A hybrid congestion control method and system based on RTT and INT for data center networks

Abstract

This invention relates to a hybrid congestion control method and system for data center networks based on RTT and INT, within the field of computer network transmission control. The invention aims to achieve high-precision network congestion control with low data center network load costs. When facing different network environments, the host and network nodes can promptly perceive the network status and select different strategies accordingly, thereby adjusting the sender's rate to an ideal state. This invention employs a congestion control scheme based on a hybrid of time-delay signals and network telemetry signals. By analyzing network information and determining the network state, time-delay-based control ensures that the algorithm can achieve overall control of the network under general conditions with low network overhead, while network telemetry-based control ensures precise control of complex network conditions. The flexible switching between the two improves the overall performance of data center network data transmission.

Description

Technical Field

[0001] This invention relates to the field of computer network transmission control, and in particular to a hybrid congestion control algorithm based on RTT and INT for data center networks. Background Technology

[0002] Data center networks are the critical infrastructure platforms for providing online services and applications on the internet today. A large number of distributed applications are deployed in these data center networks, ranging from everyday applications such as online search, social shopping, and video games to the operation of large language models, large-scale machine learning systems, high-performance computing clusters, and distributed file systems—all relying on data center networks for data storage and transmission. However, the rapid development of software and hardware technologies in recent years, along with the rapid changes in the application demands and operating modes of data center networks, has brought enormous challenges to data center network congestion control technology.

[0003] Common algorithms for addressing data center network congestion include congestion control based on Explicit Congestion Notification (ECN), congestion control based on Round-Trip Time (RTT), and congestion control based on network telemetry (INT). These algorithms use different congestion feedback signals to regulate network congestion; however, each has its own set of problems.

[0004] Congestion control algorithms based on ECN signals suffer from the complexity of setting the labeling threshold. If the threshold is too small, the algorithm overreacts to network state changes, leading to insufficient bandwidth utilization. Conversely, if the threshold is too high, the algorithm can cause frequent congestion and continuous packet loss. Furthermore, the algorithm struggles with bursty traffic and mixed large and small traffic flows. Effectively utilizing idle bandwidth without impacting primary traffic transmission remains a significant challenge when handling specific traffic patterns such as high bursts. Therefore, ECN is not chosen as the congestion signal; instead, RTT and INT are used in conjunction for congestion detection and control.

[0005] Congestion control algorithms based on latency information can ensure a generally accurate assessment of network congestion at a relatively low cost. However, they suffer from issues such as precise measurement, clock synchronization, and noise interference. The basic latency of a data center is typically a few microseconds or tens of microseconds. Accurately measuring queuing latency requires high-precision clock synchronization. Furthermore, latency in some host-side hardware processing can easily introduce noise, making the measurement and response to RTT (Round-Trip Time) insufficiently sensitive under varying network conditions, and unable to adjust the transmission rate promptly and accurately. Therefore, while RTT is a low-cost and simple-to-design congestion signal, it is difficult to make fine-grained assessments of the network, and its performance largely depends on accurate measurement of network and host latency. Therefore, when RTT cannot be accurately adjusted, more refined congestion signals are needed for network condition diagnosis.

[0006] Congestion control algorithms based on INT (Intense Threat Transmission) use real-time congestion signals actively inserted by switches, resulting in accurate and comprehensive information. Compared to ECN (Electronic Network Communication) and latency-based methods, they achieve more accurate congestion awareness. However, they suffer from higher costs. Existing INT-based schemes, regardless of network conditions, add an INT header to data packets and request switches to insert congestion information, which undoubtedly consumes bandwidth intended for normal data transmission, placing a burden on the network. Therefore, this approach combines the advantages of both RTT (Real-Time To Transmission) and INT (Intense Threat Transmission) congestion signals to compensate for their respective disadvantages. RTT signals are used to ensure normal network transmission, while INT signals are selectively used to detect network anomalies, thus achieving flexible adaptation to complex network conditions at a lower cost. Summary of the Invention

[0007] The technical problem to be solved by this invention is:

[0008] This invention proposes a hybrid congestion control method and system based on RTT and INT for data center networks. It aims to achieve high-precision network congestion control with low data center network load costs. When facing different network environments, the host and network nodes can promptly perceive the network status and select different strategies accordingly, thereby adjusting the sender's rate to an ideal state. While ensuring high network throughput, it minimizes network latency and flow completion time as much as possible.

