CN1324862C - Method for regulating congest window in communication network - Google Patents

Abstract

The invention relates to a method for adjusting a congestion window in a communication network. The method is to calculate the length of the buffer queue of the switching node, and then determine the cause of the packet loss in the network according to the length of the buffer queue, and adopt different control strategies for different reasons to control the congestion window and the start-up threshold of the switching node Adjust accordingly. The present invention not only retains the above-mentioned advantages of TCP Reno in dealing with network congestion, but also realizes the use of bandwidth estimation to set appropriate thresholds and congestion control windows for packet loss caused by link characteristics, so that network bandwidth can be effectively The use of the system improves the throughput of the system. In addition, the present invention also calculates the buffer queue length at the switch node, and judges the network congestion situation through the queue length, so as to reduce the queuing waiting time while processing packet loss in time.

Description

Method of adjusting congestion window in communication network

technical field

The invention relates to the technical field of network communication, in particular to a method for adjusting a congestion window at a transport layer in a communication network.

Background technique

The Internet (Internet) is a network that provides best-effort services, and TCP (Transport Control Protocol, Transmission Control Protocol) is a reliable transport layer protocol on it, which is widely used in email, file transfer, audio and video data Streaming and Web (World Wide Web) and other applications. According to statistics, TCP data traffic accounts for more than 95% of the Internet, so the TCP protocol is a dominant end-to-end transmission protocol in the Internet. The reliability and stability of the Internet today are inseparable from the congestion control mechanism of the TCP protocol. .

At present, the most common TCP implementation on the Internet is TCP Reno. TCP Reno is an end-to-end transport layer protocol that adopts the AIMD (Additive Increase Multiplicative Decrease) window adjustment mechanism.

In TCP Reno, the actual sending window size of the sender is: MIN(RCVWND, CWDN+NDUP), that is, the smaller value of "RCVWND" and "CWDN+NDUP" is used as the actual sending window size; where RCVWND refers to the receiver Notification window, CWDN refers to the congestion window of the sending end, NDUP is equal to 0 before the number of repeated ACKs (acknowledgments) received by the sending end is less than tcprexmtthresh (tcprexmtthresh is the number of repeated ACKs that enable TCP to start the fast retransmission mechanism), and then it is The number of duplicate ACKs received. If flow control is not considered, the sending window is equal to CWDN+NDUP.

Reno's congestion control strategy is described as follows:

① If a non-repeated ACK message is received, then:

if w<ssthresh, set w=w+1; if the current congestion window w is smaller than the slow start threshold ssthresh, each time an ACK is received, the window will be increased by the size of a TCP segment, and this process is called slow start process;

否则，如果当前的拥塞窗口大于慢启动门限值而小于通知窗口，每收到一个ACK，就将窗口增加当前窗口的倒数分之一，且这一过程称之为拥塞避免过程；

Otherwise, if the current congestion window is greater than the slow-start threshold and smaller than the notification window, each time an ACK is received, the window will be increased by one last of the current window, and this process is called congestion avoidance process;

② If repeated ACK messages are received, and the number of received ACK messages > tcprexmtthresh, then:

retranmit "next expected" segment; retransmit the message segment that needs to be retransmitted as described in the repeated ACK confirmation information, that is, the packet;

同时令慢启动门限值等于当前窗口的一半；then set w＝ssthresh；拥塞窗口等于更新的慢启动门限值；

Make the slow start threshold equal to half of the current window at the same time; then set w=ssthresh; the congestion window is equal to the updated slow start threshold;

③ If the retransmission timer times out, enter SlowStart (slow start):

The slow start threshold is equal to half of the current window;

set w=1; the congestion window is equal to one segment;

However, the aforementioned TCP Reno, which was originally designed for wired environments, does not perform satisfactorily on wireless links with bursty errors. TCP Reno believes that all packet loss is due to congestion, so random packet loss caused by changes in link characteristics is treated as congestion loss, and the window is directly reduced to half of the window before packet loss. This halving of the window is often unnecessary, resulting in a decrease in system performance and a waste of network resources.

