阅读说明：本技术 一种带宽探测阶段的发送速率调整方法及拥塞控制算法 (transmission rate adjusting method and congestion control algorithm in bandwidth detection stage ) 是由 黄家玮 彭澍 王建新 于 2019-10-25 设计创作，主要内容包括：本发明公开了一种带宽探测阶段的发送速率调整方法及拥塞控制算法,在带宽探测阶段的第一个RTT,先用增速因子g增加发送速率,然后发送一轮数据包,以进行带宽探测。根据是否在一定时间内收到该轮发送的所有数据包的ACK包来判断当前网络的拥塞状况。当前网络的状况较好时,依据当前往返延时和历史最大传输速率调整发送速率的增速因子。当前网络的状况较差时,降低发送速率发送一轮数据包,即在该带宽探测阶段的第二个RTT,降低发送速率以排空队列。在该带宽探测阶段的剩下的剩余周期,则按实时传输速率确定发送速率。本发明能兼顾传输效率、收敛性和公平性。(The invention discloses a sending rate adjusting method and a congestion control algorithm in a bandwidth detection stage. And judging the current network congestion state according to whether the ACK packets of all the data packets sent in the round are received within a certain time. And when the current network condition is better, adjusting the speed increasing factor of the sending speed according to the current round-trip delay and the historical maximum transmission speed. When the current network condition is poor, the sending rate is reduced to send a round of data packets, namely the sending rate is reduced to drain the queue at the second RTT of the bandwidth detection stage. The remaining period of the bandwidth sounding phase is followed by determining the transmission rate at the real-time transmission rate. The invention can give consideration to transmission efficiency, convergence and fairness.)

1. a sending rate adjusting method in a bandwidth detection stage is characterized by comprising the following steps:

step 1, setting the current Round number Round to 0;

Step 2, judging whether the value of Round% cp is 0 or not, wherein cp is the number of RTTs contained in each bandwidth detection stage;

If not, let the sending rate R_s＝R_rThen at a sending rate R_sSending a Round of data packets, updating Round +1, and returning to the step 2; wherein R is_rIs the real-time transmission rate;

if so, update R_s＝g×R_rWherein g is a rate-increasing factor, g>1; then at a sending rate R_sSending a Round of data packets, and updating Round + 1; after the data packet is sent in the current round, waiting for a certain time, and then entering the step 3;

And step 3: judging whether the ACK of all the data packets sent in the current round is received;

If yes, entering step 4;

if not, updating R_s＝(2-g)×R_rThen at a sending rate R_sSending a Round of data packets, updating Round +1, and returning to the step 2;

And 4, step 4: determining a transmission rate R_sWhether or not greater than historical maximum transmission rate C_max(ii) a If yes, making the speed increasing factor g equal to g_minTurning to the step 2; otherwise, turning to the step 5; wherein g is_minIs the minimum value of the speed increasing factor;

And 5: judging the current round trip time RTT_cWhether greater than or equal to KxRTT_minIf yes, let the speed increasing factor g equal to g_minTurning to the step 2; otherwise, turning to step 6;Where K is the delay threshold, RTT_minIs the minimum round-trip delay;

Step 6: judging whether the acceleration factor g is less than g_maxIf yes, changing g to g + gain, and turning to the step 2; otherwise, changing g to g-gain, and turning to the step 2; wherein g is_maxis the maximum value of the speed increasing factor, and gain is the increasing factor of the speed increasing factor;

The time length of each round of data packet transmission of the transmitting end is one RTT.

2. A congestion control algorithm is characterized by comprising a slow start stage, an emptying stage and a bandwidth detection stage; the method of claim 1 is adopted by the sending end in the bandwidth detection stage to adjust the sending rate of the bandwidth detection stage.

3. The congestion control algorithm of claim 2, wherein the sender delays the current round trip time RTT first every time an ACK packet of a data packet is received during the data packet sending process_cUpdating the difference between the time of receiving the ACK packet and the corresponding data packet sending time; then according to the formula RTT_min＝min(RTT_min，RTT_c) Updating minimum round trip time RTT_minWhere min (. cndot.) represents taking the minimum value.

4. The congestion control algorithm of claim 3, wherein the sender updates the minimum round trip delay RTT_minThe real-time transmission rate R is then also updated according to the following formula_r：

Wherein, delaviryRate_tIs the ratio of the number of ACK packets received in the RTT before T time to the RTT size, max (-) represents the maximum value, W is a time window, T is [ T-W, T-]at any time within the interval, T is the current time.