[0009] The technical solution adopted by the present invention to solve the above-mentioned technical problems is as follows:

[0010] A hybrid congestion control method based on RTT and INT for data center networks is proposed, the implementation process of which is as follows:

[0011] A congestion switching mechanism (congestion control algorithm) based on network operating status is constructed. First, the network status is identified, then the network congestion situation is assessed, and real-time congestion switching is implemented. Specifically:

[0012] Real-time traffic information is collected at the host end. By statistically analyzing the traffic information, the network congestion status is detected and evaluated. When the network is not congested or is slightly congested (under normal circumstances), RTT is used as the congestion detection signal to control the data transmission process. When the severity or duration of congestion exceeds the threshold, the INT signal is selected to notify the switch to participate in the congestion perception and adjustment. After adjustment for a period of time until the network status is observed to return to normal, the use of the INT control signal is stopped and the original congestion control based on the time delay signal RTT is switched back.

[0013] Congestion detection and rate control are performed using a hybrid congestion control mechanism combining RTT and INT.

[0014] In congestion control based on time delay signals, the sender receives the acknowledgment data packet from the receiver and calculates the time delay information in it. Then, based on the comparison between the current measurement value and the set threshold, it chooses to additively increase the network rate to improve the network bandwidth utilization when the network has low throughput. When the network is congested, it calculates the ratio of the basic round-trip time to the product of the measurement time and the gradient to dynamically adjust the sending rate of the host.

[0015] In congestion control based on network telemetry signals, the sender selects to enable INT flags to notify the switch to detect network conditions. After detecting the flags, the switch judges the network congestion status and inserts congestion information into the data packets. Subsequently, after receiving the control data packets sent from the receiver, the sender parses the network congestion information and calculates the number of bytes in flight using the bottleneck queue length in the INT information based on the bottleneck link model. It also calculates the gradient of the queue length using historical bottleneck queue information. The ratio of the bandwidth-delay product to the number of bytes in flight is used as a parameter for overall network status perception, and the ratio of the transmission rate to the queue gradient is used as a parameter for network change perception. The former ensures the algorithm's grasp of the overall network, while the latter improves the algorithm's sensitivity to network change trends. Finally, the hop count information of the traffic is combined to weight the two parameters, thereby balancing the throughput of long-distance traffic and the latency of short-distance traffic, and improving the algorithm's ability to handle high-speed bursts and complex mixed traffic patterns.

[0016] The present invention has the following beneficial technical effects:

[0017] The hybrid congestion control algorithm for data center networks based on latency and network telemetry provided by this invention achieves high-precision network congestion control with low data center network load cost. When facing different network environments, the host and network nodes can promptly perceive the network status and select different strategies accordingly, thereby adjusting the sender's rate to an ideal state. While ensuring high network throughput, it minimizes network latency and flow completion time as much as possible.

[0018] This invention employs a congestion control scheme based on a hybrid approach of time-delay signals and network telemetry signals. It analyzes network information and determines the network state. Time-delay-based control ensures the algorithm can achieve overall control of the network under normal conditions with low network overhead, while network telemetry-based control ensures precise control of complex network conditions. The flexible switching between these two approaches improves the overall performance of data center network data transmission. Attached Figure Description

[0019] Figure 1 Hybrid congestion control framework diagram;

[0020] Figure 2 Three-dimensional congestion state transition diagram;

[0021] Figure 3 A schematic diagram of a time-delay-based congestion control mechanism;

[0022] Figure 4 A schematic diagram of a congestion control mechanism based on network telemetry. Detailed Implementation

[0023] This invention presents a hybrid congestion control algorithm for data center networks based on latency and network telemetry, requiring collaboration among the sending host, switch, and receiving host. The sending host is responsible for collecting traffic information and analyzing network operation status to switch congestion control mechanisms. It further employs a congestion control algorithm based on RTT and INT signals to adjust the sending rate, alleviating network congestion. The receiving host is responsible for transmitting received network information back to the sending host. The switch is responsible for accurately detecting network operation status and transmitting congestion information to the host, thereby assisting the host in achieving precise network control. The overall solution comprises several parts: traffic information collection, network status assessment, congestion mechanism switching, latency signal control, and network telemetry detection, as detailed below.