Contents of the invention

In view of the above-mentioned problems in the prior art, the purpose of the present invention is to provide a method for adjusting the congestion window in a communication network, which can distinguish whether the cause of packet loss is random loss or congestion loss, and determine the target for packet loss according to the cause of packet generation. The resizing strategy for the window.

The purpose of the present invention is achieved through the following technical solutions:

A method for adjusting a congestion window in a communication network, comprising:

A. Calculate the length of the cache queue of the switching node in the network;

B. Determine the cause of packet loss in the network according to the length of the buffer queue, and adjust the congestion window of the switching node by using corresponding control strategies for different reasons.

Described step A comprises:

A1. Determine the bandwidth of the network according to the packets transmitted in the network;

A2. Calculate the length of the cache queue at the switching node according to the bandwidth of the network;

Described step A1 comprises:

The bandwidth of the network is determined according to the length of the message confirmed by the ACK (confirmation) message received by the sending node in the network and the interval between two adjacent ACK messages.

Described step A1 further comprises:

The length of the message confirmed by the ACK (confirmation) message received by the sending node is divided by the interval time between the ACK message and the previous ACK message, and the obtained result is used as the sampling value of the network bandwidth based on the ACK message ;

Smooth processing is performed on the obtained multiple network bandwidth sampling values to obtain the network bandwidth.

Described step A2 comprises:

Calculate the length of the buffer queue at the switching node according to the obtained network bandwidth value, the size of the packet, the current expected network bandwidth and the minimum round-trip time between the sending node and the receiving node.

Described step A1 specifically comprises:

The network bandwidth sampling value sample_BWE[k] is:

sample_BWE[k]=(ACKed*seg_size*8)/now～lastACKtime,

Among them, ACKed indicates the number of message segments confirmed by the ACK received this time, seg_size indicates the bytes contained in the message segment, now~lastACKtime indicates the time interval between receiving ACK this time and receiving ACK last time, and K is The serial number of the message confirmed by the ACK received this time;

Smooth the sampling value of the network bandwidth, estimate and determine the network bandwidth BWE[k] as:

$\mathrm{<span\; class="notranslate">\; BWE</span>}<span\; class="notranslate">\; [</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; ]</span><span\; class="notranslate">\; =</span><span\; class="notranslate">\; (</span>\frac{\mathrm{<span\; class="notranslate">\; 19</span>}}{\mathrm{<span\; class="notranslate">\; twenty\; one</span>}}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; *</span>\mathrm{<span\; class="notranslate">\; BWE</span>}<span\; class="notranslate">\; [</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; ]</span><span\; class="notranslate">\; +</span><span\; class="notranslate">\; (</span>\frac{\mathrm{<span\; class="notranslate">\; 1</span>}}{\mathrm{<span\; class="notranslate">\; twenty\; one</span>}}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; *</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; sample</span>}<span\; class="notranslate">\; \_</span>\mathrm{<span\; class="notranslate">\; BWE</span>}<span\; class="notranslate">\; [</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; ]</span><span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; sample</span>}<span\; class="notranslate">\; \_</span>\mathrm{<span\; class="notranslate">\; BWE</span>}<span\; class="notranslate">\; [</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; ]</span><span\; class="notranslate">\; )</span><span\; class="notranslate">\; ;</span>$

And, the step A2 specifically includes:

Buffer queue length at switch node N is:

$\stackrel{<span\; class="notranslate">\; \&OverBar;</span>}{\mathrm{<span\; class="notranslate">\; N</span>}}<span\; class="notranslate">\; =</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; Expected</span>}<span\; class="notranslate">\; -</span>\frac{\mathrm{<span\; class="notranslate">\; BWE</span>}}{\mathrm{<span\; class="notranslate">\; seg</span>}<span\; class="notranslate">\; \_</span>\mathrm{<span\; class="notranslate">\; size</span>}}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; *</span>\mathrm{<span\; class="notranslate">\; BaseRTT</span>}<span\; class="notranslate">\; ,</span>$

in, $\frac{\mathrm{cwnd}}{\mathrm{BaseRtt}}$ is the expected rate at the current moment, cwnd indicates the size of the congestion window, BaseRtt is the minimum round-trip time of messages between switching nodes, seg_size is the size of the message segment as mentioned above, and BWE is the current network bandwidth obtained in the above step A1 Estimated value BWE[k].

