RDMA network transmission method and system based on adaptive threshold and hybrid congestion control

Abstract

The present invention discloses an RDMA network transmission method and system based on adaptive threshold and hybrid congestion control, including: step 1: constructing an RDMA transmission link; step 2: adaptively adjusting the RTT maximum and minimum thresholds according to the real-time monitored link transmission data; step 3: dynamically and comprehensively adjusting the transmission rate through a multi-factor adjustment mechanism based on the adaptively adjusted RTT threshold and packet loss rate threshold; step 4: using an interval adaptive speed regulation mechanism, dynamically adjusting the transmission rate according to the position of the real-time RTT value within the threshold range, and returning to step 2. The technical solution of the present invention has higher adaptability and efficiency in network congestion control, and can better meet the high bandwidth and low latency requirements of data centers. Not only does it improve the transmission efficiency, but it also enhances the stability and robustness of the system in various complex network environments.

Description

Technical Field

[0001] The invention belongs to the field of network transmission, and in particular relates to an RDMA network transmission method and system based on adaptive threshold and hybrid congestion control. Background Art

[0002] With the rapid development of the Internet and cloud computing technologies, the demand for data centers continues to grow. As an important part of network infrastructure, the performance of data centers plays a vital role in the efficiency and stability of network transmission. Especially in high-bandwidth, low-latency application scenarios, such as real-time audio and video transmission, online games, and financial trading systems, data centers need to have extremely high transmission performance and reliability.

[0003] Traditional network congestion control algorithms, such as TCP (Transmission Control Protocol), often show great limitations when dealing with high-bandwidth, low-latency network environments. In order to meet the needs of modern data centers, researchers have proposed a variety of improved and optimized algorithms, among which the TIMELY algorithm has attracted widespread attention due to its congestion control strategy based on RTT (round-trip time).

[0004] However, the TIMELY algorithm uses the rate of change (gradient) of RTT to adjust the transmission rate. By detecting small changes in RTT, it can respond to changes in network conditions in a timely manner, thereby achieving efficient congestion control. The core mechanisms of TIMELY include the following: 1. RTT gradient detection: predict network congestion by monitoring the rate of change of RTT in real time; 2. Multiplicative adjustment: dynamically adjust the transmission rate according to the RTT gradient to maintain low latency and high throughput; 3. Fast response speed: can quickly detect and respond to small changes in RTT and adapt to rapid changes in network conditions.

[0005] However, the TIMELY algorithm also exposes some limitations in practical applications: 1. Single factor adjustment: TIMELY mainly relies on the RTT gradient for adjustment, ignoring other network performance indicators such as packet loss rate and bandwidth utilization. This makes TIMELY insufficiently responsive and has limited adjustment effects when facing complex network conditions; 2. Fixed threshold: TIMELY uses a fixed RTT threshold for adjustment and lacks dynamic adaptability. When network conditions change drastically, fixed thresholds are difficult to respond effectively, resulting in decreased transmission efficiency; 3. Simple adjustment mechanism: TIMELY's adjustment mechanism is relatively simple and lacks detailed adjustment strategies. In complex network environments, the adjustment of the transmission rate is not smooth enough, which may lead to unstable network performance. Summary of the invention

[0006] In order to overcome these limitations of TIMELY, the present invention proposes a series of improvement and optimization measures, and provides an RDMA network transmission method and system based on adaptive threshold and hybrid congestion control to improve the adaptability and transmission efficiency of the system in various network environments.

[0007] The present invention provides the following technical solutions:

[0008] On the one hand, the present invention provides an RDMA network transmission method based on adaptive threshold and hybrid congestion control, comprising:

[0009] Step 1: Build an RDMA transmission link;

[0010] Step 2: Adaptively adjust the RTT maximum and minimum thresholds based on the real-time monitored link transmission data;

[0011] Step 3: Based on the adaptively adjusted RTT threshold and packet loss rate threshold, the transmission rate is dynamically and comprehensively adjusted through a multi-factor adjustment mechanism;

[0012] When the real-time RTT is greater than the maximum RTT threshold or the packet loss rate exceeds the minimum packet loss rate threshold, the transmission rate is decelerated and the process returns to step 2; when the real-time RTT is less than the minimum RTT threshold, the transmission rate is accelerated and the process returns to step 2;

[0013] If the real-time RTT is between the RTT minimum threshold and the RTT maximum threshold, proceed to step 4;

[0014] Step 4: Adopt the interval adaptive speed regulation mechanism to dynamically adjust the transmission rate according to the position of the real-time RTT value within the threshold range, and return to step 2;

[0015] RTT refers to the round-trip delay of data transmission. By introducing an adaptive threshold adjustment mechanism, the RTT threshold can be automatically optimized according to the real-time changes in network conditions, so that bandwidth can be fully utilized when the network conditions are good and congestion can be effectively prevented when the network conditions are poor. This mechanism improves the robustness and adaptability of the system, ensuring efficient transmission performance under different network conditions.

[0016] If the real-time RTT is between the RTT minimum threshold and the RTT maximum threshold, by real-time monitoring of the position of the RTT value within the threshold range, the system can subtly and accurately adjust the transmission rate to make the rate change more gradual and smooth. When the RTT is small, the rate increases significantly, which can quickly improve the transmission efficiency; as the RTT gradually approaches the maximum threshold, the deceleration increases significantly, avoiding excessive acceleration under high load and keeping the system stable.

[0017] Furthermore, the adaptive speed regulation mechanism within the interval is used to dynamically adjust the transmission rate according to the position of the real-time RTT value within the threshold range by using the following formula:

[0018]

[0019] Among them, rate * and rate represent the transmission rates before and after dynamic adjustment respectively; α1 represents the precise acceleration factor, which is used to control the increase of the transmission rate when RTT is close to Tlow; β1 represents the precise deceleration factor, which is used to control the decrease of the transmission rate when RTT is close to Thigh; γ2 represents the influencing factor of the interval packet loss rate; PacketLossRate represents the packet loss rate; RTT represents the round-trip delay of data transmission; Thigh represents the maximum RTT threshold; Tlow represents the minimum RTT threshold.

[0020] α1 determines how quickly the transmission rate increases when the RTT is low. A larger α1 value will increase the rate faster, while a smaller α1 value will increase the rate more slowly. The recommended initial value is 0.3.

