End-network coordinated ubiquitous network congestion control method

Abstract

The present application relates to the technical field of network communication congestion control, in particular to an end-network cooperative ubiquitous network congestion control method. The method is applied to a ubiquitous network comprising end nodes, intermediate nodes and content production nodes. The end nodes generate and send interest packets. The intermediate nodes obtain queue usage state, pending interest table occupation state, available bandwidth state and interest packet satisfaction state when the data packets return, generate a first congestion score or a second congestion score, and generate a congestion marking probability in combination with the occupation contribution of the target interest flow in the pending interest table to mark the return data packets with congestion. The intermediate nodes perform interest flow splitting or rate shaping according to the congestion level, and the end nodes adjust the subsequent interest packet sending window according to the congestion score information and the congestion marking result. The present application is suitable for high dynamic and multi-path ubiquitous network congestion control.

Description

Technical Field

[0001] This invention relates to the field of network communication congestion control technology, specifically to a ubiquitous network congestion control method based on end-to-end network collaboration. Background Technology

[0002] Ubiquitous networks typically consist of end nodes, intermediate nodes, and content production nodes, enabling content requests, data forwarding, and cache reuse in dynamic communication scenarios such as satellite networks, mobile networks, and edge networks. Existing congestion control methods generally involve end nodes adjusting their sending windows, or intermediate nodes performing congestion detection based on single or limited parameters such as queue length, packet loss, and link bandwidth, and then limiting the transmission rate of interest packets or adjusting forwarding paths accordingly.

[0003] However, in ubiquitous networks characterized by frequent link state changes, unstable cache hits, and multiple paths, relying solely on a single network attribute can easily misjudge temporary queuing, path switching, or cache failure as persistent congestion. Simultaneously, end nodes struggle to promptly ascertain the local congestion levels of intermediate nodes, and intermediate nodes also find it difficult to differentiate the actual contribution of different interest flows to congestion. This results in a lack of coordinated basis for window adjustments, rate limiting, and path switching, easily leading to packet loss, underutilization of links, and unfair control among different interest flows. Therefore, a ubiquitous network congestion control method that coordinates end-network operations is needed to address these issues. Summary of the Invention

[0004] To address the above problems, this invention provides the following technical solution: a ubiquitous network congestion control method based on end-to-end network collaboration, applied to a ubiquitous network comprising end nodes, intermediate nodes, and content production nodes, the method comprising: S1. End nodes generate and send interest packets, intermediate nodes forward the interest packets, and when the data packets corresponding to the interest packets return, they obtain network status indicators corresponding to the interest flow. The network status indicators include queue usage status, pending interest table occupancy status, available bandwidth status, and interest packet satisfaction status. S2. The intermediate node generates a first congestion score based on the queue usage status and the pending interest table occupancy status, and generates a second congestion score based on the network status index when the first congestion score meets the congestion triggering condition. S3. The intermediate node determines the congestion score information based on the first congestion score or the second congestion score, and generates a congestion marking probability based on the queue overload status and the occupancy contribution of the target interest flow in the unprocessed interest table. The return data packet is marked for congestion according to the congestion marking probability to obtain the marked data packet. S4. The intermediate node generates network-side control instructions based on the congestion level to which the congestion score information belongs. When the congestion level is a severe congestion level, it generates interest-based traffic splitting control instructions. When the congestion level is a mild congestion level, it generates rate-shaping control instructions. The node then adjusts the forwarding path or forwarding rate of subsequent interest packets according to the network-side control instructions. S5. The end node receives the marked data packet, generates an end-side congestion window adjustment instruction based on the congestion score information and the congestion marking result in the marked data packet, and adjusts the sending window of subsequent interest packets according to the end-side congestion window adjustment instruction.

[0005] Furthermore, the interest package includes a content name prefix field, a random number field, a start time field, and a traffic split ratio field; The labeled data packet includes a content name prefix field, a random number field, a data field, a cache hit field, a data generation time field, a congestion flag field, and a congestion score field; the congestion flag field is used to carry the congestion labeling result, and the congestion score field is used to carry the congestion score information.

[0006] Furthermore, the network status indicators are obtained as follows: The ratio of the occupied queue length of the intermediate node to the maximum queue capacity is determined as the queue utilization rate; The ratio of the number of currently pending interest table entries in the intermediate node to the maximum capacity of the pending interest table is determined as the pending interest table occupancy rate. The available bandwidth is determined based on the maximum bandwidth of the link, the bandwidth occupied by the link in the previous statistical period, the number of data bytes in the current statistical period, and the statistical duration. The ratio of the number of received data packets to the number of interest packets sent is defined as the interest packet satisfaction rate.

[0007] Furthermore, the step of generating a first congestion score based on the queue usage status and the pending interest table occupancy status, and generating a second congestion score based on the network status indicator when the first congestion score meets the congestion triggering condition, includes: The first congestion score is generated based on a weighted average of queue utilization and pending interest table occupancy. When the first congestion score is greater than the congestion trigger threshold, a second congestion score is generated based on the weighted result of queue utilization, pending interest table occupancy, available bandwidth, and interest packet satisfaction rate. When the first congestion score is not greater than the congestion trigger threshold, the first congestion score is used as congestion score information to indicate a warning state or a no-congestion state.

[0008] Furthermore, when the marked data packet passes through multiple intermediate nodes along the return path, each intermediate node compares the congestion score generated by its own node with the existing congestion score information in the marked data packet. If the congestion score generated by its own node is higher than the existing congestion score information, it updates the congestion score information in the congestion score field, so that the congestion score information received by the end node indicates a higher degree of congestion on the return path.

[0009] Furthermore, the step of generating congestion label probabilities based on the queue overload status and the occupancy contribution of the target interest stream in the pending interest table includes: The fair share of the queue for the target interest stream is determined based on the current queue length and the number of interest streams. When the queue occupancy of the target interest stream does not exceed the fair share of the queue, no congestion mark is applied to the return data packets corresponding to the target interest stream; When the queue occupancy of the target interest flow exceeds the fair share of the queue and the current queue length is greater than the queue safety threshold, the basic congestion probability is determined based on the current queue length, the queue safety threshold, and the maximum queue length. The congestion contribution weight is determined based on the ratio of the number of entries of the target interest flow in the pending interest table to the total number of entries of all interest flows in the pending interest table. The product of the basic congestion probability and the congestion contribution weight is determined as the congestion labeling probability of the target interest flow.