Described step A comprises:

Calculate the buffer queue length of the switching node according to the expected rate of the current network, the actual rate and the minimum round-trip time of packets between the switching nodes.

Described step A further comprises:

Calculate the expected rate Expected and the actual rate Actual:

$\mathrm{<span\; class="notranslate">\; Expected</span>}<span\; class="notranslate">\; =</span>\frac{\mathrm{<span\; class="notranslate">\; cwnd</span>}}{\mathrm{<span\; class="notranslate">\; BaseRtt</span>}}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; Actual</span>}<span\; class="notranslate">\; =</span>\frac{\mathrm{<span\; class="notranslate">\; cwnd</span>}}{\mathrm{<span\; class="notranslate">\; Rtt</span>}}<span\; class="notranslate">\; ;</span>$

Among them, cwnd represents the congestion window size, BaseRtt represents the minimum round-trip time of messages between switching nodes, and RTT represents the actual round-trip time of messages between switching nodes;

Calculate and determine the buffer queue length N of the switching node as:

N=Actual*(RTT-BaseRTT)=Diff*BaseRTT,

Wherein, Diff=Expected-Actual, which is the difference between the expected rate Expected and the actual rate Actual.

Described step B comprises:

B1. If the length of the buffer queue is greater than the set queue length threshold, it is determined that the network bandwidth is saturated, and the packet loss generated in the network is caused by congestion, and the switching node is performed according to the corresponding congestion control strategy Adjustment of window and activation threshold;

B2. If the length of the buffer queue is less than the set queue length threshold, it is determined that the network bandwidth is not saturated, and the packet loss generated in the network is caused by a change in link characteristics, and according to the change based on link characteristics The control strategy adjusts the switching node's window and startup threshold.

In the method for adjusting the congestion window in the communication network, the adjustment of the window of the switching node and the start-up threshold according to the control strategy based on the link characteristic change described in step B2 is:

If three repeated ACK messages are received, adjust the slow start threshold equal to the ratio of the product of the estimated bandwidth and the minimum round-trip time to the size of the TCP segment, and if the current congestion window is greater than the slow start threshold, set the congestion window Equal to the slow start threshold;

If the retransmission timer expires, then make the congestion window equal to 1, make the slow start threshold equal to the ratio of the product of the estimated bandwidth and the minimum round-trip time to the size of the TCP segment, and if the slow start threshold is less than 2, then make the slow start threshold equal to 2.

When TCP Reno handles network congestion, by halving the congestion window, it can effectively reduce the cache length and deal with network congestion in a timely manner. The performance of this algorithm has been well proven in wired networks. It can be seen from the above-mentioned technical solution provided by the present invention that the present invention not only retains the above-mentioned advantages of TCP Reno when dealing with network congestion, but also realizes the use of bandwidth estimation to set appropriate The threshold and congestion control window can make effective use of network bandwidth and improve the throughput of the system.

In a word, the present invention can effectively distinguish the cause of packet loss to improve the throughput performance of TCP on the wireless link according to specific conditions, and can adaptively adjust the window according to the congestion status of the network to enhance the stability of the TCP protocol. In addition, the invention also calculates the buffer queue length at the switch node, and judges the network congestion situation through the queue length, so as to reduce the queue waiting time while processing the packet loss in time.

Description of drawings

Figure 1 is a flow chart of the method of the present invention.

Detailed ways

The core idea of the method of the present invention is by calculating the length of the cache queue of the switching node, then, according to the length of the cache queue, determine the cause of packet loss in the network, and adopt different control strategies for different reasons to the switching node. The congestion window and startup threshold are adjusted accordingly. On the one hand, the present invention can use the number of received packets confirmed by the returned ACK data flow and the interval time between two adjacent ACKs to estimate the bandwidth of the network, and then use the estimated bandwidth information to calculate the buffer queue at the switching node The size of the length N; on the other hand, the size of the buffer queue length N of the switching node can also be determined according to the expected rate and the actual rate value of the current network.