[0021] β1 determines how quickly the transmission rate decreases when the RTT is high (close to or above Tlow). A larger β1 value will cause the rate to decrease faster, while a smaller β1 value will cause the rate to decrease more gradually. The recommended initial value is 0.4.

[0022] γ2 controls the impact of packet loss rate on transmission rate. A larger γ2 value will reduce the rate more when the packet loss rate is high, while a smaller γ value will reduce the impact of packet loss rate. The recommended initial value is between 0.1 and 0.5, and the specific value needs to be adjusted according to the network environment.

[0023] The packet loss rate reflects the current network transmission quality. The packet loss rate value in the current network transmission. When the packet loss rate is high, the transmission rate is reduced by γ2×PacketLossRate to improve the transmission quality and stability.

[0024] Smooth acceleration with low RTT within the interval:

[0025] When the RTT is low within the interval, a smooth and accurate rate increase is achieved, which avoids drastic rate fluctuations and improves system stability. The value of the cosine function is close to 1. This significantly increases the rate and can quickly improve the transmission efficiency under low load. As RTT gradually approaches Thigh, When it is close to 1, the value of the cosine function is close to 0, and the rate of increase slows down, avoiding over-acceleration and keeping the system stable.

[0026] High RTT rapid deceleration within the interval:

[0027] When the RTT in the interval is high, the rate can be reduced quickly and smoothly, relieving network pressure in time and preventing network congestion. The value is small, the denominator is close to 1, and the value of the fraction is close to 1. This means that when RTT just exceeds Tlow, the deceleration effect is significant, which can quickly reduce the transmission rate and relieve network pressure. As the value gradually increases, the denominator becomes larger, the value of the fraction becomes smaller, and the contribution to the reduction in rate becomes smaller, thus making the deceleration process a smooth transition and avoiding system instability caused by excessive deceleration.

[0028] Interval packet loss rate influencing factor: γ2×PacketLossRate

[0029] During the acceleration process, the packet loss rate is taken into account to avoid an increase in the packet loss rate due to excessive acceleration. Even if the RTT is low, if the packet loss rate is high, acceleration should not be performed to prevent network congestion. γ2 controls the impact of the packet loss rate on the rate. The general value range is 0.05 to 0.2, and the optimal value is determined through experiments.

[0030] Furthermore, the method of adaptively adjusting the maximum and minimum RTT thresholds is as follows:

[0031] When the RTT average is less than 50ms and the bandwidth utilization is over 80%, increase Thigh and decrease Tlow:

[0032] Thigh=Thigh×(1+α Thigh )

[0033] Tlow=Tlow×(1-α Tlow )

[0034] When the RTT average is greater than 100ms or the packet loss rate exceeds 2%, reduce Thigh and increase Tlow:

[0035] Thigh=Thigh×(1-α Thigh )

[0036] Tlow=Tlow×(1+α Tlow )

[0037] Among them, Thigh represents the maximum RTT threshold; Tlow represents the minimum RTT threshold, α Tlow represents the low threshold dynamic adjustment factor, α Thigh Indicates the high threshold dynamic adjustment factor. The dynamic adjustment factor value range is set between 0.1 and 0.3.

[0038] An adaptive threshold adjustment mechanism is introduced to dynamically adjust the maximum threshold (Thigh) and minimum threshold (Tlow) of RTT by real-time monitoring of RTT, bandwidth utilization and packet loss rate. When the network condition is good, the maximum threshold is appropriately increased and the minimum threshold is appropriately decreased to fully utilize the bandwidth; when the network condition is poor, the maximum threshold is appropriately decreased and the minimum threshold is appropriately increased to prevent congestion.

[0039] Furthermore, the specific manner of slowing down or adjusting the transmission rate is as follows:

[0040] Deceleration phase:

[0041] When the number of times that RTT is higher than Thigh exceeds N1 or the packet loss rate exceeds 5%, the transmission rate is adjusted according to the following formula and the rapid deceleration stage is entered:

[0042]

[0043] Otherwise, the transmission rate is adjusted according to the following formula and the normal deceleration stage is entered:

[0044]

[0045] Acceleration phase:

[0046] When the number of times that RTT is lower than Tlow exceeds N2, the transmission rate is adjusted according to the following formula and the rapid acceleration stage is entered:

[0047] rate * =rate+N2×δ×(1-γ3×PacketLossRate)

[0048] Otherwise, the transmission rate is adjusted according to the following formula and the normal acceleration stage is entered:

[0049] rate * =rate+δ×(1-γ3×PacketLossRate)

[0050] Among them, Thigh represents the maximum RTT threshold; Tlow represents the minimum RTT threshold; γ represents the packet loss rate deceleration factor; PacketLossRate represents the packet loss rate; RTT represents the round-trip delay of data transmission; rate * and rate represent the transmission rates before and after dynamic adjustment respectively; δ represents the incremental step size, which is used to control the acceleration amplitude; N1 and N2 represent the number of times of entering the judgment in the acceleration stage and the deceleration stage respectively; β represents the deceleration factor, which is used to control the deceleration amplitude; γ3 represents the packet loss rate acceleration influencing factor;

[0051] Compared with the prior art, this algorithm introduces the factor γ and performs complex adjustments in combination with the packet loss rate, providing a more sophisticated and flexible rate adjustment mechanism.

[0052] Through the multi-factor adjustment mechanism in the acceleration and deceleration stages, when the RTT value is lower than Thigh, the system will quickly increase the transmission rate; when the RTT value is higher than Tlow, the system will quickly reduce the transmission rate.

[0053] In a second aspect, an RDMA network transmission system based on adaptive threshold and hybrid congestion control includes:

[0054] Transmission link construction unit: constructs RDMA transmission link based on RDMA;

[0055] RTT threshold dynamic acquisition unit: Adaptively adjusts the RTT maximum and minimum thresholds based on real-time monitored link transmission data;

[0056] Multi-factor rate adjustment unit: Based on the adaptively adjusted RTT threshold and packet loss rate threshold, the transmission rate is dynamically and comprehensively adjusted through a multi-factor adjustment mechanism;

[0057] When the real-time RTT is greater than the maximum RTT threshold or the packet loss rate exceeds the minimum packet loss rate threshold, the transmission rate is decelerated and adjusted, and the RTT threshold dynamic acquisition unit is called; when the real-time RTT is less than the minimum RTT threshold, the transmission rate is accelerated and adjusted, and the RTT threshold dynamic acquisition unit is called;

[0058] If the real-time RTT is between the RTT minimum threshold and the RTT maximum threshold, the interval adaptive speed regulation unit is called; the interval adaptive speed regulation unit: adopts the interval adaptive speed regulation mechanism to dynamically adjust the transmission rate according to the position of the real-time RTT value within the threshold range, and calls the RTT threshold dynamic acquisition unit.