[0024] 1. Traffic information collection: When the sender receives a data packet, it classifies the data packet into three categories, namely normal data packets, response data packets, and control data packets according to the tags carried by the data packet, and processes them separately. For response data packets, analyze and record the timestamp information in them. For control data packets, count and analyze the routing queue, queuing delay, and other information in them. For normal data packets, no additional processing is performed.

[0025] 2. Network condition assessment: After collecting information, assess the network condition. Specifically, divide the operating condition of the network into three states: no congestion, mild congestion, and severe congestion. In the no-congestion state, the network throughput is less than or equal to the network load, and the queue length in the switch is zero. Only need to simply increase the sending rate. In the mild-congestion state, the network throughput is slightly greater than the network load, and there is packet accumulation in the switch but it never exceeds the queue threshold. Just moderately adjust the sending rate. In the severe-congestion state, the network throughput is much greater than the network load, and new arriving data packets cannot be accepted continuously in the switch buffer. At this time, it is necessary to accurately measure the network state to achieve more precise control. The sender analyzes the current and historical traffic information collected in step 1, and drives the network to switch between these three states. The switching of the three states is considered from two aspects: the severity and duration of congestion. The system will preset a low delay threshold t_min and a high delay threshold t_max. After the sender receives a data packet, it calculates its round-trip delay information t_current, and then compares it with the two set thresholds. The low threshold evaluates whether the network is congested, while the high threshold evaluates the severity of network congestion. In addition, set a counter S0 to accumulate in the congested state and clear the counter when the network returns to normal, so as to count the duration of network congestion. Then it will also be compared with the preset threshold S1. The network defaults to start from the no-congestion state. If t_current < t_min, it will continuously be in the no-congestion state. If t_min ≤ t_current ≤ t_max and S0 ≤ S1, it will be converted to the mild-congestion state. If t_current > t_max or S0 > S1, it will be converted to the severe-congestion state. When in the severe-congestion state, judge whether the collected routing queue length qlen is all 0. If not, continue to be in the current congested state until the switch empties the queue to 0 before it can be converted to the no-congestion state.

[0026] 3. Congestion Mechanism Switching: The congestion control mechanism is switched based on network assessment results. In the absence of congestion or with mild congestion, round-trip time (RTD) information is used to adjust the sending rate of network hosts. In this case, the sender disables the INT flag when sending data packets, thus informing the switch to stop using network telemetry. Upon receiving the corresponding data packets, the switch checks the flag and stops inserting INT information, thereby reducing network load costs. When severe or persistent congestion occurs, the sender enables the INT flag when sending data packets, thus activating network telemetry. Switches along the route receive the corresponding data packets, measure the network status, and insert the obtained information into the data packets, thereby achieving more accurate perception and control of the network condition.

[0027] 4. Time Delay Signal Control: This method uses time delay signals to control the transmission rate. For typical network conditions, congestion control based on RTT signals is used, consisting of two parts: congestion detection and rate control. The sender can calculate the round-trip time (RTT) of the current data transmission using the difference between the timestamp carried in the received ACK packet and its arrival time: RTT = t current -ack timestamp For the obtained RTT measurement value, after obtaining the difference based on the historical measurement value, its exponentially weighted average is transformed into a smoothed difference: RTT diff = (1-α)×rtt diff +α×(rtt current -rtt last Then, the gradient value of the time delay, rtt, is calculated. gradient =rtt diff / rtt_min. During congestion detection, the measured RTT value is compared with a pre-set threshold. If it is lower than the threshold, the sending rate is additively increased: rate = rate current +rate ai If the value is higher than the threshold, then the congestion level factor is calculated. Then, the transmission rate is dynamically adjusted based on the congestion factor: rate = α × (f × rate) current +β)+(1-α)×rate current .