5. Congestion control algorithm according to claim 4Method, characterized by a delaveryRate_tIs the ratio of the number of ACK packets received in the RTT before the time t to the RTT, wherein the RTT value is the RTT at the time t_minThe value is obtained.

6. The congestion control algorithm of claim 3, wherein the time duration of each round of sending a data packet by the sender is one RTT, and the RTT value is the RTT at the beginning of sending the round of data packet_min。

7. The congestion control algorithm according to any of claims 2 to 6, wherein the slow start phase comprises the steps of: sending end according to sending rate R_sSending a round of data packets; after the data packets in the current round are sent, waiting for a certain time, and then judging whether ACK packets of all the data packets sent in the current round are received or not; if so, updating the sending rate R_s＝R_sX p, continuing the steps; if not, then enter the evacuation phase.

8. the congestion control algorithm of claim 6, wherein after a round of data packet transmission is completed, the waiting time is equal to the RTT of the round of data packet transmission completed before determining whether the ACK of all data packets transmitted in the round is received_minThe value is obtained.

9. The congestion control algorithm of claim 7 wherein the sender updates the real-time transmission rate R_rThereafter, the historical maximum transmission rate C is also updated according to the following formula_max：

C_max＝max(C_max，R_r)。

10. The congestion control algorithm of claim 9, wherein the evacuation phase comprises the steps of: sending end updating sending rate R_s＝R_sP, then at the sending rate R_sSending a round of data packets; and determines the transmission rate R_sWhether is less than or equal toIs equal to C_max(ii) a If yes, entering a bandwidth detection stage after the data packet of the round is sent; otherwise, after the data packet in the current round is sent, the steps are continued.

Technical Field

the invention relates to a sending rate adjusting method and a congestion control algorithm in a bandwidth detection stage.

Background

In recent years, with the tremendous growth in the number of users and the bandwidth of networks, the emergence of new networks and computing technologies, new business applications and business models has placed new demands on the performance of the internet. Network technology is continuously developed, and a network support platform is gradually converted into multiple services from a single service; network data transmission changes the flow model of the network fundamentally from traditional simple data to the application modes of emerging cloud computing, online video, mobile internet and the like which are developed rapidly at present; as globalization progresses, communication across intercontinental networks is more frequent. The development trend of the next generation internet is higher bandwidth, larger delay and more traffic, which all put new demands on the existing transport protocols. Therefore, the research of the next generation internet is receiving more and more attention, and becomes a hot spot in the field of computer network research.

Most of the traditional transmission protocols adopt a TCP data transmission protocol based on packet loss, which is a reliable data transmission protocol for end-to-end best effort delivery. When network resources are unable to meet user demands, conventional TCP protocols try to provide services to users until network utilization is extremely inefficient and congestion occurs. When congestion occurs, due to the shortage of queue cache caused by the increase of the queuing length of the data packet in the router, if measures cannot be taken in time, the packet loss rate of the data packet is increased, the throughput is reduced, the network delay is increased, and the resource utilization rate is reduced. If the congestion is more serious, the congestion of the network is broken down. However, the existing congestion control is mainly based on the congestion control method of packet loss, so people often find that the speed of the network does not achieve the effect expected in design. Congestion control based on packet loss continuously occupies a cache from the connection start stage, causing more packet loss, but when packet loss is found, the transmission speed is immediately reduced to be very low, causing great waste of network resources, and causing low transmission efficiency.

This problem is alleviated in the BBR congestion control algorithm. The BBR congestion control algorithm proposes a new model to account for congestion: the BBR does not use packet loss as a congestion signal, but calculates the window size by using a method of detecting available bandwidth. Many experiments prove that the BBR has excellent performance on a wide area network, and can greatly improve the transmission speed. However, BBR also has some problems, such as slow convergence rate of throughput rate when multiple BBR streams compete for the same bottleneck link; because the BBR empties the queue as much as possible, the BBR occupies little cache, and the BBR and other protocols have fairness problems; more importantly, the BBR detects bandwidth and delay alternately, but the BBR has a period of up to 6 RTTs after entering the stationary phase, and is in a period of not actively detecting bandwidth, where it cannot find out the spare bandwidth that may exist. In the active bandwidth probing phase, the fixed sending rate of the BBR also has the problem that the sending rate cannot be increased to the maximum available bandwidth quickly or the increase is too aggressive to increase the congestion of the whole network, thereby reducing the performance.