[0010] Furthermore, the generation method of the interest-based traffic splitting control command includes: The queue utilization rate, pending interest table occupancy rate, available bandwidth, and interest packet satisfaction rate of the candidate forwarding links are obtained respectively, and a second congestion score is generated for each candidate forwarding link based on the queue utilization rate, the pending interest table occupancy rate, the available bandwidth, and the interest packet satisfaction rate. The candidate forwarding link with the lowest second congestion score is identified as the target backup link; Within the return time window, count the number of congestion marks for each interest flow and the total number of congestion marks for all interest flows to determine the congestion impact of each interest flow. The migration quantity is determined based on the congestion impact, the transfer ratio, and the number of entries of the corresponding interest flow in the pending interest table, and the subsequent interest packets corresponding to the migration quantity are redirected to the target backup link.

[0011] Furthermore, the generation method of the rate shaping control command includes: The dynamic queue congestion threshold is determined based on the queuing target time, available outgoing link bandwidth, and average packet size. The shaping transmission rate of the current interest stream is determined based on the traffic segmentation ratio carried in the traffic segmentation ratio field, the dynamic queue congestion threshold, the buffer queue length of the current interest stream, the round-trip time, the local cache hit rate and the global cache hit rate of the current interest stream. The intermediate node updates the forwarding rate of subsequent interest packets for the current interest stream according to the shaped sending rate.

[0012] Furthermore, when an intermediate node detects that a link interface is disconnected and there are data packets that cannot be forwarded, it caches the data packets that cannot be forwarded in the content cache. Based on the cached data packets after removing the data fields, it adds interface information, cache freshness information, and announcement flags, generates a cache announcement packet, and announces it to neighboring nodes, so that neighboring nodes can obtain the corresponding cached data based on the content name prefix and interface information.

[0013] Furthermore, the step of generating an end-side congestion window adjustment instruction based on the congestion score information and the congestion marking result in the marked data packet includes: When the congestion score information is determined by the second congestion score and the second congestion score is in the severe congestion range, the congestion window is reduced at double speed according to the second congestion score; When the congestion score information is determined by the second congestion score and the second congestion score is in the mild congestion range, the congestion window is linearly reduced. When the congestion score information is determined by the first congestion score and the first congestion score is within the warning congestion interval, the congestion window is conservatively increased. When the congestion score information is determined by the first congestion score and the first congestion score is in a non-congestion interval, the congestion window is increased at double speed.

[0014] Compared with the prior art, the present invention has the following beneficial effects: This ubiquitous network congestion control method, which integrates end-to-end network collaboration, performs hierarchical congestion scoring at intermediate nodes based on queue usage status, pending interest table occupancy status, available bandwidth status, and interest packet fulfillment status. The congestion scoring information and congestion labeling results are then fed back to the end nodes along with return data packets. This allows end nodes to obtain the actual congestion level along the path, avoiding misjudgments caused by relying solely on end-side packet loss or window adjustments based on a single queue status. By generating congestion labeling probabilities based on the occupancy contribution of target interest flows in the pending interest table, the method can distinguish the impact of different interest flows on congestion formation, making congestion labeling less simplistic. It operates on all data packets or all interest flows, thus giving subsequent window reduction, interest splitting, and rate shaping a more specific target. By performing interest splitting under severe congestion levels and rate shaping under mild congestion levels, and having end nodes synchronously adjust the subsequent interest packet sending windows, it forms a coordinated closed loop between local link control on the network side and global sending control on the end side. This reduces data packet accumulation and invalid rate limiting caused by link state fluctuations, buffer hit changes, or multi-path switching, and improves the timeliness of congestion control, the rationality of path utilization, and the fairness of control among different interest flows in ubiquitous networks. Attached Figure Description

[0015] Figure 1 A schematic diagram illustrating the steps of the ubiquitous network congestion control method with end-to-end network collaboration provided by the present invention; Figure 2 This is a schematic diagram of satellite data transmission in a ubiquitous network. Figure 3 Diagram of the overall architecture for multi-indicator congestion control with end-to-end network coordination; Figure 4 A diagram showing the field format of interest packets and data packets; Figure 5 A schematic diagram of a single-consumer, single-path network scenario; Figure 6 A comparison chart of throughput under a single consumer, single path, and 10ms link latency. Figure 7 A comparison chart of throughput under a single consumer, single path, and 30 millisecond link latency. Figure 8 A comparison chart of throughput under a single consumer, single path, and 50 millisecond link latency. Figure 9 This is a schematic diagram of a multi-consumer single-path network scenario. Figure 10 A throughput graph for multiple consumers, single path, and multiple flows; Figure 11 A comparison chart of throughput across multiple consumers on a single path; Figure 12 A diagram illustrating a multi-consumer, multi-path network scenario; Figure 13A throughput graph for multiple consumers, multiple paths, and multiple data streams; Figure 14 A comparison chart of throughput for different multi-consumer, multi-path solutions; Figure 15 This is a real satellite scene image; Figure 16 Map showing the data download volume of various ground stations; Figure 17 A timeline showing the completion time of data downloads for various ground stations. Detailed Implementation

[0016] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0017] To enable those skilled in the art to better understand the present invention, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0018] Please see Figure 1 , Figure 1 A schematic diagram illustrating the steps of the ubiquitous network congestion control method with end-to-end network collaboration provided by the present invention; the ubiquitous network congestion control method with end-to-end network collaboration is applied to a ubiquitous network including end nodes, intermediate nodes, and content production nodes, and the method includes: S1. End nodes generate and send interest packets, intermediate nodes forward the interest packets, and when the data packets corresponding to the interest packets return, they obtain network status indicators corresponding to the interest flow. The network status indicators include queue usage status, pending interest table occupancy status, available bandwidth status, and interest packet satisfaction status. Specifically, such as Figure 2 As shown, end nodes can be user nodes, ground station nodes, or other content request nodes in a satellite communication scenario; intermediate nodes can be satellite nodes, routing nodes, or edge forwarding nodes; and content production nodes can be servers providing target content data, satellite payload nodes, or ground content source nodes. End nodes generate interest packets based on content name prefixes and send these packets to content production nodes or intermediate nodes that cache the corresponding content via intermediate nodes. When the corresponding data packet returns along the reverse path established by the interest packet, the intermediate node reads its local forwarding queue, pending interest table, link bandwidth statistics, and the send / receive records of the interest packet and data packet at the moment the data packet passes through it, thereby obtaining network status indicators related to the current interest flow. Figure 3The diagram illustrates the collaborative architecture between end nodes, intermediate nodes, and content production nodes. The end nodes are responsible for sending interest packets, adjusting congestion windows, and handling timeouts. The intermediate nodes are responsible for congestion detection, congestion marking, link disconnection caching, rate shaping, and interest splitting. The content production nodes return the corresponding data content according to the content name in the interest packet.