[0059] Furthermore, the adaptive speed regulation mechanism within the interval is used to dynamically adjust the transmission rate according to the position of the real-time RTT value within the threshold range by using the following formula:

[0060]

[0061] Among them, rate *and rate represent the transmission rates before and after dynamic adjustment respectively; α1 represents the precise acceleration factor, which is used to control the increase of the transmission rate when RTT is close to Tlow; β1 represents the precise deceleration factor, which is used to control the decrease of the transmission rate when RTT is close to Thigh; γ2 represents the influencing factor of the interval packet loss rate; PacketLossRate represents the packet loss rate; RTT represents the round-trip delay of data transmission; Thigh represents the maximum RTT threshold; Tlow represents the minimum RTT threshold.

[0062] Furthermore, the method of adaptively adjusting the maximum and minimum RTT thresholds is as follows:

[0063] When the RTT average is less than 50ms and the bandwidth utilization is over 80%, increase Thigh and decrease Tlow:

[0064] Thigh=Thigh×(1+α Thigh )

[0065] Tlow=Tlow×(1-α Tlow )

[0066] When the RTT average is greater than 100ms or the packet loss rate exceeds 2%, reduce Thigh and increase Tlow:

[0067] Thigh=Thigh×(1-α Thigh )

[0068] Tlow=Tlow×(1+α Tlow )

[0069] Among them, Thigh represents the maximum RTT threshold; Tlow represents the minimum RTT threshold, α Tlow represents the low threshold dynamic adjustment factor, α Thigh Indicates the high threshold dynamic adjustment factor. The dynamic adjustment factor value range is set between 0.1 and 0.3.

[0070] Furthermore, the specific method of slowing down or accelerating the transmission rate is as follows:

[0071] Deceleration phase:

[0072] When the number of times that RTT is higher than Thigh exceeds N1 or the packet loss rate exceeds 5%, the transmission rate is adjusted according to the following formula and the rapid deceleration stage is entered:

[0073]

[0074] Otherwise, the transmission rate is adjusted according to the following formula and the normal deceleration stage is entered:

[0075]

[0076] Acceleration phase:

[0077] When the number of times that RTT is lower than Tlow exceeds N2, the transmission rate is adjusted according to the following formula and the rapid acceleration stage is entered:

[0078] rate * =rate+N2×δ×(1-γ3×PacketLossRate)

[0079] Otherwise, the transmission rate is adjusted according to the following formula and the normal acceleration stage is entered:

[0080] rate * =rate+δ×(1-γ3×PacketLossRate)

[0081] Among them, Thigh represents the maximum RTT threshold; Tlow represents the minimum RTT threshold; γ represents the packet loss rate deceleration factor; PacketLossRate represents the packet loss rate; RTT represents the round-trip delay of data transmission; rate * and rate represent the transmission rates before and after dynamic adjustment, respectively; δ represents the incremental step size, which is used to control the acceleration amplitude; N1 and N2 represent the number of times the judgment is entered in the acceleration stage and the deceleration stage, respectively; β represents the deceleration factor, which is used to control the deceleration amplitude; and γ3 represents the packet loss rate acceleration influencing factor.

[0082] According to a third aspect, an electronic device includes:

[0083] one or more processors;

[0084] a memory storing one or more computer programs;

[0085] The processor calls the computer program to implement:

[0086] The steps of the above RDMA network transmission method based on adaptive threshold and hybrid congestion control.

[0087] In a fourth aspect, a computer-readable storage medium stores a computer program, wherein the computer program is called by a processor to implement:

[0088] The above steps of the RDMA network transmission method based on adaptive threshold and hybrid congestion control.

[0089] Beneficial Effects

[0090] Compared with the existing method, the advantages of the present invention are:

[0091] 1. An adaptive threshold adjustment mechanism is introduced to dynamically adjust the maximum threshold (Thigh) and minimum threshold (Tlow) of RTT by real-time monitoring of RTT, bandwidth utilization and packet loss rate. When the network condition is good, the maximum threshold is appropriately increased and the minimum threshold is appropriately decreased to make full use of the bandwidth; when the network condition is poor, the maximum threshold is appropriately decreased and the minimum threshold is appropriately increased to prevent congestion.

[0092] 2. It combines delay-based and packet loss-based control strategies. By monitoring RTT, it detects mild congestion and performs preventive control. By monitoring the packet loss rate in real time, when the packet loss rate or RTT exceeds a certain threshold, it executes a rapid deceleration strategy to reduce the sending rate and prevent severe congestion.

[0093] 3. Through the combination of multiple acceleration factors and influencing factors, efficient response to different types of congestion is achieved, making rate adjustment smoother and more accurate. Specifically, it includes the main acceleration factor, the secondary packet loss rate influencing factor, the secondary acceleration factor, the main deceleration factor, the deceleration stage packet loss rate influencing factor, the secondary deceleration factor, etc.

[0094] 4. Real-time monitoring of RTT, packet loss rate and RTT gradient, and dynamic adjustment of the sending rate according to changes in these indicators, providing higher network adaptability and robustness. This innovation ensures that the system can respond to changes in network conditions in a timely manner and optimize transmission performance.

[0095] 5. In-range adaptive speed regulation mechanism: By real-time monitoring of the position of the RTT value within the threshold range, the system can subtly and accurately adjust the transmission rate to make the rate change more gradual and smooth. When the RTT is small, the rate increases significantly, which can quickly improve the transmission efficiency; as the RTT gradually approaches the maximum threshold, the deceleration increases significantly, avoiding excessive acceleration under high load and maintaining system stability.