Therefore, it is an urgent problem to provide a new transmission protocol with transmission efficiency, convergence and fairness in consideration for the next generation internet.

Disclosure of Invention

in order to solve the defects of the prior art, the invention provides a sending rate adjusting method and a congestion control algorithm in a bandwidth detection stage, which can give consideration to transmission efficiency, convergence and fairness.

The technical scheme of the invention comprises the following steps:

A sending rate adjusting method in a bandwidth detection stage comprises the following steps:

Step 1, setting the current Round number Round to 0;

Step 2, judging whether the value of Round% cp is 0 or not, wherein cp is the number of RTTs contained in each bandwidth detection stage;

If not, let the sending rate R_s＝R_rThen at a sending rate R_sSending a Round of data packets, updating Round +1, and returning to the step 2; wherein R is_rIs the real-time transmission rate;

If so, update R_s＝g×R_r(i.e. increasing the transmission rate R by a multiplicative growth factor g_sIs g × R_rTo detect whether there is spare bandwidth in the current network), where g is the speed-up factor and g is the spare bandwidth>1; then at a sending rate R_sSending a Round of data packets, and updating Round + 1; after the data packet is sent in the current round, waiting for a certain time, and then entering the step 3;

And step 3: judging whether the ACK of all the data packets sent in the current round is received;

If yes, entering step 4;

If not, updating R_s＝(2-g)×R_r(i.e. sending rate R)_sReduced to (2-g). times.R_r) Then at a sending rate R_ssending a Round of data packets to drain the queue as completely as possible, updating Round +1, and returning to the step 2;

and 4, step 4: judgment of R_sWhether or not greater than historical maximum transmission rate C_max(ii) a If yes, making the speed increasing factor g equal to g_minturning to the step 2; otherwise, turning to the step 5; wherein g is_minIs the minimum value of the speed increasing factor;

And 5: judging the current round trip time RTT_cWhether greater than or equal to KxRTT_minIf yes, let the speed increasing factor g equal to g_minTurning to the step 2; otherwise, turning to step 6; where K is the delay threshold, RTT_minIs the minimum round-trip delay;

Step 6: judging whether the acceleration factor g is less than g_maxIf yes, changing g to g + gain, and turning to the step 2; otherwise, changing g to g-gain, and turning to the step 2; wherein g is_maxIs the maximum value of the speed increasing factor, and gain is the increasing factor of the speed increasing factor;

the time length of each round of data packet transmission of the transmitting end is one RTT.

The sending rate adjusting method in the bandwidth detection stage can be directly applied to the bandwidth detection stage in the BBR congestion control algorithm.

The invention also provides a congestion control algorithm, which comprises a slow start stage, an emptying stage and a bandwidth detection stage; the method of claim 1 is adopted by the sending end to adjust the sending rate in the bandwidth detection stage.

Further, in the process of sending data packets, the sending end firstly delays the current round trip time RTT (round trip time) every time an ACK packet (non-repeated ACK packet) of one data packet is received_cUpdating the difference between the time of receiving the ACK packet and the corresponding data packet sending time; then according to the formula RTT_min＝min(RTT_min，RTT_c) Updating minimum round trip time RTT_minWhere min (. cndot.) represents taking the minimum value.

Further, the sending endUpdating minimum round trip time RTT_minThe real-time transmission rate R is then also updated according to the following formula_r：

Wherein, delaviryRate_tis the ratio of the number of ACK packets received in a RTT before T time to the RTT size, max ((-) represents the maximum value, W is a time window, equal to 6-10 times RTT, T is [ T-W, T-]At any time in the interval, T is the current time, and the purpose of the formula is to take the maximum transmission rate over a period of time.

Further, DeliveryRate_tIs the ratio of the number of ACK packets received in the RTT before the time t to the RTT, wherein the RTT value is the RTT at the time t_minThe value is obtained.

Further, the time length of each round of data packet transmission by the transmitting end is one RTT, and the value of the RTT is the RTT when the round of data packet transmission starts_min。

further, the slow start phase comprises the steps of: sending end updating sending rate R_s＝R_sX p at the transmission rate R_sSending a round of data packets; after the data packets in the current round are sent, waiting for a certain time, and then judging whether ACK packets of all the data packets sent in the current round are received or not; if yes, continuing to perform the steps; if not, then enter the evacuation phase.