[0019] Furthermore, in one embodiment provided in this application, the interest package includes a content name prefix field, a random number field, a start time field, and a traffic split ratio field; The labeled data packet includes a content name prefix field, a random number field, a data field, a cache hit field, a data generation time field, a congestion flag field, and a congestion score field; the congestion flag field is used to carry the congestion labeling result, and the congestion score field is used to carry the congestion score information.

[0020] Specifically, such as Figure 4 As shown, the content name prefix field in the interest packet indicates the content object requested by the end node, the random number field distinguishes different request instances with the same content name prefix, the start time field records the sending time of the interest packet entering the network, and the traffic splitting ratio field indicates the proportion of the current interest flow in all interest flows sent by the end node. After receiving the interest packet, the intermediate node can use the content name prefix field to determine the forwarding direction or buffer hit status, use the random number field to distinguish duplicate requests, use the start time field to calculate the time the interest packet has spent in the network, and use the traffic splitting ratio field to provide the end-side traffic ratio basis for subsequent rate shaping and buffer occupancy control.

[0021] The content name prefix field and random number field in the tagged data packet are used for matching with previously sent interest packets. The data field carries the requested content, the cache hit field indicates whether the data packet was returned after being cached by an intermediate node, the data generation time field records the time when the data packet was generated or returned, the congestion flag field carries the congestion marking result generated by the intermediate node, and the congestion score field carries the congestion score information calculated or updated by the intermediate node. After receiving the tagged data packet, the end node can determine whether the data packet has been marked as a congestion-related data packet by the intermediate node through the congestion flag field, and determine the high degree of congestion on the return path through the congestion score field.

[0022] Furthermore, in one embodiment provided in this application, the network status indicator is obtained as follows: The ratio of the occupied queue length of the intermediate node to the maximum queue capacity is determined as the queue utilization rate; The ratio of the number of currently pending interest table entries in the intermediate node to the maximum capacity of the pending interest table is determined as the pending interest table occupancy rate. The available bandwidth is determined based on the maximum bandwidth of the link, the bandwidth occupied by the link in the previous statistical period, the number of data bytes in the current statistical period, and the statistical duration. The ratio of the number of received data packets to the number of interest packets sent is defined as the interest packet satisfaction rate.

[0023] Specifically, intermediate nodes maintain network status statistics for each outgoing interface, each candidate link, or each statistical period. Let the first node be... The length of the occupied queue of each intermediate node is The maximum capacity of the queue is Then the queue utilization rate is:

[0024] in, This indicates the queue usage status; a higher value indicates a more severe queuing of data packets at intermediate nodes. If... If the value is zero or the queue is not initialized, the intermediate node will not use the queue utilization rate for that period in the congestion scoring, or the state will be set to a preset safe value to avoid misjudgment caused by abnormal parameters.

[0025] Let the first The number of interest table entries currently pending processing at each intermediate node is: The maximum capacity of the interest table to be processed is The occupancy rate of the interest table to be processed is:

[0026] in, This indicates the occupancy status of interest packets that have not yet returned data in intermediate nodes. The occupancy rate of the pending interest table can reflect the queuing pressure that may be formed by future return data packets, so this indicator can provide a congestion warning before data packets actually pile up.

[0027] Assume the maximum bandwidth of the link is The bandwidth occupied by the links in the previous statistical period was The number of data bytes counted in the current statistical period is The statistical duration is The smoothing coefficient is The available bandwidth status is as follows:

[0028] in, This indicates the remaining capacity of the current link. In actual implementation, intermediate nodes can... The bandwidth state is normalized to a value in the same direction as the congestion level, allowing it to participate in a weighted calculation along with queue utilization, pending interest table occupancy, and interest packet fulfillment rate. A low available bandwidth indicates that the link can continue to carry a limited amount of data; a high available bandwidth indicates that the link still has the capacity to forward interest packets and data packets.

[0029] Let the number of data packets received in the statistics window be... The number of interest packets sent is The interest package satisfaction rate is:

[0030] in, This is used to characterize whether interest packets sent by end nodes or intermediate nodes can be effectively satisfied by corresponding data packets. If interest packets are continuously sent but insufficient corresponding data packets are returned, it may indicate that the link is congested, the path is switched, the buffer has expired, or the data packets are being dropped. Figure 5 The diagram illustrates a single-consumer, single-path network scenario where the consumer requests content from the producer via an intermediate satellite node. Link bandwidth, link latency, and random packet loss rate are used as test conditions to observe the throughput changes of different congestion control methods under a single transmission path. Figure 6 , Figure 7 and Figure 8 The throughput comparison results are shown for link delays of 10 ms, 30 ms, and 50 ms, respectively. In this type of graph, the horizontal change corresponds to the simulation running time, and the vertical change corresponds to the link throughput. By comparing the curve changes under different link delays, we can observe the ability of this method to utilize the available bandwidth of the link under short, medium, and high delay conditions, as well as the throughput convergence after the combined effect of end-side window adjustment and intermediate node rate shaping.

[0031] S2. The intermediate node generates a first congestion score based on the queue usage status and the pending interest table occupancy status, and generates a second congestion score based on the network status index when the first congestion score meets the congestion triggering condition. Specifically, when data packets return or the statistical period arrives, the intermediate node first performs a congestion assessment based on the queue usage status and the pending interest table occupancy status. The queue usage status directly reflects the current data packet backlog, while the pending interest table occupancy status reflects the unmet interest packet size. Both are relatively less affected by short-term factors such as cache hits and temporary link jitter, making them suitable as the basis for congestion triggering judgment. The intermediate node first generates a first congestion score and compares it with the congestion triggering threshold. Only when the first congestion score meets the congestion triggering condition does it further utilize the available bandwidth status and interest packet fulfillment status to generate a second congestion score. Through this hierarchical scoring method, the intermediate node can reduce computational load under low load or no congestion conditions and perform fine-grained multi-indicator judgments when congestion may occur.

