:22 seconds 233 milliseconds 630 microseconds, and they are arranged in the deterministic queue T3; the 4th, 6th and 11th packets are best-effort packets arriving at port C3; the 7th and 10th packets are time-sensitive packets for the second service flow 2, their transmission time at this node is 21:32:22 seconds 233 milliseconds 640 microseconds, and they are arranged in the deterministic queue T3.

[0050] Reference Figure 3 , this example is Figure 1 Use scenarios and Figure 2 In the packet arrival scenario, the steps to implement traffic burst shaping include:

[0051] Step 1: Set up two layers of packet queues at each output port of the network packet switching node.

[0052] 1.1) Set the first-layer packet queue to M deterministic queues T i and 1 best-effort queue T BE , each deterministic queue T i The queue capacity is C, the best-effort queue is T BE The queue capacity is C # , 1≤i≤M;

[0053] In this example, based on the transmission time slot width and node port transmission bandwidth of the scenario, it is assumed that the queue capacity of each deterministic queue is C = 60KByte, and the best effort queue T BE The queue capacity is C # =30MByte, the number of deterministic queues M=3, set up three deterministic queues T1, T2, T3 and one best-effort queue T BE ;

[0054] 1.2) Set the second-layer packet queue to M queue sets S i , each queue set S i Includes a high priority shaping queue S i,1 and a low-priority shaping queue S i,2, is the shaping queue S i,j Allocate different logical bandwidths, S i,1 and S i,2 The logical bandwidth ratio is x:y, and each shaping queue S i,j The capacity is C * , 1≤j≤2;

[0055] In this example, based on the queue capacity C of the deterministic queue, assuming but not limited to the bandwidth ratio reserved for network control signaling is r = 0.1, the queue capacity C of each shaping queue is * =27KByte, set up three queue sets S1, S2, and S3 corresponding to the deterministic queues T1, T2, and T3 respectively. The first queue set S1 includes a high-priority shaping queue S 1,1 and a low-priority shaping queue S 1,2 , S 1,1 With S 1,2 The ratio of logical bandwidth is 9:1. The second queue set S2 includes a high priority shaping queue S 2,1 and a low-priority shaping queue S 2,2 , S 2,1 With S 2,2 The ratio of logical bandwidth is 9:1. The third queue set S3 includes a high priority shaping queue S 3,1 and a low-priority shaping queue S 3,2 , S 3,1 With S 3,2 The logic bandwidth ratio is 9:1, and we get Figure 4 The two-tier queue structure shown in the figure consists of three deterministic queues T1, T2, T3 and one best-effort queue T BE The second-level queue consists of 6 shaping queues S 1,1 , S 1,2 , S 2,1 , S 2,2 , S 3,1 , S 3,2 composition.

[0056] Step 2: For each queue set S in the second layer i The groups in maintain a weight value set W i .

[0057] In this example, a weight value set W1 is maintained for packets in the second layer queue set S1, a weight value set W2 is maintained for packets in the second layer queue set S2, and a weight value set W3 is maintained for packets in the second layer queue set S3.

[0058] Step 3: Determine the queue type to be imported based on the packet type arriving at the port

[0059] For the packet F arriving at each output port of the switching node k , check its group type:

[0060] If group F k For network control signaling, the packet is directly imported into the deterministic queue T that is currently transmitting packets in the current transmission time slot. i middle;

[0061] If group F k For time-sensitive packets, the transmission time information of the segment identifier SID of the time-sensitive packet SR extension header at this node is obtained to determine the specific deterministic queue T to which the packet is imported. i , then proceed to step 4;

[0062] If group F k For best-effort packets, the packets are directed to the best-effort queue T. BE middle.

[0063] In this example, for the 13 packets that arrive suddenly at the C3 port of the third switching node C, the first and 12 packets are determined to be network control signaling and are directly imported into the current transmission deterministic queue T3; the fourth, sixth, and eleventh packets are determined to be best-effort packets and are imported into the best-effort queue T3. BE ; Determine that the 2nd, 3rd, 5th, 7th, 8th, 9th, 10th, and 13th packets are time-sensitive packets, obtain the transmission time information at this switching node in the packet header, determine that the 5th packet is imported into the deterministic queue T1, determine that the 2nd and 9th packets are imported into the deterministic queue T2, and determine that the 3rd, 7th, 8th, 10th, and 13th packets are imported into the deterministic queue T3. This example only applies to the deterministic queue T3 and its corresponding shaping queue S 3,1 , S 3,2 The group changes in other deterministic queues and their corresponding shaping queues can be implemented with reference to this example.