After the length N of the buffer queue of the corresponding switching node is obtained, the congestion status of the network can be judged according to the value of N. That is, when packet loss occurs, judge the cause of packet loss according to the congestion status of the network. If packet loss occurs and N is greater than the set queue length threshold, it is considered that packet loss is caused by network congestion; if packet loss occurs, N is less than the set queue length threshold, it is considered that the packet loss is due to the influence of the change of link characteristics. For packet loss caused by different reasons, different control strategies can be adopted to control the congestion window and start threshold adjustment.

For having further understanding to the present invention, now in conjunction with accompanying drawing, and take the length value of the buffer queue of calculating switching node according to network bandwidth as example to further illustrate the present invention, referring to Fig. 1, specifically comprise following process:

Step 1: Sampling the network bandwidth value sample_BWE[k];

sample_BWE[k]=(ACKed*seg_size*8)/now～lastACKtime,

Among them, ACKed indicates the number of message segments confirmed by the ACK received this time, seg_size indicates the bytes contained in the message segment, now~lastACKtime indicates the time interval between receiving ACK this time and receiving ACK last time, and K is The serial number of the message confirmed by the ACK received this time, the ACKed value and seg_size value can be obtained according to the characteristics of the network, and the now~lastACKtime values can be obtained based on the actual monitoring of the receiving time of the ACK message.

Step 2: smoothing the obtained network bandwidth sampling value to obtain the estimated current network bandwidth value BWE[k];

$\mathrm{<span\; class="notranslate">\; BWE</span>}<span\; class="notranslate">\; [</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; ]</span><span\; class="notranslate">\; =</span><span\; class="notranslate">\; (</span>\frac{\mathrm{<span\; class="notranslate">\; 19</span>}}{\mathrm{<span\; class="notranslate">\; twenty\; one</span>}}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; *</span>\mathrm{<span\; class="notranslate">\; BWE</span>}<span\; class="notranslate">\; [</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; ]</span><span\; class="notranslate">\; +</span><span\; class="notranslate">\; (</span>\frac{\mathrm{<span\; class="notranslate">\; 1</span>}}{\mathrm{<span\; class="notranslate">\; twenty\; one</span>}}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; *</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; sample</span>}<span\; class="notranslate">\; \_</span>\mathrm{<span\; class="notranslate">\; BWE</span>}<span\; class="notranslate">\; [</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; ]</span><span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; sample</span>}<span\; class="notranslate">\; \_</span>\mathrm{<span\; class="notranslate">\; BWE</span>}<span\; class="notranslate">\; [</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; ]</span><span\; class="notranslate">\; )</span><span\; class="notranslate">\; .</span>$

Step 3: Calculate the cache queue at the switching node according to the obtained bandwidth value of the current network;

$\stackrel{<span\; class="notranslate">\; \&OverBar;</span>}{\mathrm{<span\; class="notranslate">\; N</span>}}<span\; class="notranslate">\; =</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; Expected</span>}<span\; class="notranslate">\; -</span>\frac{\mathrm{<span\; class="notranslate">\; BWE</span>}}{\mathrm{<span\; class="notranslate">\; seg</span>}<span\; class="notranslate">\; \_</span>\mathrm{<span\; class="notranslate">\; size</span>}}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; *</span>\mathrm{<span\; class="notranslate">\; BaseRTT</span>}<span\; class="notranslate">\; ;</span>$

in $\frac{\mathrm{cwnd}}{\mathrm{BaseRtt}}$ is the expected rate at the current moment, cwnd indicates the size of the congestion window, BaseRtt is the minimum round-trip time of the packet sent by the switching node, seg_size is the packet size of the packet sent by the switching node, and BWE is the estimated value of the network bandwidth at the current moment BWE[ k], the cwnd value is a parameter that has been applied in the network transmission, and the present invention can directly acquire and apply it, and the BaseRtt value is obtained by tracking and comparing the RTT.

Step 4: When packet loss occurs in the network, judge whether the length value N of the buffer queue is greater than the set queue length threshold value β, if greater, then perform step 5, otherwise, perform step 6, the described The queue length threshold β is usually an empirical value, which can be obtained through network testing, and the value can be 3 for the current network.