[0032] Furthermore, in one embodiment provided by this application, the step of generating a first congestion score based on the queue usage status and the pending interest table occupancy status, and generating a second congestion score based on the network status indicator when the first congestion score meets the congestion triggering condition, includes: The first congestion score is generated based on a weighted average of queue utilization and pending interest table occupancy. When the first congestion score is greater than the congestion trigger threshold, a second congestion score is generated based on the weighted result of queue utilization, pending interest table occupancy, available bandwidth, and interest packet satisfaction rate. When the first congestion score is not greater than the congestion trigger threshold, the first congestion score is used as congestion score information to indicate a warning state or a no-congestion state.

[0033] Specifically, the intermediate nodes first calculate the first congestion score:

[0034] in, This indicates the first congestion score. Indicates queue utilization weight. This represents the weight of the pending interest table occupancy rate. The above weights can be pre-configured based on network type, node cache capacity, maximum queue length, maximum capacity of the pending interest table, and link stability, or updated during network operation based on historical congestion status. If... If the value is not greater than the congestion trigger threshold, the intermediate node will use the first congestion score as the congestion score information, and further determine the current state as a no-congestion state or a warning state based on the range in which the first congestion score is located.

[0035] when When the congestion score exceeds the congestion trigger threshold, the intermediate node further calculates a second congestion score:

[0036] in, This indicates the second congestion score. Indicates the available bandwidth state weight. This indicates that the interest package satisfies the state weight. In actual implementation, it can be... and After converting the data into a normalized state variable that is in the same direction as the congestion level, it is substituted into the above formula so that a larger value for the second congestion score indicates a higher degree of congestion. Intermediate nodes determine the current congestion level as mild or severe based on the range in which the second congestion score falls. Figures 6 to 8 The throughput comparison shown under different link delays can be used to illustrate the impact of the hierarchical scoring strategy on throughput convergence and link utilization under different delay conditions. When the link delay increases, relying solely on end-side feedback can easily lead to window adjustment lag, while intermediate nodes can provide more timely congestion status feedback in the return path through the first congestion score and the second congestion score.

[0037] S3. The intermediate node determines the congestion score information based on the first congestion score or the second congestion score, and generates a congestion marking probability based on the queue overload status and the occupancy contribution of the target interest flow in the unprocessed interest table. The return data packet is marked for congestion according to the congestion marking probability to obtain the marked data packet. Specifically, when the first congestion score does not trigger the second-stage refined analysis, the intermediate node determines the first congestion score as the congestion score information; when the first congestion score triggers the second-stage refined analysis, the intermediate node determines the second congestion score as the congestion score information. The intermediate node writes the congestion score information into the congestion score field of the return data packet and determines the congestion labeling probability of the return data packet corresponding to the target interest flow based on whether the current queue exceeds the queue safety threshold and the occupancy contribution of the target interest flow in the pending interest table. Subsequently, the intermediate node writes the congestion flag field of the return data packet according to this congestion labeling probability, forming a labeled data packet. This processing method ensures that data packets corresponding to different interest flows have different congestion labeling probabilities, avoiding indiscriminate labeling of all data packets or all interest flows.

[0038] Furthermore, in one embodiment provided in this application, when the marked data packet passes through multiple intermediate nodes along the return path, each intermediate node compares the congestion score generated by its own node with the existing congestion score information in the marked data packet, and updates the congestion score information in the congestion score field when the congestion score generated by its own node is higher than the existing congestion score information, so that the congestion score information received by the end node indicates a higher degree of congestion on the return path.

[0039] Specifically, such as Figure 2As shown, in satellite networks or other ubiquitous networks, data packets may pass through multiple intermediate nodes before returning to the end node. Each intermediate node calculates its own congestion score when a data packet passes through it and reads the existing congestion score field from the labeled data packet. If the congestion score generated by the node is lower than or equal to the existing congestion score, the congestion score field remains unchanged to avoid low-congestion nodes overwriting feedback from high-congestion nodes; if the congestion score generated by the node is higher than the existing congestion score, the congestion score field is updated to the congestion score generated by the node. Therefore, the labeled data packets ultimately received by the end node can represent the high level of congestion on the return path, and the end node does not need to know the complete state of each intermediate node to perform global window adjustments based on this score.

[0040] Furthermore, in one embodiment provided in this application, the step of generating the congestion label probability based on the queue overload state and the occupancy contribution of the target interest stream in the pending interest table includes: The fair share of the queue for the target interest stream is determined based on the current queue length and the number of interest streams. When the queue occupancy of the target interest stream does not exceed the fair share of the queue, no congestion mark is applied to the return data packets corresponding to the target interest stream; When the queue occupancy of the target interest flow exceeds the fair share of the queue and the current queue length is greater than the queue safety threshold, the basic congestion probability is determined based on the current queue length, the queue safety threshold, and the maximum queue length. The congestion contribution weight is determined based on the ratio of the number of entries of the target interest flow in the pending interest table to the total number of entries of all interest flows in the pending interest table. The product of the basic congestion probability and the congestion contribution weight is determined as the congestion labeling probability of the target interest flow.

[0041] Specifically, the intermediate node records the current queue length as... The current number of interest streams is recorded as Then the fair share of the queue for the target interest stream can be expressed as:

[0042] in, This indicates the fair share of the target interest flow in the current queue state. The queue occupancy of the target interest flow can be estimated by the number of packets belonging to that interest flow in the current queue, the buffer length of packets belonging to that interest flow, or the number of pending interest table entries corresponding to that interest flow. When the queue occupancy of the target interest flow does not exceed... Even if there is some queuing at the nodes, the intermediate nodes will not mark the return data packets corresponding to the target interest flow as congested, so as to avoid the small interest flow being suppressed prematurely.