Further, after a round of data packet transmission is finished, before judging whether the ACK of all data packets transmitted in the round is received, the waiting time length is equal to the RTT of the round of data packet transmission when the round of data packet transmission is finished_minthe value is obtained.

further, the sending end updates the real-time transmission rate R_rThereafter, the historical maximum transmission rate C is also updated according to the following formula_max：

C_max＝max(C_max，R_r)。

Further, the emptying phase comprises the steps of: sending end updating sending rate R_s＝R_s/p，Then at a sending rate R_sSending a round of data packets; judging the sending rate R of the round_swhether or not it is less than or equal to C_max(ii) a If yes, entering a bandwidth detection stage after the data packet of the round is sent; otherwise, after the data packet in the current round is sent, the steps are continued.

Has the advantages that:

in the first RTT of each bandwidth detection stage, the transmission rate is increased by the speed increasing factor g, and then a round of data packets are transmitted to perform bandwidth detection. And when an ACK packet is received, measuring the current round-trip delay and the real-time transmission rate according to the information in the ACK packet, and recording the historical maximum transmission rate. And judging the current network congestion state according to whether the ACK packets of all the data packets sent in the round are received within a certain time. When the current network condition is good, the speed increasing factor of the sending rate is adjusted according to the current round-trip delay and the historical maximum transmission rate, which specifically comprises the following steps: when the sending rate is larger than the historical maximum real-time transmission rate or the current round-trip delay exceeds a set threshold (KxRTT)_min) While, the speed increasing factor g is adjusted to the minimum value g_minTo reduce the increase in transmission rate; if the sending rate is not greater than the historical maximum transmission rate and the current round-trip delay is not greater than the set threshold, then the speed-increasing factor g is less than the maximum value g according to whether the speed-increasing factor g is less than the maximum value g_maxTo increase or decrease the increase factor to increase or decrease the amplification of the transmission rate (if less than g)_maxThe value of g is increased by gain, otherwise it is deemed to have been sent at a high rate and decreased by gain to ensure that link congestion conditions are not exacerbated). When the current network condition is poor, the sending rate is reduced to send a round of data packets, namely the sending rate is reduced to drain the queue at the second RTT of the bandwidth detection stage. The remaining cp-1 (or cp-2) RTTs in the bandwidth sounding phase determine the sending rate at the real-time transmission rate. After cp RTTs of each bandwidth probing phase pass (Round% cp ═ 0), the next bandwidth probing phase is entered, and the above procedure is repeated.

The invention improves the transmission efficiency, improves the bandwidth utilization rate, and reduces the queuing delay, the average round-trip delay and the packet loss rate; meanwhile, the bandwidth occupied by each flow can be converged to the fair share as soon as possible, and the convergence and fairness of the flows are improved compared with the existing wide area network congestion control algorithm.

drawings

Fig. 1 is a flow chart of a congestion control algorithm in an embodiment of the present invention.

Fig. 2 is a diagram of the validity verification result of the congestion control algorithm (named BBR-DYNAMIC) according to the present invention.

FIG. 2(a) is a graph comparing the throughput of the BBR-DYNAMIC protocol of the present invention (BBR-DYNAMIC) with that of the BBR protocol in a 100Mbps bandwidth environment with 50ms delay.

FIG. 2(b) is a graph comparing the throughput of the BBR-DYNAMIC protocol of the present invention (BBR-DYNAMIC) with that of the BBR protocol in the environment of 100ms delay and 100Mbps bandwidth.

Fig. 3 is a test scenario network topology diagram.

Fig. 4 is a protocol packet loss indicator test chart, wherein the congestion control algorithm of the present invention is named BBR DYNAMIC.

Fig. 4(a) is a comparison chart of packet loss rate statistics when the basic delay of the bottleneck link is 50ms and the bandwidth is 100 Mbps.

Fig. 4(b) is a comparison chart of packet loss rate statistics when the basic delay of the bottleneck link is 100ms and the bandwidth is 100 Mbps.

Fig. 4(c) is a comparison chart of packet loss rate statistics when the basic delay of the bottleneck link is 150ms and the bandwidth is 100 Mbps.

Fig. 4(d) is a comparison chart of packet loss rate statistics when the basic delay of the bottleneck link is 200ms and the bandwidth is 100 Mbps.