Step 5: Determine the length value of the buffer queue N>β, at this time, it is considered that the network bandwidth is saturated, and the cause of packet loss is mainly due to congestion loss, and the congestion window and slow start threshold are adjusted according to the congestion control strategy in the TCP-Reno algorithm. The adjustment process is the same as described in the background art.

Step 6: Determine the length value of the cache queue N<β, at this time, it is considered that the network bandwidth is not saturated, the network is not congested, and the cause of packet loss is likely to be caused by changes in link characteristics. Effects usually include:

①Link error rate: The wireless host uses wireless or infrared wave transmission for communication, which is very susceptible to environmental interference. The BER (bit error rate) on the wireless network is 10 times or worse than that of the wired network. As long as there is a wireless link, it will affect the performance of the entire network;

②Bandwidth: Bandwidth is a scarce resource in wireless networks. Compared with typical wired Ethernet 10M to 1000M (or even higher), the bandwidth of wireless LAN is relatively small (2M), and the change is still large;

③ Mobility: The wireless host moves frequently during communication, which may easily lead to data loss or even temporary interruption of communication. TCP interprets this as congestion and triggers the congestion control mechanism, but this is unnecessary because once the movement stops, the wireless host starts receiving data again, which causes TCP performance to degrade.

For the packet loss caused by the change of link characteristics, it is necessary to adjust the congestion window and the slow start threshold according to the following control strategy based on the change of link characteristics. The specific processing process is as follows;

If the switching node receives three duplicate ACK packets, perform the following operations:

ssthresh=(bwe*BaseRTT)/(seg_size);/Make the slow start threshold equal to the ratio of the product of estimated network bandwidth and minimum round-trip time to the TCP segment size;

if(cwnd＞ssthresh);/If the current congestion window is greater than the slow start threshold;

cwnd=ssthresh;/make the congestion window equal to the slow start threshold;

At the same time, if the switching node receives repeated ACK messages, it needs to re-send the corresponding message and start the corresponding timer. When the timer expires, the following operations are performed:

cwnd=1;/make the congestion window equal to 1;

ssthresh=(bwe*BaseRTT)/(seg_size);/Make the slow start threshold equal to the ratio of the product of the estimated bandwidth and the minimum round-trip time to the TCP segment size;

if(ssthresh<2)

ssthresh=2; /If the slow start threshold is less than 2, set the slow start threshold equal to 2.

In the above process, steps 1 to 3 describe the acquisition process of the buffer queue length value of the switching node. In TCP Vegas, the buffer queue length value can also be obtained by the following method, but in order to ensure the subsequent congestion window And the realization of the start-up threshold value adjustment mechanism, the network bandwidth value in the current network also needs to be obtained while obtaining the buffer queue length value through the following method;

That is, the buffer queue length value of the switching node can also be obtained through the following calculation process:

First, calculate the current expected rate Expected and the actual rate Actual of the network, specifically: expected rate $\frac{\mathrm{cwnd}}{\mathrm{BaseRtt}},$ actual rate $\mathrm{Actual}=\frac{\mathrm{cwnd}}{\mathrm{Rtt}};$

Among them, cwnd represents the size of the congestion window, BaseRtt represents the minimum round-trip time of the message sent by the switching node, RTT represents the actual round-trip time of the message sent by the switching node, and RTT is obtained through actual monitoring;

Then calculate the difference Diff between the expected rate Expected and the actual rate Actual as: Diff=Expected-Actual;

Based on the length N of the cache queue of the switching node, the formula for calculating the RTT is:

RTT=BaseRTT+N/Actual;

According to the above formula, it can be concluded that the length N of the cache queue is:

N=Actual*(RTT-BaseRTT)=Diff*BaseRTT.

Certainly, in the present invention, other methods can also be used to obtain the value of the length N of the buffer queue of the switching node, and then, the congestion window and the start-up threshold can be adjusted according to the obtained N value to meet the needs of network transmission.

It can be seen from the above description process of the specific implementation of the present invention that the present invention not only retains the algorithmic advantages of TCP Reno when dealing with network congestion, for example, by halving the congestion window, it can effectively reduce the buffer queue length and timely process Network congestion conditions, etc. At the same time, it also effectively utilizes network bandwidth estimation to set appropriate thresholds and congestion control windows for packet loss caused by changes in link characteristics, so that network bandwidth can be effectively utilized and system throughput can be improved.