[0043] When the queue occupancy of the target interest stream exceeds And the current queue length Greater than the queue safety threshold At that time, intermediate nodes calculate the basic congestion probability:

[0044] in, This represents the base congestion probability determined by the degree of queue overload. This indicates the maximum length of the queue. If... ,but Set to zero; if the calculation result exceeds the probability range, then limit it to zero. Within the interval. Subsequently, the intermediate nodes calculate the occupancy contribution weight of the target interest stream in the interest table to be processed:

[0045] in, This indicates the number of entries in the interest table to be processed in this node for the target interest stream. This represents the total number of entries in the interest table to be processed at this node for all interest flows. If A value of 0 indicates that there are currently no unsatisfied interest flows available for statistical analysis, and intermediate nodes do not generate congestion marking probabilities. The congestion marking probability of the target interest flow is:

[0046] in, Used to determine whether the return data packets corresponding to the target interest stream are written to the congestion flag. Because The probability of marking an interest stream is related to the extent to which the target interest stream occupies resources in the pending interest table. Therefore, interest streams that occupy more resources in the pending interest table will have a higher marking probability, while interest streams that occupy fewer resources will not be equally suppressed due to global congestion.

[0047] Figure 9 The diagram illustrates a multi-consumer single-path network scenario, which includes multiple consumers, multiple producers, and intermediate satellite nodes, with multiple interest flows sharing the same path and competing for link resources. Figure 10 The diagram illustrates the throughput variations of multiple flows using this method in this scenario. Figure 10In the initial stage, only the first data stream exists, which converges quickly and occupies most of the link bandwidth. Around 20 seconds later, the second data stream starts, and the two interest streams begin competing for the same path resource. After a brief adjustment, the two data streams tend to allocate link bandwidth fairly. Around 40 seconds later, the third data stream starts, and the three interest streams compete for the same path resource. After convergence, each of the three data streams maintains approximately one-third of the bandwidth usage until the simulation ends. This result corresponds to the aforementioned fair share judgment and the probabilistic labeling mechanism based on the contribution of the interest table to congestion, indicating that the intermediate node can generate differentiated congestion labels based on the contribution of different interest streams to congestion formation, rather than allowing the interest stream that starts first or occupies more resources to maintain an advantage in the long term.

[0048] Figure 11 The figure shows a comparison of throughput results for different congestion control schemes in a multi-consumer single-path scenario. This figure compares the throughput performance of our proposed method with the first, second, third, and fourth comparison schemes under the same single-path multi-flow contention condition. The figure shows that although the instantaneous throughput of a single data flow may not always be the highest in our method, the total throughput of multiple data flows is higher, and the throughput of the three data flows is closer, which can prevent a single interest flow from excessively occupying link resources. These experimental results correspond to the multi-index congestion detection, target interest flow congestion contribution calculation, and congestion label probability generation mechanism, indicating that the intermediate node considers both queue status and the contribution of pending interest table occupancy when generating congestion labels, thus balancing total throughput and inter-flow fairness.

[0049] S4. The intermediate node generates network-side control instructions based on the congestion level to which the congestion score information belongs. When the congestion level is a severe congestion level, it generates interest-based traffic splitting control instructions. When the congestion level is a mild congestion level, it generates rate-shaping control instructions. The node then adjusts the forwarding path or forwarding rate of subsequent interest packets according to the network-side control instructions. Specifically, intermediate nodes map congestion score information to different congestion levels. When the congestion score indicates severe congestion, it means that the current forwarding path may be experiencing queue overflow, insufficient bandwidth, or deterioration in the satisfaction status of interest packets. In this case, the intermediate node generates an interest-based traffic splitting control command to migrate some subsequent interest packets from the current high-load path to a candidate path with a lower congestion score. When the congestion score indicates mild congestion, it means that the current link is still usable, but the short-term traffic exceeds the port's sending capacity or the network experiences slight fluctuations. In this case, the intermediate node generates a rate-shaping control command to limit or smooth the forwarding rate of subsequent interest packets on this link, without immediately performing large-scale path migration. Figure 3 The rate shaping mechanism and forwarding offloading function shown correspond to the two types of network-side control commands mentioned above.

[0050] Furthermore, in one embodiment provided in this application, the generation method of the interest-based traffic control command includes: The queue utilization rate, pending interest table occupancy rate, available bandwidth, and interest packet satisfaction rate of the candidate forwarding links are obtained respectively, and a second congestion score is generated for each candidate forwarding link based on the queue utilization rate, the pending interest table occupancy rate, the available bandwidth, and the interest packet satisfaction rate. The candidate forwarding link with the lowest second congestion score is identified as the target backup link; Within the return time window, count the number of congestion marks for each interest flow and the total number of congestion marks for all interest flows to determine the congestion impact of each interest flow. The migration quantity is determined based on the congestion impact, the transfer ratio, and the number of entries of the corresponding interest flow in the pending interest table, and the subsequent interest packets corresponding to the migration quantity are redirected to the target backup link.

[0051] Specifically, intermediate nodes first calculate a second congestion score for each candidate forwarding link, and then determine the candidate forwarding link with the lowest second congestion score as the target backup link. This selection method does not simply choose the link with fewer hops, the shortest path, or historical reachability, but rather judges the actual availability of candidate links based on queue utilization, pending interest table occupancy, available bandwidth, and interest packet fulfillment rate. For interest flows that need to be split, intermediate nodes calculate their congestion impact based on the return time window. Let the packet reception time be... The data was generated at the time of... The return time window is:

[0052] in, Used to define the time range for congestion labeling statistics. Assume the target interest flow falls within the time window. The number of congestion markers is The total number of interest streams marked as congested within this time window is The congestion impact of the target interest stream is then:

[0053] in, This indicates the relative impact of the target interest flow on the current congestion. When If no valid congestion marker is formed within the statistical window, the intermediate node will not perform traffic diversion. Let the diversion ratio be... The number of entries for the target interest stream in the interest table to be processed is The number of migrations is:

[0054] in, This indicates the number of subsequent interest packets that need to be migrated to the target backup link, or the corresponding scale of the interest table migration to be processed. Intermediate nodes can select the appropriate number of interest packets from subsequent arrivals with the same content name prefix, the same interest stream identifier, or the same end node request sequence, and redirect them to the target backup link.