Fig. 5 is a protocol average round trip delay indicator test chart, wherein the congestion control algorithm of the present invention is named BBRDYNAMIC.

Fig. 5(a) is a comparison graph of the average round-trip delay for a bottleneck link with a base delay of 50ms and a bandwidth of 100 Mbps.

Fig. 5(b) is a comparison graph of the average round-trip delay when the bottleneck link base delay is 100ms and the bandwidth is 100 Mbps.

Fig. 5(c) is a comparison graph of the average round-trip delay for a bottleneck link base delay of 150ms and a bandwidth of 100 Mbps.

fig. 5(d) is a comparison graph of the average round-trip delay for a bottleneck link with a base delay of 200ms and a bandwidth of 100 Mbps.

FIG. 6 is a test chart of the throughput improvement indicator of the protocol, wherein the congestion control algorithm of the present invention is named BBR-DYNAMIC.

Detailed Description

The invention will be further described with reference to the accompanying drawings.

referring to fig. 1, fig. 1 is a flow chart of an embodiment of the present invention.

In this embodiment, the congestion control algorithm process is as follows:

First, parameter initialization

The parameter initialization comprises: initializing real-time transmission rate R based on experience_rSending rate R_sHistorical maximum transmission rate C_maxMinimum round trip delay RTT_minCurrent round trip delay RTT_cAnd a speed-increasing factor g; setting the multiplicative growth coefficient p (p) according to experience>1) RTT number cp, delay threshold K and maximum value g of speed increasing factor contained in bandwidth detection stage_maxMinimum value g of acceleration factor_minAnd a speed-increasing factor amplification gain; the sending rate is the rate of sending data packets by the sending end, and the real-time transmission rate is the actual transmission rate of the data packets in the network and may be less than the sending rate;

in this embodiment, the real-time transmission rate R is initialized_rand a transmission rate R_sall are 1, historical maximum real-time transmission rate C_maxthe unit of the three is 0, and the unit of the three is a data packet/s; setting the multiplicative growth coefficient p to 2/ln 2; setting the number cp of RTT (round trip time) contained in a bandwidth detection stage as 8; setting a delay threshold K to be 3/2; initializing a speed increasing factor g to be 1; setting the maximum value g of the speed increasing factor_max1.5, minimum value of acceleration factor g_min1.1, 0.1 for increasing gain of speed-increasing factor;

Minimum round trip delay RTT_minAnd current round trip delay RTT_cThe initial value of (c) can be determined as follows: firstly, a data packet is sent, round trip delay is calculated according to the time of sending the data packet and the time of receiving ACK of the data packet, and the round trip delay is taken as the minimum round trip delay RTT_minAnd current round trip delay RTT_can initial value of (1);

Second, slow start phase

2.1) the transmitting end updates the transmission rate R_s＝R_sX p at the transmission rate R_ssending a round of data packets; after the data packet is sent in the current round, waiting for RTT_minThen step 2.2);

2.2) the sending end judges whether the ACK packets of all the data packets sent in the current round are received or not, and judges the current network state according to whether the ACK packets of all the data packets are received or not;

If yes, returning to the step 2.1);

The above process is started slowly in a manner similar to TCP, and with the minimum round trip delay RTT_minFor the rate adjustment period, the transmission rate R is increased by taking p as a multiplicative growth coefficient of the transmission rate_sIs R_s×p；

If not, entering an emptying stage;

Third, evacuation phase

3.1) the transmitting end updates the transmission rate R_s＝R_sP, then at the sending rate R_sSending a round of data packets;

3.2) judging the sending rate R of the current round_sWhether or not it is less than or equal to C_max(ii) a If yes, entering a bandwidth detection stage after the data packet of the round is sent; otherwise, after the data packet of the round is sent, turning to the step 3.2);

The above procedure is the round trip delay RTT_minfor the rate adjustment period, 1/p is used as a multiplicative reduction coefficient of the transmission rate, and the transmission rate R is reduced_sis R_sP to drain the queue;

Fourth, bandwidth detection phase

Step 1, setting the current Round number Round to 0;

step 2, the sending end judges whether the value of Round% cp is 0;

If not, let R_s＝R_rAt a sending rate R_sSending a Round of data packets, updating Round +1, and returning to the step 2;

If so, update R_s＝g×R_rThen at a sending rate R_ssending a Round of data packets, and updating Round + 1; after the data packet is sent in the current round, waiting for RTT_minThen step 3 is carried out;