[0064] Step 4: For time-sensitive packets, determine the deterministic queue T to which they are imported. i Afterwards, the queue entry priorities are divided and time-sensitive packets are imported into shaping queues of different priorities.

[0065] 4.1) Extract the packet length information L from the time-sensitive packet header and perform normalization to obtain the normalized packet length W i ' ,k ;

[0066] The methods for achieving normalization processing include maximum and minimum normalization, zero mean normalization, and neural network normalization. In this example, the normalization processing method used for the packet length information L is, but not limited to, maximum and minimum normalization, and the normalized packet length is obtained. Where MTU is the maximum transmission unit of the network, MTU = 1518 bytes. For the 3rd, 7th, 8th, 10th, and 13th packets to be imported into deterministic queue T3, the length information L is extracted from their IPv6 header payload field. The extracted packet lengths of the 3rd, 8th, and 13th packets are 900 bytes, and the normalized packet lengths are 0.593. The extracted packet lengths of the 7th and 10th packets are 400 bytes, and the normalized packet lengths are 0.264.

[0067] 4.2) Extract the remaining hop count information W″ in the time-sensitive packet header i,k ;

[0068] For the 3rd, 7th, 8th, 10th, and 13th packets to be imported into the deterministic queue T3, the remaining hop count information W" is extracted from the segment remaining field of their SR extension headers. i,k In this example, the number of remaining hops for the 3rd, 8th, and 13th groups extracted is 2, and the number of remaining hops for the 7th and 10th groups extracted is 1;

[0069] 4.3) Extract the transmission time information t from the time-sensitive packet header at this node send , and compare it with the system current time t current Take the difference and get the transmission time deviation W″ i,k =t current -t send ;

[0070] In this example, the current system time is 21:32:22:233 milliseconds, 640 microseconds. For the 3rd, 7th, 8th, 10th, and 13th packets to be imported into deterministic queue T3, the transmission time information at this switching node is extracted from the segment identifier SID of their SR extension header. The transmission time of the 3rd, 8th, and 13th packets at this switching node is 21:32:22:233 milliseconds, 630 microseconds, with a transmission time deviation of 0.00001s. The transmission time of the 7th and 10th packets at this node is 21:32:22:233 milliseconds, 640 microseconds, with a transmission time deviation of 0s.

[0071] 4.4) Normalize the packet length W′ i,k , the remaining number of hops W″ i,k and transmission time deviation W″′ i,k Weighted summation is performed to obtain the weight value W of the time-sensitive group entry i,k=aW′ i,k +bW″ i,k +cW″′ i,k , where a is the normalized packet length W′ i,k The weighted coefficient of b is the remaining hop number W″ i,k The weighting coefficient of , c is the transmission time deviation W″′ i,k The weighting coefficient of

[0072] In this example, it is assumed but not limited to that the weight coefficients a=1, b=1, c=1000, and the normalized packet length W′ of the packet is obtained. i,k , the remaining number of hops W″ i,k and transmission time deviation W″′ i,k After that, the calculated entry weight values ​​of the 3rd, 8th, and 13th groups are 2.603, and the calculated entry weight values ​​of the 7th and 10th groups are 1.264;

[0073] 4.5) Add the weight value W to the time-sensitive group i,k and weight value set W i The statistic P i For comparison:

[0074] If W i,k ≥P i , then the time-sensitive packets are imported into the high-priority shaping queue S i,1 and enqueue it with weight W i,k Add weight value set W i ;

[0075] Otherwise, the time-sensitive packet is directed to the low-priority shaping queue S. i,2 and enqueue it with weight W i,k Add weight value set W i ;

[0076] In this example, let but not be limited to the statistic P i is the weight value set W i the median;

[0077] Compare the third group's queue weight value 2.603 with the median of the queue weight value set W3 = {}, that is, the set is empty at this time, and the third group is queued into the high priority shaping queue S 3,1 , the enqueuing weight value 2.603 of the third group is added to the enqueuing weight value set W3;

[0078] Compare the 7th group's enqueuing weight value 1.264 with the median of the enqueuing weight value set W3 = {2.603}. The median is 2.603, 1.264 < 2.603, and the 7th group is enqueued in the low-priority shaping queue S. 3,2, the 7th group's entry weight value 1.264 is added to the entry weight value set W3;

[0079] Compare the 8th group's enqueuing weight value 2.603 with the median of the enqueuing weight value set W3 = {1.264, 2.603}. The median is 1.934, 2.603>1.934, and the 8th group is enqueued in the high-priority shaping queue S. 3,1 , the 8th group's entry weight value 2.603 is added to the entry weight value set W3;