[0055] Figure 12 The diagram illustrates a multi-consumer, multi-path network scenario where multiple consumers access multiple producers through intermediate satellite nodes, and the same consumer can form multiple path data streams. Figure 13 The results show the changes in multi-data stream throughput of this method in a multi-consumer, multi-path scenario. Figure 13 In the initial stage, there are two sub-streams that converge quickly and approximately equally share the link bandwidth. After about 20 seconds, a new data stream starts, competing for path resources with the existing two sub-streams. After convergence, the new data stream and the existing two sub-streams form a relatively stable bandwidth allocation relationship. After about 60 seconds, when some data streams end, the remaining two sub-streams re-compete for link resources and maintain a relatively balanced throughput after convergence. This process corresponds to the candidate link scoring, target backup link selection, and migration quantity determination mechanism, indicating that intermediate nodes can dynamically distribute subsequent interest packets based on multi-path status and flow-level congestion impact, rather than passively narrowing the window by end nodes after path congestion.

[0056] Figure 14 This figure shows a comparison of multi-data-stream throughput results for different congestion control schemes in a multi-path scenario. It compares the throughput and fairness of the proposed method with the first, second, third, and fourth comparison schemes under multi-path, multi-stream contention conditions. The figure shows that the throughput of the two streams in the proposed method is closer, and the overall throughput is higher than the total throughput of the corresponding data streams in the various comparison schemes. In contrast, while some comparison schemes may achieve higher instantaneous throughput for a single stream, the differences between different streams are significant, easily leading to uneven distribution of link resources. These results demonstrate that the proposed method uses multi-index scoring to understand the current congestion state, future link load, and link carrying capacity, and selects interest-based traffic splitting or rate shaping based on congestion levels, thus simultaneously considering link utilization and inter-stream fairness in a multi-path scenario.

[0057] Furthermore, in one embodiment provided in this application, the generation method of the rate shaping control command includes: The dynamic queue congestion threshold is determined based on the queuing target time, available outgoing link bandwidth, and average packet size. The shaping transmission rate of the current interest stream is determined based on the traffic segmentation ratio carried in the traffic segmentation ratio field, the dynamic queue congestion threshold, the buffer queue length of the current interest stream, the round-trip time, the local cache hit rate and the global cache hit rate of the current interest stream. The intermediate node updates the forwarding rate of subsequent interest packets for the current interest stream according to the shaped sending rate.

[0058] Specifically, intermediate nodes prioritize rate shaping under mild congestion conditions. Let the queuing target time be... The available bandwidth of the outgoing link is The average size of the data packets is The dynamic queue congestion threshold is:

[0059] in, This indicates the target length of the packet queue allowed under the current link conditions. This threshold is not a fixed constant but updates dynamically as the available bandwidth of the outgoing link changes. When the available bandwidth of the link decreases, This corresponding reduction causes intermediate nodes to lower their interest packet forwarding rate earlier; when the available bandwidth of the link recovers, The capacity is increased accordingly, enabling intermediate nodes to utilize the restored link capacity.

[0060] Let the current time be The current available bandwidth of the link is The proportion of target interest streams carried by the traffic segmentation ratio field is The average buffer quota per stream is The current buffer queue length of the target interest stream is Round-trip delay is The local cache hit rate of the target interest stream is Global cache hit rate The adjustment coefficient is and Then the current interest stream's shaping transmission rate is:

[0061] in, This indicates the forwarding rate of subsequent interest packets from the current interest stream by the intermediate node; The number of interest streams for the current activity. The average per-flow buffer quota is allocated based on the dynamic queue congestion threshold. This indicates the target buffer occupancy allocated to this interest stream based on the traffic segmentation ratio; This indicates the length of the buffer currently occupied by the interest stream; Used to convert buffer offset into a transmission rate adjustment amount; This is used to reflect the difference between the local cache hit rate of an interest stream and the global cache hit rate. When the local cache hit rate of an interest stream is high, intermediate nodes allow it to maintain a high forwarding rate because its subsequent packets are more likely to be returned from the local cache. When the local cache hit rate of an interest stream is lower than the global average and the queue occupancy is high, intermediate nodes reduce the shaping and sending rate of that interest stream to avoid the continuous accumulation of remote return data.

[0062] Furthermore, in one embodiment provided in this application, when an intermediate node detects that the link interface is disconnected and there are data packets that cannot be forwarded, it caches the data packets that cannot be forwarded in the content cache, and adds interface information, cache freshness information and announcement flags to the cached data packets after removing the data fields, generates a cache announcement packet and announces it to the neighboring nodes, so that the neighboring nodes can obtain the corresponding cached data according to the content name prefix and interface information.

[0063] Specifically, such as Figure 2 and Figure 15 As shown, in real satellite topologies or other highly dynamic ubiquitous networks, link interfaces may be disconnected due to satellite movement, inter-satellite link switching, obstruction, beam coverage changes, or route updates. When an intermediate node detects that an interface is disconnected and it has data packets that cannot be forwarded through that interface, the intermediate node writes the data packet to its local content cache and generates a cache announcement packet. The cache announcement packet can remove the complete data fields while retaining matching information such as the content name prefix and random number, and add interface information, cache freshness information, and an announcement flag. The interface information indicates which interface neighboring nodes can reach the cached data through, the cache freshness information determines whether the cached content can still be used to respond to subsequent interest packets, and the announcement flag indicates that the data packet is a cache announcement packet rather than a regular data packet. After receiving the cache announcement packet, the neighboring node updates its local forwarding or cache index according to the content name prefix. When a subsequent end node requests the same content, the neighboring node can obtain the corresponding cached data from the intermediate node that announced the cache based on the interface information, thereby avoiding repeatedly requesting the same data from the remote content production node after the link is disconnected.

[0064] Figure 15 The simulation presents a real satellite scenario, which includes satellite nodes, ground station nodes, and content request paths. It is used to simulate a communication environment with frequent changes in satellite links, parallel requests from multiple ground stations, and the simultaneous existence of single and multiple paths. Figure 16 The results show a comparison of data download volume at various ground stations in this real satellite scenario. The horizontal axis corresponds to different ground stations or different schemes, and the vertical axis corresponds to the cumulative data download volume completed within the specified operating time. Figure 16In the data analysis, the cumulative data download volume of this method across multiple ground stations was higher than that of several comparative schemes, with the data download volume at multiple ground stations approaching the target download scale. In some multi-path ground station scenarios, the download volume of the comparative schemes was significantly lower than that of this method due to insufficient congestion assessment or inadequate path control. This result corresponds to the comprehensive transmission capability resulting from the combined effects of multi-index congestion scoring, link disconnection buffering notification, interest-based traffic splitting, and rate shaping, indicating that in real satellite topologies, this method can reduce download interruptions or invalid waiting caused by link switching and local congestion.