And step 3: judging whether the ACK of all the data packets is received;

If yes, entering step 4;

If not, updating R_s＝(2-g)×R_rThen at a sending rate R_sSending a Round of data packets to drain the queue as completely as possible, updating Round +1, and returning to the step 2;

And 4, step 4: the sending end judges the sending rate R_sWhether or not greater than historical maximum transmission rate C_max(ii) a If yes, making the speed increasing factor g equal to g_minturning to the step 2; otherwise, turning to the step 5;

And 5: the sending end judges the current round trip time RTT_cWhether greater than or equal to KxRTT_minIf yes, let the speed increasing factor g equal to g_minTurning to the step 2; otherwise, turning to step 6;

Step 6: if the acceleration factor g is smaller than g_maxIf so, changing g to g + gain, and turning to the step 2; otherwise, changing g to g-gain, and turning to the step 2;

The time length of each round of data packet transmission of the transmitting end is one RTT, and the value of the RTT is the RTT when the round of data packet transmission starts_min。

And (4) combining the steps 4-6, adjusting the acceleration factor g in the bandwidth detection stage by the following formula:

When a sending end receives an ACK packet (non-repeated ACK packet) of a data packet in the process of sending the data packet, firstly, the difference between the time of receiving the ACK packet and the sending time of the corresponding data packet is calculated and used as the current round trip delay RTT_c(ii) a Then according to RTT_min＝min(RTT_min，RTT_c) Updating minimum round trip time RTT_minwhere min (-) denotes taking the minimum value; then, the following formula is usedNew real time transmission rate R_r：

Wherein, delaviryRate_tIs one RTT before time t_minNumber of ACK packets received within time and RTT_minThe ratio of the magnitudes, max (·) represents the maximum value, W is a time window equal to 6-10 times the RTT_mint is [ T-W, T]At any time in the interval, T is the current time, and the purpose of the formula is to take the maximum transmission rate over a period of time.

Further, the sending end updates the real-time transmission rate R_rThereafter, the historical maximum transmission rate C is also updated according to the following formula_max：

C_max＝max(C_max，R_r)。

fig. 2 is a diagram of a protocol validity verification experiment result, which aims to verify that a congestion control algorithm BBR-DYNAMIC in the present invention can effectively perform DYNAMIC sending rate adjustment in a bandwidth detection phase with respect to a BBR, thereby ensuring transmission efficiency. The experimental environment was as follows: the 1 machine and 1 server are connected to a switch (configured WANem WAN Environment simulator) with 1000KB cache. The connection to send data is established using the Iperf tool for a total of 30 seconds. The experiment is respectively tested by BBR and BBR-DYNAMIC, the bandwidth and the time delay of the bottleneck link are changed, and the change situation of the throughput rate of the bottleneck link is observed.

As can be seen from FIG. 2(a), under the experimental environment with 50ms delay and 100Mbps bandwidth, 80Mbps UDP stream is set to be sent every 8 seconds and then suspended for 2 seconds, the BBR-DYNAMIC can be increased to the maximum available bandwidth more quickly when the UDP stream is suspended, i.e. the maximum available bandwidth is converged to the fair share, the whole bandwidth utilization rate is improved by 22.67%, and the change curve of the throughput rate is more stable. This is because of the mechanism of dynamically adjusting the transmission rate of the BBR-DYNAMIC, and when there is spare bandwidth, the BBR-DYNAMIC can increase the value of the speed-increasing factor g to increase the transmission rate and seize more bandwidth. When the UDP flow is withdrawn, BBR-DYNAMIC can grow to the actual bandwidth faster than BBR, and the waste of bandwidth resources can not be caused.

in FIG. 2(b), under the experimental environment with 100ms delay and 100Mbps bandwidth, the UDP stream of 80Mbps suspended for 1 second is set to be sent every 9 seconds, the BBR-DYNAMIC of the present invention can increase to the maximum available bandwidth more quickly, and the effect is more obvious than that in FIG. 2 (a). This is because the feedback received becomes slower when the delay becomes higher, so the BBR-DYNAMIC of the present invention, which employs a relatively aggressive speedup strategy when bandwidth is found to be free, will increase to the actual bandwidth more quickly than the BBR employing a fixed speedup strategy.