[0028] 5. Network Telemetry Probing: For situations with severe or prolonged network congestion, in-band network telemetry is used for more accurate network condition detection. This telemetry probing is implemented jointly by the sending end, switches supporting INT functionality, and the receiving end. When the sending end sends a data packet, it adds an INT header field to the packet header. Then, during the transmission of the data packet from the sender to the receiver, each switch on the data link adds metadata information based on the INT flag field carried in the data packet, recording the switch's current status, including the queue length (qlen), transmission bytes (txBytes), and link bandwidth capacity (B) of the egress port. When the receiver receives a data packet, it copies all the INT information recorded by the switches into the control data packet sent back to the sender. The sender can then use this information to more accurately assess the network congestion situation and adjust the window more precisely. The sending end is responsible for initiation and control. When sending data, it adds an INT header field to the data packet header to initiate congestion detection and implement selective congestion switching mechanisms. When receiving data, it needs to parse the corresponding INT header to obtain the relevant information, and then calculate the congestion level and adjust the transmission rate based on this information in the data packet. The switch is responsible for network congestion detection. Upon receiving a data packet, it determines whether to perform congestion detection based on whether the packet carries an INT flag. After receiving the INT flag, the switch collects network information and monitors the queue length. Then, the switch writes the current network information, including the queue length, into the data packet and forwards it to the next switch. All switches along the path operate in this manner until the data packet reaches the receiving end. The receiving end is responsible for information feedback. After parsing the INT information, it copies the INT information into the generated data packet and sends it to the sender.

[0029] Network telemetry detection needs to handle situations with severe or prolonged congestion. After parsing the state of network nodes, the sender uses this state as a basis to more accurately assess the degree of network congestion and make precise rate control from the perspectives of network state change trends and traffic characteristics. By combining the current and historical network states, the gradient information of the switch queues can be calculated, thereby enhancing the host's perception of network change trends. Combining location information allows for targeted balancing of traffic characteristics and demands at different distances, thus improving the algorithm's dynamic adaptation to sudden traffic changes and mixed traffic of long and short distances. Specifically, the sender iterates through the queue lengths of switches along the path, selects the maximum value as the bottleneck queue length, and then calculates the number of network flying bytes based on the transmission rate and basic latency, using this as a basic parameter to measure network conditions: Inflight = qlen + T × txRate. Then, the difference between the current bottleneck queue and the previous bottleneck queue is calculated: qlen. gra =(α×(qlen) curr -qlenlast ) / time_diff+(1-α)×qlen gra The ratio of two measured values ​​to their target value can be used as a measure of congestion level. The former, the ratio of bandwidth-delay product to the number of bytes flowed, allows for control over the overall network status, while the latter, the ratio of transmission rate to queue change rate, allows for perception of network change trends. Then, hop count information is used as a coefficient to perform a weighted average of the two to obtain the adjustment parameters. This achieves a balance between traffic at different distances, ensuring that traffic at closer distances responds more drastically to network changes, achieving low latency for short-distance traffic without compromising throughput for long-distance traffic. Finally, this parameter guides a more accurate update of the transmission rate; the adjusted rate is determined by both the current transmission rate and the adjustment parameter u. Rate update formula: Rate = γ × (Rate...) current ×u+β)+(1-γ)×Rate current .

[0030] Verification has shown that the method proposed in this invention solves the technical problem identified in this invention. Practical application has verified the technical effects and practicality claimed by this invention. Simulation experiments and practical applications have both verified the technical effects claimed by this invention: by comprehensively utilizing the advantages of both RTT and INT congestion signals to compensate for their respective disadvantages, using the RTT signal to ensure normal network transmission, and selectively enabling the INT signal to detect network anomalies, this method achieves flexible adaptation to complex network conditions at a relatively low cost.

[0031] The algorithm (method) proposed in this invention is the underlying technical core of this invention, and various products can be derived based on the algorithm.

[0032] Based on the algorithm (method) proposed in this invention, a hybrid congestion control system based on RTT and INT for data center networks is developed using a programming language. The system has program modules corresponding to the steps of the above technical solution, and executes the steps in the above-mentioned hybrid congestion control method based on RTT and INT for data center networks when running.