[0065] S5. The end node receives the marked data packet, generates an end-side congestion window adjustment instruction based on the congestion score information and the congestion marking result in the marked data packet, and adjusts the sending window of subsequent interest packets according to the end-side congestion window adjustment instruction.

[0066] Specifically, after receiving a labeled data packet, the end node reads the congestion score field and the congestion flag field. If the congestion flag field indicates that the data packet is not labeled and the congestion score information is in the no-congestion range, the end node can increase the subsequent interest packet sending window. If the congestion flag field indicates that the data packet is labeled, or the congestion score information is in the warning, light, or severe congestion range, the end node adjusts the window accordingly based on the different levels. In this way, the local congestion awareness of intermediate nodes can be fed back to the end nodes with the return data packets, and the global sending window adjustment of the end nodes can reduce the amount of subsequent interest packets injected, thus forming a closed loop of end-network collaboration. Figure 17 The results show a comparison of data download completion times for various ground stations in a real satellite scenario. The horizontal axis corresponds to different ground stations or different schemes, and the vertical axis corresponds to the time required to complete the corresponding data download task. Figure 17 In the comparison, our method achieved shorter download completion times across multiple ground stations compared to several other schemes. In multi-path ground station scenarios, some of the comparison schemes experienced significantly increased completion times due to their inability to simultaneously address path switching, local queue backlog, and inter-flow fairness. Our method, however, achieves faster download completion by coordinating end-side window adjustment with intermediate node interest-based traffic splitting and rate shaping. This result corresponds to the collaborative relationship between end-side congestion window adjustment and network-side control commands.

[0067] Furthermore, in one embodiment provided in this application, the step of generating an end-side congestion window adjustment instruction based on the congestion score information and the congestion marking result in the marked data packet includes: when the congestion score information is determined by a second congestion score and the second congestion score is in the severe congestion range, performing a double-rate reduction of the congestion window according to the second congestion score; when the congestion score information is determined by the second congestion score and the second congestion score is in the mild congestion range, performing a linear reduction of the congestion window; when the congestion score information is determined by a first congestion score and the first congestion score is in the early warning congestion range, performing a conservative increase of the congestion window; and when the congestion score information is determined by the first congestion score and the first congestion score is in the no-congestion range, performing a double-rate increase of the congestion window.

[0068] Specifically, the end node records the current congestion window as When the congestion score information is determined by the second congestion score, and When falling into a severely congested region, the end node performs a doubling reduction:

[0069] in, This is the window reduction factor. The higher the value, the greater the window shrinkage, causing end nodes to quickly reduce the amount of subsequent interest packets injected. When the congestion score information is determined by the second congestion score, and When falling into a lightly congested region, the end node performs linear reduction:

[0070] This processing is used to handle short-term traffic exceeding transmission capacity or slight link fluctuations, avoiding excessive reduction of the transmission window due to mild congestion. When congestion score information is determined by the first congestion score, and When falling into the congestion warning range, the endpoint performs a conservative increase:

[0071] This process is used to maintain a slow window growth when there is potential risk but no actual congestion has yet occurred. When the congestion score information is determined by the first congestion score, and When the endpoint falls into a non-congested region, it executes a doubling of its increment rate:

[0072] in, This is the window growth factor, used to improve the utilization of available bandwidth when the link is unobstructed. If the adjusted window is smaller than the preset minimum window, the end node will restore it to the minimum window; if the adjusted window is larger than the preset maximum window, the end node will limit it to the maximum window to avoid uncontrolled transmission of interest packets due to abnormal expansion or contraction of the window on the end side.

[0073] In one embodiment, the congestion trigger threshold is set to 0.5. When the first congestion score is greater than the congestion trigger threshold, the intermediate node generates a second congestion score and determines a mild congestion interval or a severe congestion interval based on the second congestion score. The severe congestion interval is the interval where the second congestion score is greater than 0.8 and less than or equal to 1, and the mild congestion interval is the interval where the second congestion score is greater than 0.5 and less than or equal to 0.8. When the first congestion score is not greater than the congestion trigger threshold, the intermediate node determines a warning congestion interval or a no-congestion interval based on the first congestion score. The warning congestion interval is the interval where the first congestion score is greater than 0.3 and less than or equal to 0.5, and the no-congestion interval is the interval where the first congestion score is greater than or equal to 0 and less than or equal to 0.3. The above congestion trigger threshold and each score interval can be adjusted according to the link type, node buffer capacity, and network operating status.

[0074] Through the above processing Figures 6 to 8 Corresponding single-consumer single-path latency variation scenarios, Figures 10 to 11 Corresponding multi-consumer single-path fair competition scenarios, Figures 13 to 14 Corresponding multi-consumer, multi-path traffic splitting scenarios, and Figures 16 to 17 The corresponding real satellite topology download scenarios can all be explained by the same end-to-end network collaborative control chain: intermediate nodes identify the source of congestion through multi-index scoring and flow-level marking, the network side performs interest-based traffic splitting or rate shaping according to the congestion level, and end nodes adjust the subsequent interest packet sending window according to the congestion information carried by the return data packets, thereby reducing the situation of misjudged congestion, invalid rate limiting, and excessive occupation of link resources by a single interest flow.

[0075] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0076] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A ubiquitous network congestion control method based on end-to-end network collaboration, applied to ubiquitous networks containing end nodes, intermediate nodes, and content production nodes, characterized in that: The method includes: S1. End nodes generate and send interest packets, intermediate nodes forward the interest packets, and when the data packets corresponding to the interest packets return, they obtain network status indicators corresponding to the interest flow. The network status indicators include queue usage status, pending interest table occupancy status, available bandwidth status, and interest packet satisfaction status. S2. The intermediate node generates a first congestion score based on the queue usage status and the pending interest table occupancy status, and generates a second congestion score based on the network status index when the first congestion score meets the congestion triggering condition. S3. The intermediate node determines the congestion score information based on the first congestion score or the second congestion score, and generates a congestion marking probability based on the queue overload status and the occupancy contribution of the target interest flow in the unprocessed interest table. The return data packet is marked for congestion according to the congestion marking probability to obtain the marked data packet. S4. The intermediate node generates network-side control instructions based on the congestion level to which the congestion score information belongs. When the congestion level is a severe congestion level, it generates interest-based traffic splitting control instructions. When the congestion level is a mild congestion level, it generates rate-shaping control instructions. The node then adjusts the forwarding path or forwarding rate of subsequent interest packets according to the network-side control instructions. S5. The end node receives the marked data packet, generates an end-side congestion window adjustment instruction based on the congestion score information and the congestion marking result in the marked data packet, and adjusts the sending window of subsequent interest packets according to the end-side congestion window adjustment instruction.