[0096] Through the above improvements, the technical solution of the present invention has higher adaptability and efficiency in network congestion control, and can better meet the needs of data centers for high bandwidth and low latency. It not only improves transmission efficiency, but also enhances the stability and robustness of the system in various complex network environments. BRIEF DESCRIPTION OF THE DRAWINGS

[0097] Figure 1 It is a schematic flow chart of the method provided by the example of the present invention;

[0098] Figure 2 This is the RTT change analysis diagram described in the example of the present invention;

[0099] Figure 3 This is a transmission rate split diagram described in an example of the present invention. DETAILED DESCRIPTION

[0100] The present invention will be further described below with reference to the accompanying drawings and examples.

[0101] like Figure 1 As shown, an RDMA network transmission method based on adaptive threshold and hybrid congestion control provided by an example of the present invention includes:

[0102] Step 1: Build an RDMA transmission link;

[0103] Step 2: Adaptively adjust the RTT maximum and minimum thresholds based on the real-time monitored link transmission data;

[0104] The method of adaptively adjusting the RTT maximum and minimum thresholds is as follows:

[0105] When the RTT average is less than 50ms and the bandwidth utilization is over 80%, increase Thigh and decrease Tlow:

[0106] Thigh=Thigh×(1+α Thigh )

[0107] Tlow=Tlow×(1-α Tlow )

[0108] When the RTT average is greater than 100ms or the packet loss rate exceeds 2%, reduce Thigh and increase Tlow:

[0109] Thigh=Thigh×(1-α Thigh )

[0110] Tlow=Tlow×(1+α Tlow )

[0111] Among them, Thigh represents the maximum RTT threshold; Tlow represents the minimum RTT threshold, α Tlow represents the low threshold dynamic adjustment factor, α Thigh Indicates the high threshold dynamic adjustment factor. The dynamic adjustment factor value range is set between 0.1 and 0.3.

[0112] An adaptive threshold adjustment mechanism is introduced to dynamically adjust the maximum threshold (Thigh) and minimum threshold (Tlow) of RTT by real-time monitoring of RTT, bandwidth utilization and packet loss rate. When the network condition is good, the maximum threshold is appropriately increased and the minimum threshold is appropriately decreased to fully utilize the bandwidth; when the network condition is poor, the maximum threshold is appropriately decreased and the minimum threshold is appropriately increased to prevent congestion.

[0113] Step 3: Based on the adaptively adjusted RTT threshold and packet loss rate threshold, the transmission rate is dynamically and comprehensively adjusted through a multi-factor adjustment mechanism;

[0114] When the real-time RTT is greater than the maximum RTT threshold or the packet loss rate exceeds the minimum packet loss rate threshold, the transmission rate is decelerated and the process returns to step 2; when the real-time RTT is less than the minimum RTT threshold, the transmission rate is accelerated and the process returns to step 2;

[0115] The specific ways to slow down or adjust the transmission rate are as follows:

[0116] Deceleration phase:

[0117] When the number of times that RTT is higher than Thigh exceeds N1 or the packet loss rate exceeds 5%, the transmission rate is adjusted according to the following formula and the rapid deceleration stage is entered:

[0118]

[0119] Otherwise, the transmission rate is adjusted according to the following formula and the normal deceleration stage is entered:

[0120]

[0121] Acceleration phase:

[0122] When the number of times that RTT is lower than Tlow exceeds N2, the transmission rate is adjusted according to the following formula and the rapid acceleration stage is entered:

[0123] rate * =rate+N2×δ×(1-γ3×PacketLossRate)

[0124] Otherwise, the transmission rate is adjusted according to the following formula and the normal acceleration stage is entered:

[0125] rate * =rate+δ×(1-γ3×PacketLossRate)

[0126] Among them, Thigh represents the maximum RTT threshold; Tlow represents the minimum RTT threshold; γ represents the packet loss rate deceleration factor; PacketLossRate represents the packet loss rate; RTT represents the round-trip delay of data transmission; rate * and rate represent the transmission rates before and after dynamic adjustment respectively; δ represents the incremental step size, which is used to control the acceleration amplitude; N1 and N2 represent the number of times of entering the judgment in the acceleration stage and the deceleration stage respectively; β represents the deceleration factor, which is used to control the deceleration amplitude; γ3 represents the packet loss rate acceleration influencing factor;

[0127] If the real-time RTT is between the RTT minimum threshold and the RTT maximum threshold, proceed to step 4;

[0128] Step 4: Adopt the interval adaptive speed regulation mechanism to dynamically adjust the transmission rate according to the position of the real-time RTT value within the threshold range, and return to step 2.

[0129] The adaptive speed regulation mechanism within the interval is used to dynamically adjust the transmission rate according to the position of the real-time RTT value within the threshold range by using the following formula:

[0130]

[0131] Among them, rate * and rate represent the transmission rates before and after dynamic adjustment respectively; α1 represents the precise acceleration factor, which is used to control the increase of the transmission rate when RTT is close to Tlow; β1 represents the precise deceleration factor, which is used to control the decrease of the transmission rate when RTT is close to Thigh; γ2 represents the influencing factor of the interval packet loss rate; PacketLossRate represents the packet loss rate; RTT represents the round-trip delay of data transmission; Thigh represents the maximum RTT threshold; Tlow represents the minimum RTT threshold.

[0132] α1 determines how quickly the transmission rate increases when the RTT is low. A larger α1 value will increase the rate faster, while a smaller α1 value will increase the rate more slowly. The recommended initial value is 0.3.

[0133] β1 determines how quickly the transmission rate decreases when the RTT is high (close to or above Tlow). A larger β1 value will cause the rate to decrease faster, while a smaller β1 value will cause the rate to decrease more gradually. The recommended initial value is 0.4.

[0134] γ2 controls the impact of packet loss rate on transmission rate. A larger γ2 value will reduce the rate more when the packet loss rate is high, while a smaller γ value will reduce the impact of packet loss rate. The recommended initial value is between 0.1 and 0.5, and the specific value needs to be adjusted according to the network environment.

[0135] The packet loss rate reflects the current network transmission quality. The packet loss rate value in the current network transmission. When the packet loss rate is high, the transmission rate is reduced by γ2×PacketLossRate to improve the transmission quality and stability.

[0136] Smooth acceleration with low RTT within the interval:

[0137] When the RTT is low within the interval, a smooth and accurate rate increase is achieved, which avoids drastic rate fluctuations and improves system stability. The value of the cosine function is close to 1. This significantly increases the rate and can quickly improve the transmission efficiency under low load. As RTT gradually approaches Thigh, When it is close to 1, the value of the cosine function is close to 0, and the rate of increase slows down, avoiding over-acceleration and keeping the system stable.