[0080] Compare the 10th group's enqueuing weight value 1.264 with the median of the enqueuing weight value set W3 = {1.264, 2.603, 2.603}. The median is 2.603, 1.264 < 2.603, and the 10th group is enqueued in the low-priority shaping queue S. 3,2 , the 10th group's enqueuing weight value 1.264 is added to the enqueuing weight value set W3;

[0081] Compare the 13th group's enqueuing weight value 2.603 with the median of the enqueuing weight value set W3 = {1.264, 1.264, 2.603, 2.603}. The median is 1.934, 2.603>1.934, and the 13th group is enqueued in the high-priority shaping queue S. 3,1 , the 13th group's entry weight value 2.603 is added to the entry weight value set W3;

[0082] After prioritizing the time-sensitive packets and importing them into the shaping queue, the diagram of the time-sensitive packet queue is as follows: Figure 5 shown.

[0083] Step 5: At the beginning of each transmission time slot, the shaping queue S i,j The time-sensitive packets in are imported into the deterministic queue T of the current transmission i middle.

[0084] After time-sensitive packets are imported into shaping queues of different priorities, each transmission time slot obtains the deterministic queue T of the current transmission at the beginning. i , and within T i Under the condition of remaining transmission capacity, the corresponding shaping queue S i,j According to the allocated logical bandwidth, the time-sensitive packets in each shaping queue are imported into the current transmission deterministic queue T. i And the queue weight value W corresponding to the time-sensitive group i,k From the weight value set W i leave.

[0085] In this example, according to the shaping queue S set in step 1.2), 3,1 With S3,2 The logical bandwidth ratio is 9:1, and the CBS mechanism is used to shape the queue S 3,1 、S 3,2 The time-sensitive packets in the queue are imported into the deterministic queue T3 in the order of the 3rd, 7th, 8th, and 13th, and the corresponding enqueue weights of these packets, 2.603, 1.934, 2.603, and 2.603, are removed from the weight value set W3. After these packets are imported into the deterministic queue T3, the remaining transmission capacity of T3 is 100 bytes. The packet length of the 10th packet is greater than the remaining transmission capacity of T3, so the 10th packet is not imported into the deterministic queue T3. The result of importing the time-sensitive packets from the shaping queue into the deterministic queue is as follows: Figure 6 shown.

[0086] Step 6: After each transmission time slot ends, the remaining time-sensitive packets in the shaping queue are processed.

[0087] After each transmission time slot ends, its deterministic queue T is determined i Corresponding shaping queue S i,j Are there any groups remaining?

[0088] If yes, first queue the high priority shaping queue S i,1 The remaining packets in are imported into the deterministic queue T i , and then the low priority shaping queue S i,2 The remaining packets in are imported into the deterministic queue T i , complete the shaping of burst traffic;

[0089] Otherwise, there is no traffic conflict.

[0090] In this example, at the end of the transmission time slot of the deterministic queue T3, the shaping queue S 3,2 There is still a 10th packet left in the transmission time slot. The 10th packet is imported into the deterministic queue T3 to complete the shaping of the burst traffic. At the end of the transmission time slot, the remaining packets in the deterministic queue T3 are as follows: Figure 7 shown.

[0091] In this example, network control signaling is transmitted in the most timely manner, packets of service flow 1 are classified as high priority and transmitted first in the specified transmission time slot, delay and delay jitter are guaranteed, and packet delay and delay jitter of service flow 2 do not exceed the performance indicators.

[0092] The above description is only a specific example of the present invention. Obviously, after understanding the content and principles of the present invention, professionals in this field may make various modifications and changes in form and details without departing from the principles and structure of the present invention. For example, the number of deterministic queues, normalization processing method, selection of weight coefficients, logical bandwidth allocation method of priority queues, and transmission method between priority queues, etc., but these modifications and changes based on the principles of the present invention are still within the scope of protection of the claims of the present invention.