Fig. 3 is a network topology used in the experiment, which is specifically as follows: the whole network consists of three parts of a sending end, a switch and a receiving end, and is connected by links with four kinds of delays and two kinds of bandwidths in total. In the experiment, 1, 3, 5 and 10 streams are simultaneously sent at the sending end respectively, and the change conditions of the packet loss, the time delay and the throughput rate of a link are tested.

fig. 4 is a protocol packet loss indicator test chart, which is used to show the performance of the BBR DYNAMIC in reducing link packet loss, and illustrates that the BBR DYNAMIC of the present invention has better fairness and convergence than the BBR protocol.

Fig. 4(a) shows a comparison chart of packet loss statistics when the basic delay of the bottleneck link is 50ms and the bandwidth is 100 Mbps. The BBR DYNAMIC judges whether a queuing phenomenon exists according to the minimum round-trip delay of whether the current round-trip delay exceeds a delay threshold K times, if so, the BBR DYNAMIC transmits at a reduced speed, and more queues are not caused to avoid packet loss. It can be seen that the packet loss rates of the two protocols increase with the increase of the number of streams, but the packet loss rate of the BBR DYNAMIC of the present invention is always lower than that of the BBR protocol, which indicates that the BBR DYNAMIC can reduce the packet loss rate on the bottleneck link.

Fig. 4(b) shows a comparison chart of packet loss statistics when the basic delay of the bottleneck link is 100ms and the bandwidth is 100 Mbps. It can be seen that, due to uncertainty of packet loss, the difference between packet loss rates of the two protocols is small, but the packet loss performance of the BBRDYNAMIC of the present invention is always better than that of the BBR protocol.

fig. 4(c) shows a comparison chart of the packet loss rate statistics when the bottleneck link has a basic delay of 150ms and a bandwidth of 100 Mbps. It can be seen that as the delay increases, the packet loss rates of the two protocols increase to different degrees, but the packet loss rates of the BBRDYNAMIC of the present invention are always lower than that of the BBR protocol.

Fig. 4(d) shows a comparison chart of packet loss statistics when the basic delay of the bottleneck link is 200ms and the bandwidth is 100 Mbps. It can be seen from these four figures that the performance of the BBR DYNAMIC of the present invention is good in reducing the link packet loss problem.

FIG. 5 is a test chart of the average round-trip delay index of the protocol, which is used to show the performance of BBR DYNAMIC in reducing the average round-trip delay problem of the link according to the present invention, and shows that BBR DYNAMIC of the present invention has better fairness and convergence than BBR protocol. With a bandwidth of 100 Mbps. Fig. 5(a) - (d) are graphs comparing the average round trip delay of the bottleneck link base delay of 50ms, 100ms, 150ms, and 200ms, respectively.

The BBR DYNAMIC judges whether a queuing phenomenon exists according to the minimum round-trip delay of whether the current round-trip delay exceeds a delay threshold K times, if so, the BBR DYNAMIC sends the current round-trip delay at a reduced speed, and more queues are not caused so as to reduce the round-trip delay on a link. The comparison of the average round-trip delay of the bottleneck link of the two protocols in the environment of 4 different basic delays can be intuitively seen, the average delay of the bottleneck link of the BBR DYNAMIC is lower than the average delay of the bottleneck link of the BBR, which shows that the BBR DYNAMIC can actually reduce the queuing of the link, thereby reducing the average round-trip delay.

Fig. 6 is a test chart of the throughput improvement index of the protocol. The experiment tests the improvement percentage of the throughput rate of BBR-DYNAMIC compared with BBR under the conditions of 100Mbps bandwidth and 1, 3, 5 and 10 different streams under four delays of 50ms, 100ms, 150ms and 200 ms. It can be seen that the BBR-DYNAMIC of the invention has more obvious effect of improving the throughput rate under the condition of high delay. When the bandwidth is found to be vacant, the BBR-DYNAMIC adopting a relatively aggressive speed increasing strategy can be increased to the actual bandwidth more quickly than the BBR adopting a fixed speed increasing strategy, so that the improvement of the throughput of the BBR-DYNAMIC is more obvious under the condition of high delay. And under the condition that a plurality of streams compete for bottleneck bandwidth simultaneously, the spare bandwidth can be quickly occupied by the plurality of streams, at the moment, the improvement effect of the throughput is not as obvious as that of a single stream, and the improvement of the throughput is lower and lower along with the increase of the number of the streams. However, it can be seen from the figure that the BBR-DYNAMIC of the present invention can also effectively improve the throughput even in the case of a plurality of streams.