[0033] The developed system (software) computer program is stored on a computer-readable storage medium. This computer program is configured to implement the steps of the aforementioned hybrid congestion control method for data center networks based on RTT and INT when invoked by a processor. In other words, the invention is materialized on a carrier, becoming a computer program product.

[0034] A hybrid congestion control device based on RTT and INT for data center networks is disclosed. The device includes at least one processor and a memory communicatively connected to the at least one processor. The memory stores instructions executable by the at least one processor, which enable the at least one processor to perform the aforementioned hybrid congestion control method based on RTT and INT for data center networks. By selecting RTT and INT to assist each other in congestion detection and control, the device achieves this.

[0035] Various implementations of the systems and techniques described herein can be implemented in digital electronic circuit systems, integrated circuit systems, application-specific integrated circuits (ASICs), computer hardware, firmware, software, and / or combinations thereof. These various implementations may include: implementations in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which may be a dedicated or general-purpose programmable processor, capable of receiving data and instructions from a storage system, at least one input device, and at least one output device, and transferring data and instructions to the storage system, the at least one input device, and the at least one output device.

[0036] The computational programs (also referred to as programs, software, software applications, or code) of this invention include machine instructions of a programmable processor and can be implemented using high-level procedural and / or object-oriented programming languages, and / or assembly / machine languages. As used herein, the terms "machine-readable medium" and "computer-readable medium" refer to any computer program product, device, and / or apparatus (e.g., disk, optical disk, memory, programmable logic device PLD) for providing machine instructions and / or data to a programmable processor, including machine-readable media that receive machine instructions as machine-readable signals. The term "machine-readable signal" refers to any signal for providing machine instructions and / or data to a programmable processor.

[0037] It should be understood that the various processes shown above can be used to rearrange, add, or delete steps. For example, the steps described in this application can be executed in parallel, sequentially, or in different orders, as long as the desired result of the technical solution disclosed in this application can be achieved, they are all within the protection scope of this invention.

Claims

1. A hybrid congestion control method based on RTT and INT for data center networks, characterized in that, The implementation process of the method is as follows: To construct a congestion switching mechanism based on network operating status, the network status is first identified, then the network congestion situation is assessed, and real-time congestion switching is implemented. Specifically: Real-time traffic information is collected at the host end. By statistically analyzing the traffic information, the network congestion status is detected and evaluated. When the network is not congested or slightly congested, RTT is used as a congestion detection signal to control the data transmission process. When the severity or duration of congestion exceeds the threshold, the INT signal is selected to notify the switch to participate in the congestion perception and adjustment. After adjustment for a period of time until the network status is observed to return to normal, the use of the INT control signal is stopped and the original congestion control based on the time delay signal RTT is switched back. Congestion detection and rate control are performed using a hybrid congestion control mechanism combining RTT and INT. In congestion control based on time delay signals, the sender receives the acknowledgment data packet from the receiver and calculates the time delay information in it. Then, based on the comparison between the current measurement value and the set threshold, it chooses to additively increase the network rate to improve the network bandwidth utilization when the network has low throughput. When the network is congested, it calculates the ratio of the basic round-trip time to the product of the measurement time and the gradient to dynamically adjust the sending rate of the host. In congestion control based on network telemetry signals, the sender selects to enable INT flags to notify the switch to detect network conditions. After detecting the flags, the switch judges the network congestion status and inserts congestion information into the data packets. Subsequently, after receiving the control data packets sent from the receiver, the sender parses the network congestion information and calculates the number of flying bytes based on the bottleneck link model using the bottleneck queue length in the INT information. It also calculates the gradient of the queue length using historical bottleneck queue information. The ratio of the bandwidth-delay product to the number of flying bytes is used as a parameter for overall network status perception to ensure the algorithm's grasp of the overall network. The ratio of the transmission rate to the queue gradient is used as a parameter for network change perception to improve the algorithm's sensitivity to network change trends. Finally, the parameters for overall network status perception and network change perception are weighted by combining the hop count information of the traffic to balance the throughput of long-distance traffic and the latency of short-distance traffic, thereby improving the algorithm's ability to handle high-speed bursts and complex mixed traffic patterns.