[0138] High RTT rapid deceleration within the interval:

[0139] When the RTT in the interval is high, the rate can be reduced quickly and smoothly, relieving network pressure in time and preventing network congestion. The value is small, the denominator is close to 1, and the value of the fraction is close to 1. This means that when RTT just exceeds Tlow, the deceleration effect is significant, which can quickly reduce the transmission rate and relieve network pressure. As the value gradually increases, the denominator becomes larger, the value of the fraction becomes smaller, and the contribution to the reduction in rate becomes smaller, thus making the deceleration process a smooth transition and avoiding system instability caused by excessive deceleration.

[0140] Interval packet loss rate influencing factor: γ2×PacketLossRate

[0141] During the acceleration process, the packet loss rate is taken into account to avoid an increase in the packet loss rate due to excessive acceleration. Even if the RTT is low, if the packet loss rate is high, acceleration should not be performed to prevent network congestion. γ2 controls the impact of the packet loss rate on the rate. The general value range is 0.05 to 0.2, and the optimal value is determined through experiments.

[0142] By introducing an adaptive threshold adjustment mechanism, the RTT threshold can be automatically optimized according to real-time changes in network conditions, thereby making full use of bandwidth when network conditions are good and effectively preventing congestion when network conditions are poor. This mechanism improves the robustness and adaptability of the system, ensuring efficient transmission performance under different network conditions.

[0143] If the real-time RTT is between the RTT minimum threshold and the RTT maximum threshold, by real-time monitoring of the position of the RTT value within the threshold range, the system can subtly and accurately adjust the transmission rate to make the rate change more gradual and smooth. When the RTT is small, the rate increases significantly, which can quickly improve the transmission efficiency; as the RTT gradually approaches the maximum threshold, the deceleration increases significantly, avoiding excessive acceleration under high load and keeping the system stable.

[0144] Compared with the prior art, this algorithm introduces the factor γ and performs complex adjustments in combination with the packet loss rate, providing a more sophisticated and flexible rate adjustment mechanism.

[0145] Through the multi-factor adjustment mechanism in the acceleration and deceleration stages, when the RTT value is lower than Thigh, the system will quickly increase the transmission rate; when the RTT value is higher than Tlow, the system will quickly reduce the transmission rate.

[0146] Experimental verification

[0147] A simulated RDMA network transmission system is provided, which adopts adaptive threshold and multi-factor adjustment mechanism for effective congestion control. In 100 transmission cycles, the system dynamically adjusts the transmission rate according to the real-time round-trip time (RTT) and packet loss rate to test the performance and adjustment effect under different network conditions.

[0148] This experiment was conducted in a controlled network simulation environment, using software tools to generate network traffic, simulate network latency and packet loss. This environment allows us to precisely control various network parameters in order to simulate different network conditions, and monitor and record various parameters during the transmission process.

[0149] In order to comprehensively evaluate the performance of the RDMA transmission system, the following parameters are set:

[0150] Thigh (high RTT threshold): Initially set to 100ms, it is used to determine the upper limit of RTT. If this value is exceeded, the transmission rate will be reduced.

[0151] Tlow (low RTT threshold): Initially set to 50ms, it is used to determine the lower limit of RTT. If it is lower than this value, the transmission rate will increase.

[0152] β (deceleration factor): set to 0.3 to reduce the transmission rate when Thigh is exceeded.

[0153] α1 (precision acceleration factor): set to 0.3, used to adjust the sensitivity of the transmission rate and respond to real-time changes in RTT.

[0154] β1 (precision slowdown factor): set to 0.3, used to reduce the transmission rate when RTT approaches or exceeds Thigh.

[0155] γ (Packet loss rate deceleration factor): Set to 0.2 to reduce the drop in transmission rate when packet loss is detected.

[0156] γ2 (interval packet loss rate impact factor): set to 0.2 to adjust the impact of packet loss rate in a specific interval on the transmission rate.

[0157] γ3 (packet loss rate acceleration factor): set to 0.1 to prevent the packet loss rate from increasing due to excessively fast growth.

[0158] v (speed factor): set to 0.3 to increase the transmission rate when the RTT is significantly lower than Tlow.

[0159] α Tlow and α Thigh(Dynamic threshold adjustment factor): set to 0.03 and 0.02 respectively, used to dynamically adjust the values ​​of Thigh and Tlow according to the network status.

[0160] N1 and N2 represent the number of times the judgment is entered in the acceleration stage and the deceleration stage, respectively, and are set to 5 here.

[0161] During the 100 transmission tests, the system will automatically record the following data after each transmission:

[0162] RTT (round trip time), packet loss rate, transmission rate, dynamically adjusted Thigh and Tlow

[0163] These data will be used to analyze the network behavior and system response to evaluate the effectiveness of adaptive threshold adjustment strategies and hybrid congestion control strategies.

[0164] Start the experiment according to the method flow of the present invention:

[0165] Use a network simulator to generate the initial network state and start the first data transmission.

[0166] Record the initial RTT, packet loss rate, and transmission rate.

[0167] Dynamic adjustment stage:

[0168] ·Continuously monitor network status and dynamically adjust transmission rate based on the comparison between real-time RTT and set Thigh and Tlow.

[0169] If the RTT exceeds Thigh, reduce the transmission rate according to β1; if the RTT is lower than Tlow, increase the transmission rate according to δ.

[0170] Use γ, γ2, and γ3 to adjust the transmission rate according to the packet loss rate.

[0171] Data recording and feedback:

[0172] After each transmission, the RTT, packet loss rate, transmission rate and dynamically adjusted threshold are recorded.

[0173] Analyze data and optimize parameter settings.

[0174] RTT change analysis, such as Figure 2 As shown, in Figure 2In the figure, the X-axis label: Transmission Number, represents the number of simulated transmissions, from 1 to 100, and each dot represents the result of a simulated transmission; the Y-axis label: RTT (ms), represents the round-trip time of the network, in milliseconds (ms). This is an indicator that measures the time required for a signal to travel from the sender to the receiver and then back to the sender, reflecting the delay status of the network. In addition to the dot data showing RTT, there are two lines representing dynamic thresholds (Thigh and Tlow): the dotted line (Thigh Dynamic): shows how the upper threshold Thigh is dynamically adjusted as the transmission proceeds. The change of this line reflects the system's adaptive adjustment of the upper threshold when network conditions change; the solid line (Tlow Dynamic): shows the dynamic change of the lower threshold Tlow. Although the change is not large in this particular experiment, it shows that the system can adjust this threshold to respond to network conditions when necessary.