Claims

1. A time-deterministic traffic burst shaping method for distributed networks, characterized in that: These include: (1) Set up two layers of packet queues at each output port of the network packet switching node: (1a) Set the first-layer packet queue to M deterministic queues T i and 1 best-effort queue T BE , each deterministic queue T i The queue capacity is C, the best-effort queue is T BE The queue capacity is C # , 1≤i≤M; (1b) Set the second-layer packet queue to M queue sets S i , each queue set S i Includes a high priority shaping queue S i,1 and a low-priority shaping queue S i,2 , is the shaping queue S i,j Allocate different logical bandwidths, S i,1 and S i,2 The logical bandwidth ratio is x:y, and each shaping queue S i,j The capacity is C * , 1≤j≤2; (2) Each queue set S in the second layer i The groups in maintain a weight value set W i ; (3) For the packet F arriving at each output port of the switching node k , check its group type: If group F k For network control signaling, the packet is directly imported into the deterministic queue T that is currently transmitting packets in the current transmission time slot. i middle; If group F k For time-sensitive packets, the transmission time information of the segment identifier SID of the time-sensitive packet SR extension header at this node is obtained to determine the specific deterministic queue T to which the packet is imported. i , execute (4); If group F k For best-effort packets, the packets are directed to the best-effort queue T. BE middle; (4) For time-sensitive packets, determine the deterministic queue T to which they are imported. i After that, the queue priority is divided and time-sensitive packets are imported into shaping queues of different priorities: (4a) Extract the packet length information L from the time-sensitive packet header and perform normalization to obtain the normalized packet length W′ i,k ; (4b) Extract the remaining hop count information W″ in the time-sensitive packet header i,k ; (4c) Extract the transmission time information t at this node from the time-sensitive packet header send , and compare it with the system current time t current Take the difference and get the transmission time deviation W″′ i,k =t current -t send ; (4d) Normalize the packet length W′ i,k , the remaining number of hops W″ i,k and transmission time deviation W″′ i,k Weighted summation is performed to obtain the weight value W for time-sensitive group entry i,k =aW′ i,k +bW″ i,k +cW″′ i,k ; (4e) Enqueue the time-sensitive group with weight W i,k and weight value set W i The statistic P i For comparison: If W i,k ≥P i , then the time-sensitive packets are imported into the high-priority shaping queue S i,1 and enqueue it with weight W i,k Add weight value set W i ; Otherwise, the time-sensitive packet is directed to the low-priority shaping queue S. i,2 and enqueue it with weight W i,k Add weight value set W i ; (5) After the time-sensitive packets are imported into the shaping queues of different priorities, each transmission time slot obtains the deterministic queue T of the current transmission at the beginning. i , and within T i Under the condition of remaining transmission capacity, the corresponding shaping queue S i,j According to the allocated logical bandwidth, the time-sensitive packets in each shaping queue are imported into the current transmission deterministic queue T. i And the queue weight value W corresponding to the time-sensitive group i,k From the weight value set W i leave; (6) After each transmission time slot ends, determine its deterministic queue T i Corresponding shaping queue S i,j Are there any groups remaining? If yes, first queue the high priority shaping queue S i,1 The remaining packets in are imported into the deterministic queue T i , and then the low priority shaping queue S i,2 The remaining packets in are imported into the deterministic queue T i , complete the shaping of burst traffic; Otherwise, there is no traffic conflict.

2. The method according to claim 1, characterized in that The deterministic queue T in (1a) i The queue capacity C is calculated as follows: C=B*τ Where τ is the size of the transmission time slot and B is the transmission bandwidth of this output port.

3. The method according to claim 1, characterized in that The best effort queue T in (1a) BE The queue capacity C # ,It is set according to the hardware conditions of the network packet switching nodes.

4. The method according to claim 1, characterized in that The shaping queue S in (1b) i,j The queue capacity C * , which is calculated as follows: Where r is the bandwidth ratio reserved for network control signaling.

5. The method according to claim 1, characterized in that The network control signaling in (3) includes routing protocol packets, Internet Control Message Protocol ICMP packets and Resource Reservation Protocol RSVP packets.

6. The method according to claim 1, characterized in that The packet length information of the time-sensitive packet header is extracted and normalized in (4a), which is implemented as follows: (4a1) obtaining the length information L of the packet from the IPv6 header payload field of the time-sensitive packet; (4a2) The packet length L is normalized according to the following formula to obtain the normalized packet length W i ' ,k : Among them, MTU is the maximum transmission unit of the network.

7. The method according to claim 1, characterized in that The remaining hop count information W in (4b) i ' , ' k , which is obtained from the segment remainder field of the packet SR extended header.

8. The method according to claim 1, characterized in that The weight value set W in (4e) i The statistic P i Includes median, mean and mode.

9. The method according to claim 1, characterized in that The current transmission deterministic queue T in (5) i The remaining transmission capacity of the deterministic queue T i The amount of data that can be transmitted within the remaining transmission time of the current transmission time slot.

10. The method according to claim 1, characterized in that The deterministic queue T for the current transmission in (5) i Corresponding shaping queue S i,j According to the allocated logical bandwidth, the time-sensitive packets in each shaping queue are imported into the T i In the middle, the credit-based shaper CBS mechanism is used, that is, the high priority shaping queue S i,1 and low priority shaping queue S i,2 In an alternating manner, the time-sensitive packets in each queue enter the T i , the two queues are imported into T i The ratio of the number of packets is the ratio of the logical bandwidth allocated to the two queues, x:y.

Images