2. The hybrid congestion control method based on RTT and INT for data center networks according to claim 1, characterized in that, The hybrid congestion control method is based on the collaborative work of the sending host, the switch, and the receiving host. The sending host is responsible for collecting traffic information and analyzing the network operating status to switch congestion control mechanisms. Furthermore, it uses a congestion control algorithm based on RTT and INT signals to adjust the sending rate of the sending host and alleviate network congestion. The receiving host is responsible for sending the received network information back to the sending host. The switch is responsible for accurately detecting the network's operating status and transmitting congestion information to the host, thereby assisting the host in achieving precise control over the network.

3. A hybrid congestion control method based on RTT and INT for data center networks according to claim 2, characterized in that, In constructing a congestion switching mechanism based on network operating status, the process of traffic information collection is as follows: when the sender receives a data packet, it will classify the data packet into three categories according to the tags carried by the data packet, namely normal data packets, response data packets, and control data packets, and process them separately; for response data packets, the timestamp information is analyzed and recorded; for control data packets, the routing queue and queuing delay information is statistically analyzed; and no additional processing is performed on normal data packets.

4. A hybrid congestion control method based on RTT and INT for data center networks according to claim 1, 2, or 3, characterized in that, In constructing a congestion switching mechanism based on network operating status, the network condition is assessed after information is collected. The network condition assessment process is as follows: Specifically, network operation is categorized into three states: no congestion, mild congestion, and severe congestion. In the no-congestion state, network throughput is less than or equal to network load, and the queue length in the switch is zero; simply increasing the sending rate is sufficient. In the mild congestion state, network throughput is slightly greater than network load, and there is a backlog of data packets in the switch, but it never exceeds the queue threshold; adjusting the sending rate appropriately is sufficient. In the severe congestion state, network throughput is much greater than network load, and the switch buffer cannot continue to accept newly arriving data packets. At this point, accurate measurement of the network state is required to achieve more precise control. The sender analyzes the collected current and historical traffic information to drive the network to switch between these three states. The transition between the three states is considered from the perspectives of the severity and duration of congestion. The system will pre-set a low latency threshold t_min and a high latency threshold t_max. After receiving a data packet, the sender calculates its round-trip latency information t_current, and then compares t_current with the two set thresholds. The low threshold assesses whether the network is congested, while the high threshold assesses the severity of the network congestion. A counter S0 is incremented during congestion and reset to zero when the network recovers, thus tracking the duration of network congestion. The counter S0 is then compared to a pre-defined threshold S1. The network starts from a congestion-free state by default. If t_current < t_min, it remains in a congestion-free state. If t_min ≤ t_current ≤ t_max and S0 ≤ S1, it transitions to a lightly congested state. If t_current > t_max or S0 > S1, it transitions to a heavily congested state. During a heavily congested state, the length of the collected route queue qlen is checked to see if it is all zero. If not, the network remains in the current congested state until the switch empties the queue to zero, at which point it transitions to a congestion-free state.

5. A hybrid congestion control method based on RTT and INT for data center networks according to claim 4, characterized in that, In constructing a congestion switching mechanism based on network operating status, the congestion switching process is as follows: The congestion control mechanism is switched based on network assessment results. During periods of no congestion or mild congestion, round-trip time (RTD) information is used to adjust the transmission rate of network hosts. In this case, the sender marks INT (Intent to Telemetry) as disabled when sending data packets, thus informing the switch to stop using network telemetry. Upon receiving the corresponding data packet, the switch checks the mark and stops inserting INT information, thereby reducing network load costs. When severe or persistent network congestion occurs, the sender marks INT as enabled when sending data packets, thus activating network telemetry. Switches along the route receive the corresponding data packet, measure the network status, and insert the obtained information into the data packet, thereby achieving more accurate perception and control of the network condition.

6. A hybrid congestion control method based on RTT and INT for data center networks according to claim 5, characterized in that, In the process of using a hybrid congestion control mechanism combining RTT and INT for congestion detection and rate control, The transmission rate is controlled by using time delay signals. For general network conditions, congestion control based on RTT signals is used, which is divided into two parts: congestion detection and rate control. The sender can calculate the round-trip time (RTT) of this data transmission by using the difference between the timestamp carried in the received ACK packet and its arrival time. For the obtained RTT measurements, after obtaining the difference based on historical measurements, an exponentially weighted average is used to transform it into a smoothed difference: Then, the gradient value of the time delay is calculated. During congestion detection, the measured RTT value is compared with a pre-set threshold. If it is lower than the threshold, the transmission rate is increased additively. If the value is higher than the threshold, then the congestion level factor is calculated. Then, the transmission rate is dynamically adjusted based on the congestion factor: .