[0175] From the collected data, we can see that RTT fluctuated significantly at the beginning of the experiment, and then with the automatic adjustment of parameters, the fluctuation gradually decreased and tended to be stable.

[0176] Specifically:

[0177] Initial fluctuations in RTT: At the beginning of the experiment, RTT quickly rose to more than 100ms, triggering the deceleration mechanism and resulting in an initial drop in the transmission rate. During this stage, RTT fluctuated greatly, mainly affected by the initial instability of the simulated network environment. Mid-term adjustment effect: With the adjustment of the adaptive threshold, especially after Thigh and Tlow were dynamically adjusted according to network conditions, RTT began to fluctuate within a more stable range. For example, in the middle of the transmission process, when RTT gradually adapted to the adjustment of Thigh, even in the case of high packet loss rate, RTT could remain near Thigh, showing the immediate feedback of parameter adjustment and the adaptability of the system.

[0178] Late stability: In the later stages of the experiment, the stability of RTT was further improved. This is due to the alpha Tlow and α Thigh Slight adjustments to the parameters enable the system to respond more accurately to changes in RTT, thereby optimizing the transmission strategy.

[0179] In summary, when the technical solution of the present invention is used for data transmission, the system can respond to changes in RTT more accurately.

[0180] Transfer rate analysis, such as Figure 3As shown in the figure, the X-axis label: Transmission Number, represents the order of simulated transmission, a total of 100 times; the Y-axis label: Transmission Rate, represents the data transmission rate, the unit is usually Mbps (megabits per second) or other suitable units. This indicator reflects the speed of data transmission in the network and is an important indicator for measuring network performance.

[0181] The change of transmission rate directly reflects the effect of congestion control strategy.

[0182] Analyzing the data shows:

[0183] Responsive adjustment of the rate: When RTT exceeds Thigh, the transmission rate drops significantly, which is a direct result of the beta1 factor. When RTT drops below Tlow, an increase in the transmission rate is observed, indicating that the delta factor effectively promotes the recovery of the rate.

[0184] The correlation between rate and packet loss rate: During periods of high packet loss rate, the transmission rate is usually adjusted sharply, especially the gamma series factors. The transmission rate is adjusted to cope with packet loss and reduce possible data transmission interruptions or delays.

[0185] Transmission rate in the stable stage: In the middle and late stages of the experiment, with the optimization of system parameters and the adaptation of the network environment, the transmission rate tended to stabilize. This shows that the system has been adjusted to a relatively ideal state based on real-time data, effectively balancing the rate and stability.

[0186] This experiment shows that the adaptive adjustment of the RTT threshold and the comprehensive use of multiple congestion control parameters provided by the technical solution of the present invention can significantly improve the performance of the RDMA transmission system in an unstable network environment. Dynamic adjustment of Thigh and Tlow, combined with real-time rate adjustment strategies, not only optimizes data transmission efficiency, but also enhances the system's resilience to network fluctuations. Future work can further explore the optimal configuration of these parameters under different network environments, as well as how to automatically adjust these parameters through machine learning algorithms to achieve more advanced network adaptive control.

[0187] The present invention also provides an RDMA network transmission system based on adaptive threshold and hybrid congestion control, including:

[0188] Transmission link construction unit: constructs RDMA transmission link based on RDMA;

[0189] RTT threshold dynamic acquisition unit: Adaptively adjusts the RTT maximum and minimum thresholds based on real-time monitored link transmission data;

[0190] Multi-factor rate adjustment unit: Based on the adaptively adjusted RTT threshold and packet loss rate threshold, the transmission rate is dynamically and comprehensively adjusted through a multi-factor adjustment mechanism;

[0191] When the real-time RTT is greater than the maximum RTT threshold or the packet loss rate exceeds the minimum packet loss rate threshold, the transmission rate is decelerated and adjusted, and the RTT threshold dynamic acquisition unit is called; when the real-time RTT is less than the minimum RTT threshold, the transmission rate is accelerated and adjusted, and the RTT threshold dynamic acquisition unit is called;

[0192] If the real-time RTT is between the RTT minimum threshold and the RTT maximum threshold, the interval adaptive speed regulation unit is called; the interval adaptive speed regulation unit: adopts the interval adaptive speed regulation mechanism to dynamically adjust the transmission rate according to the position of the real-time RTT value within the threshold range, and calls the RTT threshold dynamic acquisition unit.

[0193] The adaptive speed regulation mechanism within the interval is used to dynamically adjust the transmission rate according to the position of the real-time RTT value within the threshold range by using the following formula:

[0194]

[0195] Among them, rate * and rate represent the transmission rates before and after dynamic adjustment respectively; α1 represents the precise acceleration factor, which is used to control the increase of the transmission rate when RTT is close to Tlow; β1 represents the precise deceleration factor, which is used to control the decrease of the transmission rate when RTT is close to Thigh; γ2 represents the influencing factor of the interval packet loss rate; PacketLossRate represents the packet loss rate; RTT represents the round-trip delay of data transmission; Thigh represents the maximum RTT threshold; Tlow represents the minimum RTT threshold.

[0196] The method of adaptively adjusting the RTT maximum and minimum thresholds is as follows:

[0197] When the RTT average is less than 50ms and the bandwidth utilization is over 80%, increase Thigh and decrease Tlow:

[0198] Thigh=Thigh×(1+α Thigh )

[0199] Tlow=Tlow×(1-α Tlow )

[0200] When the RTT average is greater than 100ms or the packet loss rate exceeds 2%, reduce Thigh and increase Tlow:

[0201] Thigh=Thigh×(1-α Thigh )

[0202] Tlow=Tlow×(1+α Tlow )

[0203] Among them, Thigh represents the maximum RTT threshold; Tlow represents the minimum RTT threshold, α Tlow represents the low threshold dynamic adjustment factor, α Thigh Indicates the high threshold dynamic adjustment factor. The dynamic adjustment factor value range is set between 0.1 and 0.3.