2. The ubiquitous network congestion control method with end-to-end network collaboration according to claim 1, characterized in that: The interest package includes a content name prefix field, a random number field, a start time field, and a traffic split ratio field; The labeled data packet includes a content name prefix field, a random number field, a data field, a cache hit field, a data generation time field, a congestion flag field, and a congestion score field; the congestion flag field is used to carry the congestion labeling result, and the congestion score field is used to carry the congestion score information.

3. The ubiquitous network congestion control method with end-to-end network collaboration according to claim 1, characterized in that, The network status indicators are obtained as follows: The ratio of the occupied queue length of the intermediate node to the maximum queue capacity is determined as the queue utilization rate; The ratio of the number of currently pending interest table entries in the intermediate node to the maximum capacity of the pending interest table is determined as the pending interest table occupancy rate. The available bandwidth is determined based on the maximum bandwidth of the link, the bandwidth occupied by the link in the previous statistical period, the number of data bytes in the current statistical period, and the statistical duration. The ratio of the number of received data packets to the number of interest packets sent is defined as the interest packet satisfaction rate.

4. The ubiquitous network congestion control method with end-to-end network collaboration according to claim 3, characterized in that, The step of generating a first congestion score based on the queue usage status and the pending interest table occupancy status, and generating a second congestion score based on the network status indicators when the first congestion score meets the congestion triggering conditions, includes: The first congestion score is generated based on a weighted average of queue utilization and pending interest table occupancy. When the first congestion score is greater than the congestion trigger threshold, a second congestion score is generated based on the weighted result of queue utilization, pending interest table occupancy, available bandwidth, and interest packet satisfaction rate. When the first congestion score is not greater than the congestion trigger threshold, the first congestion score is used as congestion score information to indicate a warning state or a no-congestion state.

5. The ubiquitous network congestion control method with end-to-end network collaboration according to claim 2, characterized in that, When the labeled data packet passes through multiple intermediate nodes along the return path, each intermediate node compares the congestion score generated by its own node with the existing congestion score information in the labeled data packet. If the congestion score generated by its own node is higher than the existing congestion score information, it updates the congestion score information in the congestion score field so that the congestion score information received by the end node indicates a higher degree of congestion on the return path.

6. The ubiquitous network congestion control method with end-to-end network collaboration according to claim 1, characterized in that, The step of generating congestion label probabilities based on queue overload status and the occupancy contribution of the target interest stream in the pending interest table includes: The fair share of the queue for the target interest stream is determined based on the current queue length and the number of interest streams. When the queue occupancy of the target interest stream does not exceed the fair share of the queue, no congestion mark is applied to the return data packets corresponding to the target interest stream; When the queue occupancy of the target interest flow exceeds the fair share of the queue and the current queue length is greater than the queue safety threshold, the basic congestion probability is determined based on the current queue length, the queue safety threshold, and the maximum queue length. The congestion contribution weight is determined based on the ratio of the number of entries of the target interest flow in the pending interest table to the total number of entries of all interest flows in the pending interest table. The product of the basic congestion probability and the congestion contribution weight is determined as the congestion labeling probability of the target interest flow.

7. The ubiquitous network congestion control method with end-to-end network collaboration according to claim 1, characterized in that, The generation methods of the interest-based traffic splitting control commands include: The queue utilization rate, pending interest table occupancy rate, available bandwidth, and interest packet satisfaction rate of the candidate forwarding links are obtained respectively, and a second congestion score is generated for each candidate forwarding link based on the queue utilization rate, the pending interest table occupancy rate, the available bandwidth, and the interest packet satisfaction rate. The candidate forwarding link with the lowest second congestion score is identified as the target backup link; Within the return time window, count the number of congestion marks for each interest flow and the total number of congestion marks for all interest flows to determine the congestion impact of each interest flow. The migration quantity is determined based on the congestion impact, the transfer ratio, and the number of entries of the corresponding interest flow in the pending interest table, and the subsequent interest packets corresponding to the migration quantity are redirected to the target backup link.

8. The ubiquitous network congestion control method with end-to-end network collaboration according to claim 2, characterized in that, The generation method of the rate shaping control command includes: The dynamic queue congestion threshold is determined based on the queuing target time, available outgoing link bandwidth, and average packet size. The shaping transmission rate of the current interest stream is determined based on the traffic segmentation ratio carried in the traffic segmentation ratio field, the dynamic queue congestion threshold, the buffer queue length of the current interest stream, the round-trip time, the local cache hit rate and the global cache hit rate of the current interest stream. The intermediate node updates the forwarding rate of subsequent interest packets for the current interest stream according to the shaped transmission rate.

9. The ubiquitous network congestion control method with end-to-end network coordination according to claim 2, characterized in that: When an intermediate node detects that a link interface is disconnected and there are data packets that cannot be forwarded, it caches the data packets that cannot be forwarded in the content cache. Based on the cached data packets after removing the data fields, it adds interface information, cache freshness information, and announcement flags, generates a cache announcement packet, and announces it to neighboring nodes, so that neighboring nodes can obtain the corresponding cached data based on the content name prefix and interface information.

10. The ubiquitous network congestion control method with end-to-end network collaboration according to claim 1, characterized in that, The step of generating an end-side congestion window adjustment instruction based on the congestion score information and the congestion marking results in the marked data packet includes: When the congestion score information is determined by the second congestion score and the second congestion score is in the severe congestion range, the congestion window is reduced at double speed according to the second congestion score; When the congestion score information is determined by the second congestion score and the second congestion score is in the mild congestion range, the congestion window is linearly reduced. When the congestion score information is determined by the first congestion score and the first congestion score is within the warning congestion interval, the congestion window is conservatively increased. When the congestion score information is determined by the first congestion score and the first congestion score is in a non-congestion interval, the congestion window is increased at double speed.

Images