7. A hybrid congestion control method based on RTT and INT for data center networks according to claim 6, characterized in that, In the congestion detection and rate control using a hybrid congestion control mechanism of RTT and INT, for cases where network congestion is severe or lasts for a long time, in-band network telemetry is used to detect the network condition more accurately. The network telemetry detection part is jointly implemented by the sending end, the switch that supports INT function, and the receiving end. When the sending end sends a data packet, it adds an INT header field to the message header; Then, during the transmission of data packets from the sender to the receiver, each switch on the data link adds metadata information to record the current status of the switch based on the INT flag field carried by the data packet, including the queue length qlen of the egress port, the number of transmitted bytes txBytes, and the link bandwidth capacity B. When the receiver receives the data packet, it copies all the INT information recorded by the switch into the control data packet sent back to the sender. Afterward, the sender can use this information to make a more accurate assessment of the network congestion and thus make more accurate adjustments to the window. The sending end is responsible for initiation and control. When sending data, it adds an INT header field to the packet header to initiate congestion detection and selectively switch congestion mechanisms. When receiving data, it needs to parse the corresponding INT header to obtain the relevant information, and then calculate the congestion level and adjust the sending rate based on this information in the packet. The switch is responsible for network congestion awareness. When it receives a packet, it selects whether to perform congestion detection based on whether the packet carries an INT flag. After receiving the INT flag, the switch will collect network information and monitor the queue length. Then, the switch needs to write the current network information, including the queue length, into the packet and finally forward the packet to the next switch. All switches along the path operate in this way until the packet reaches the receiving end. The receiving end is responsible for information feedback. After parsing the INT information, it is responsible for copying the INT information into the generated packet and sending it to the sender. Network telemetry detection needs to handle situations with severe or prolonged congestion. After the sender parses the state of the network nodes, it uses the network state as a basis to more accurately assess the degree of network congestion and make accurate rate control from the perspectives of network state change trends and traffic characteristics. By combining the current network state with historical network states, the gradient information of the switch queue can be calculated, thereby enhancing the host's perception of network change trends. Combined with location information, the characteristics and needs of traffic at different distances can be balanced in a targeted manner, thereby improving the algorithm's dynamic adaptation to sudden changes in traffic and mixed traffic of long and short distances.

8. A hybrid congestion control method based on RTT and INT for data center networks according to claim 7, characterized in that, The specific process of the rate control is as follows: Specifically, the sender iterates through the queue lengths of the switches along the route, selects the maximum value as the bottleneck queue length, and then calculates the number of network flying bytes based on the transmission rate and basic latency, using this as a basic parameter to measure network conditions. Then calculate the difference between the current bottleneck queue and the previous bottleneck queue: The ratio of two measured values ​​to their target value can be used as a measure of congestion level. , The ratio of bandwidth-delay product to the number of bytes flown is used to monitor the overall network status. To perceive network change trends, the ratio of transmission rate to queue change rate is used. Then, hop count information is used as a coefficient to perform a weighted average of the two to obtain the adjustment parameter. This achieves a balance between traffic at different distances, ensuring that traffic at closer distances responds more drastically to network changes, thus achieving low latency for short-distance traffic without compromising throughput for long-distance traffic; finally, parameters are used. The transmission rate is updated more accurately, and the adjusted rate is determined by both the current transmission rate and the adjustment parameter u; the rate update formula is: .

9. A hybrid congestion control system based on RTT and INT for data center networks, characterized in that: The system has a program module corresponding to the steps of any one of the claims 1-8 above, and executes the steps in the hybrid congestion control method based on RTT and INT for data center networks when it runs.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program configured to, when invoked by a processor, implement the steps of the hybrid congestion control method based on RTT and INT for data center networks as described in any one of claims 1-8.

Images