[0204] The specific methods for slowing down or accelerating the transmission rate are as follows:

[0205] Deceleration phase:

[0206] When the number of times that RTT is higher than Thigh exceeds N1 or the packet loss rate exceeds 5%, the transmission rate is adjusted according to the following formula and the rapid deceleration stage is entered:

[0207]

[0208] Otherwise, the transmission rate is adjusted according to the following formula and the normal deceleration stage is entered:

[0209]

[0210] Acceleration phase:

[0211] When the number of times that RTT is lower than Tlow exceeds N2, the transmission rate is adjusted according to the following formula and the rapid acceleration stage is entered:

[0212] rate * =rate+N2×δ×(1-γ3×PacketLossRate)

[0213] Otherwise, the transmission rate is adjusted according to the following formula and the normal acceleration stage is entered:

[0214] rate * =rate+δ×(1-γ3×PacketLossRate)

[0215] Among them, Thigh represents the maximum RTT threshold; Tlow represents the minimum RTT threshold; γ represents the packet loss rate deceleration factor; PacketLossRate represents the packet loss rate; RTT represents the round-trip delay of data transmission; rate * and rate represent the transmission rates before and after dynamic adjustment, respectively; δ represents the incremental step size, which is used to control the acceleration amplitude; N1 and N2 represent the number of times the judgment is entered in the acceleration stage and the deceleration stage, respectively; β represents the deceleration factor, which is used to control the deceleration amplitude; and γ3 represents the packet loss rate acceleration influencing factor.

[0216] Among them, please refer to the above method content for the specific implementation process of each module or unit, the present invention will not be repeated here, and the division of the above functional modules / units is only for example illustration. In some embodiments, some functional modules can be merged, and some functional modules / units can be split. Each functional module / unit can be implemented in software or hardware or a combination of software and hardware. Among them, the hardware and software equipment includes but is not limited to general-purpose computer equipment, programmable gate arrays, digital signal processors, microprocessors and their corresponding programming or burning software.

[0217] The present invention also provides an electronic device, including:

[0218] one or more processors;

[0219] a memory storing one or more computer programs;

[0220] The processor calls the computer program to implement:

[0221] The steps of the above RDMA network transmission method based on adaptive threshold and hybrid congestion control.

[0222] For the specific implementation process of each step, please refer to the description of the above method.

[0223] The present invention also provides a computer-readable storage medium storing a computer program, wherein the computer program is called by a processor to implement:

[0224] The above steps of the RDMA network transmission method based on adaptive threshold and hybrid congestion control.

[0225] For the specific implementation process of each step, please refer to the description of the above method.

[0226] The readable storage medium is a computer-readable storage medium, which may be an internal storage unit of the software and hardware device described in any of the aforementioned embodiments, such as a hard disk or memory of a controller. The readable storage medium may also be an external storage device of the controller, such as a plug-in hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (Flash Card), etc. equipped on the controller. Furthermore, the readable storage medium may also include both an internal storage unit of the controller and an external storage device. The readable storage medium is used to store the computer program and other programs and data required by the controller. The readable storage medium may also be used to temporarily store data that has been output or is to be output.

[0227] Based on this understanding, the technical solution of the present invention, in essence or in other words, the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product, which is stored in a storage medium and includes several instructions for enabling a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned readable storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc., various media that can store program codes.

[0228] Those skilled in the art will appreciate that the embodiments of the present application can be provided as methods, systems, or computer program products. Therefore, the present application can adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present application can adopt the form of a computer program product implemented on one or more computer-readable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program codes. The present application is a device for realizing the function specified in one flow or multiple flows of the flow chart and / or one or multiple blocks of the block diagram with reference to the instructions executed by the processor according to the method, device (system) and computer program product of the embodiment of the present application. These computer program instructions can also be stored in a computer-readable memory that can guide a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer-readable memory produce a product including an instruction device, which realizes the function specified in one flow or multiple flows of the flow chart and / or one or multiple blocks of the block diagram. These computer program instructions may also be loaded onto a computer or other programmable data processing device so that a series of operational steps are executed on the computer or other programmable device to produce a computer-implemented process, whereby the instructions executed on the computer or other programmable device provide steps for implementing the functions specified in one or more processes in the flowchart and / or one or more boxes in the block diagram.

[0229] It should be emphasized that the examples described in the present invention are illustrative rather than restrictive, and therefore the present invention is not limited to the examples described in the specific embodiments. Any other embodiments derived by those skilled in the art based on the technical solution of the present invention that do not depart from the purpose and scope of the present invention, whether modified or replaced, also fall within the scope of protection of the present invention.

Claims

1. A RDMA network transmission method based on adaptive threshold and hybrid congestion control, characterized in that: include: Step 1: Build an RDMA transmission link; Step 2: Adaptively adjust the RTT maximum and minimum thresholds based on the real-time monitored link transmission data; Step 3: Based on the adaptively adjusted RTT threshold and packet loss rate threshold, the transmission rate is dynamically and comprehensively adjusted through a multi-factor adjustment mechanism; When the real-time RTT is greater than the maximum RTT threshold or the packet loss rate exceeds the minimum packet loss rate threshold, the transmission rate is decelerated and the process returns to step 2; when the real-time RTT is less than the minimum RTT threshold, the transmission rate is accelerated and the process returns to step 2; If the real-time RTT is between the RTT minimum threshold and the RTT maximum threshold, proceed to step 4; Step 4: Adopt the interval adaptive speed regulation mechanism to dynamically adjust the transmission rate according to the position of the real-time RTT value within the threshold range, and return to step 2; Among them, RTT represents the round-trip delay of data transmission; The method of adaptively adjusting the maximum and minimum RTT thresholds is as follows: When the RTT average is less than 50ms and the bandwidth utilization is over 80%, increase Thigh and decrease Tlow: Thigh=Thigh×(1+α Thigh ) Tlow=Tlow×(1-α Tlow ) When the RTT average is greater than 100ms or the packet loss rate exceeds 2%, reduce Thigh and increase Tlow: Thigh=Thigh×(1-α Thigh ) Tlow=Tlow×(1+α Tlow ) Among them, Thigh represents the maximum RTT threshold; Tlow represents the minimum RTT threshold, α Tlow represents the low threshold dynamic adjustment factor, α Thigh Indicates the high threshold dynamic adjustment factor. The dynamic adjustment factor value range is set between 0.1 and 0.

3.

2. The method according to claim 1, characterized in that The adaptive speed regulation mechanism within the interval is used to dynamically adjust the transmission rate according to the position of the real-time RTT value within the threshold range by using the following formula: Among them, rate * and rate represent the transmission rates before and after dynamic adjustment respectively; α1 represents the precise acceleration factor, which is used to control the increase in the transmission rate when RTT is close to Tlow; β1 represents the precise deceleration factor, which is used to control the decrease in the transmission rate when RTT is close to Thigh; γ2 represents the influencing factor of the interval packet loss rate; PacketLossRate represents the packet loss rate.

3. The method according to claim 1, characterized in that: The specific ways to slow down or adjust the transmission rate are as follows: Deceleration phase: When the number of times that RTT is higher than Thigh exceeds N1 or the packet loss rate exceeds 5%, the transmission rate is adjusted according to the following formula and enters the rapid deceleration stage: Otherwise, the transmission rate is adjusted according to the following formula and the normal deceleration stage is entered: Acceleration phase: When the number of times that RTT is lower than Tlow exceeds N2, the transmission rate is adjusted according to the following formula and the rapid acceleration stage is entered: rate * =rate+N2×δ×(1-γ3×PacketLossRate) Otherwise, the transmission rate is adjusted according to the following formula and the normal acceleration stage is entered: rate * =rate+δ×(1-γ3×PacketLossRate) Among them, γ represents the packet loss rate deceleration factor; PacketLossRate represents the packet loss rate; rate * and rate represent the transmission rates before and after dynamic adjustment, respectively; δ represents the incremental step size, which is used to control the acceleration amplitude; N1 and N2 represent the number of times the judgment is entered in the acceleration stage and the deceleration stage, respectively; β represents the deceleration factor, which is used to control the deceleration amplitude; and γ3 represents the packet loss rate acceleration influencing factor.

4. An RDMA network transmission system based on adaptive threshold and hybrid congestion control, characterized in that: include: Transmission link construction unit: constructs RDMA transmission link based on RDMA; RTT threshold dynamic acquisition unit: Adaptively adjusts the RTT maximum and minimum thresholds based on real-time monitored link transmission data; Multi-factor rate adjustment unit: Based on the adaptively adjusted RTT threshold and packet loss rate threshold, the transmission rate is dynamically and comprehensively adjusted through a multi-factor adjustment mechanism; When the real-time RTT is greater than the maximum RTT threshold or the packet loss rate exceeds the minimum packet loss rate threshold, the transmission rate is decelerated and adjusted, and the RTT threshold dynamic acquisition unit is called; when the real-time RTT is less than the minimum RTT threshold, the transmission rate is accelerated and adjusted, and the RTT threshold dynamic acquisition unit is called; If the real-time RTT is between the RTT minimum threshold and the RTT maximum threshold, the adaptive speed control unit in the interval is called; In-interval adaptive speed regulation unit: Adopts the in-interval adaptive speed regulation mechanism to dynamically adjust the transmission rate according to the position of the real-time RTT value within the threshold range, and calls the RTT threshold dynamic acquisition unit; RTT represents the round-trip delay of data transmission; The method of adaptively adjusting the maximum and minimum RTT thresholds is as follows: When the RTT average is less than 50ms and the bandwidth utilization is over 80%, increase Thigh and decrease Tlow: Thigh=Thigh×(1+α Thigh ) Tlow=Tlow×(1-α Tlow ) When the RTT average is greater than 100ms or the packet loss rate exceeds 2%, reduce Thigh and increase Tlow: Thigh=Thigh×(1-α Thigh ) Tlow=Tlow×(1+α Tlow ) Among them, Thigh represents the maximum RTT threshold; Tlow represents the minimum RTT threshold, α Tlow represents the low threshold dynamic adjustment factor, α Thigh Indicates the high threshold dynamic adjustment factor. The dynamic adjustment factor value range is set between 0.1 and 0.

3.

5. The system according to claim 4, characterized in that The adaptive speed regulation mechanism within the interval is used to dynamically adjust the transmission rate according to the position of the real-time RTT value within the threshold range by using the following formula: Among them, rate * and rate represent the transmission rates before and after dynamic adjustment respectively; α1 represents the precise acceleration factor, which is used to control the increase in the transmission rate when RTT is close to Tlow; β1 represents the precise deceleration factor, which is used to control the decrease in the transmission rate when RTT is close to Thigh; γ2 represents the influencing factor of the interval packet loss rate; PacketLossRate represents the packet loss rate.

6. The system according to claim 4, characterized in that The specific methods for slowing down or accelerating the transmission rate are as follows: Deceleration phase: When the number of times that RTT is higher than Thigh exceeds N1 or the packet loss rate exceeds 5%, the transmission rate is adjusted according to the following formula and enters the rapid deceleration stage: Otherwise, the transmission rate is adjusted according to the following formula and the normal deceleration stage is entered: Acceleration phase: When the number of times that RTT is lower than Tlow exceeds N2, the transmission rate is adjusted according to the following formula and the rapid acceleration stage is entered: rate * =rate+N2×δ×(1-γ3×PacketLossRate) Otherwise, the transmission rate is adjusted according to the following formula and the normal acceleration stage is entered: rate * =rate+δ×(1-γ3×PacketLossRate) Among them, γ represents the packet loss rate deceleration factor; PacketLossRate represents the packet loss rate; rate * and rate represent the transmission rates before and after dynamic adjustment, respectively; δ represents the incremental step size, which is used to control the acceleration amplitude; N1 and N2 represent the number of times the judgment is entered in the acceleration stage and the deceleration stage, respectively; β represents the deceleration factor, which is used to control the deceleration amplitude; and γ3 represents the packet loss rate acceleration influencing factor.

7. An electronic device, characterized in that: include: one or more processors; a memory storing one or more computer programs; The processor calls the computer program to implement: The steps of the RDMA network transmission method based on adaptive threshold and hybrid congestion control as described in any one of claims 1-3.

8. A computer-readable storage medium, characterized in that: A computer program is stored, which is called by a processor to implement: The steps of an RDMA network transmission method based on adaptive threshold and hybrid congestion control as described in any one of claims 1-3.

Images