The above is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any person skilled in the art within the technical scope disclosed in the present invention can easily think of changes or Replacement should be covered within the protection scope of the present invention. Therefore, the protection scope of the present invention should be determined by the protection scope of the claims.

Claims (10)

1. A method for adjusting a congestion window in a communication network, characterized in that it comprises: A. Calculate the length of the cache queue of the switching node in the network; B. According to the length of the buffer queue, it is determined that the cause of packet loss in the network is congestion or a change in link characteristics, and corresponding control strategies are used to adjust the congestion window of the switching node for different reasons. 2. The method for adjusting the congestion window in a communication network according to claim 1, characterized in that said step A comprises: A1. Determine the bandwidth of the network according to the packets transmitted in the network; A2. Calculate the length of the cache queue at the switching node according to the bandwidth of the network. 3. The method for adjusting the congestion window in a communication network according to claim 2, characterized in that said step A1 comprises: The bandwidth of the network is determined according to the length of the message confirmed by the ACK message received by the sending node in the network and the interval between two adjacent ACK messages. 4. The method for adjusting the congestion window in a communication network according to claim 2 or 3, characterized in that said step A1 further comprises: The length of the message confirmed by the ACK message received by the sending node is divided by the interval time between the ACK message and the previous ACK message, and the obtained result is used as a sampling value based on the network bandwidth of the ACK message; Smooth processing is performed on the obtained multiple network bandwidth sampling values to obtain the network bandwidth. 5. The method for adjusting the congestion window in a communication network according to claim 4, characterized in that said step A2 comprises: Calculate the length of the buffer queue at the switching node according to the obtained network bandwidth value, the size of the packet, the current expected network bandwidth and the minimum round-trip time between the sending node and the receiving node. 6. The method for adjusting the congestion window in a communication network according to claim 5, characterized in that said step A1 specifically comprises: The network bandwidth sampling value sample_BWE[k] is: sample_BWE[k]=(ACKed*seg_size*8)/now～lastACKtime, Among them, ACKed indicates the number of message segments confirmed by the ACK received this time, seg_size indicates the bytes contained in the message segment, now~lastACKtime indicates the time interval between receiving ACK this time and receiving ACK last time, and K is The serial number of the message confirmed by the ACK received this time; Smooth the sampling value of the network bandwidth, estimate and determine the network bandwidth BWE[k] as: $\mathrm{<span\; class="notranslate">\; BWE</span>}<span\; class="notranslate">\; [</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; ]</span><span\; class="notranslate">\; =</span><span\; class="notranslate">\; (</span>\frac{\mathrm{<span\; class="notranslate">\; 19</span>}}{\mathrm{<span\; class="notranslate">\; twenty\; one</span>}}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; *</span>\mathrm{<span\; class="notranslate">\; BWE</span>}<span\; class="notranslate">\; [</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; ]</span><span\; class="notranslate">\; +</span><span\; class="notranslate">\; (</span>\frac{\mathrm{<span\; class="notranslate">\; 1</span>}}{\mathrm{<span\; class="notranslate">\; twenty\; one</span>}}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; *</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; sample</span>}<span\; class="notranslate">\; \_</span>\mathrm{<span\; class="notranslate">\; BWE</span>}<span\; class="notranslate">\; [</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; ]</span><span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; sample</span>}<span\; class="notranslate">\; \_</span>\mathrm{<span\; class="notranslate">\; BWE</span>}<span\; class="notranslate">\; [</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; ]</span><span\; class="notranslate">\; )</span><span\; class="notranslate">\; ;</span>$ And, the step A2 specifically includes: Buffer queue length at switch node N is: $\stackrel{<span\; class="notranslate">\; \&OverBar;</span>}{\mathrm{<span\; class="notranslate">\; N</span>}}<span\; class="notranslate">\; =</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; Expected</span>}<span\; class="notranslate">\; -</span>\frac{\mathrm{<span\; class="notranslate">\; BWE</span>}}{\mathrm{<span\; class="notranslate">\; seg</span>}<span\; class="notranslate">\; \_</span>\mathrm{<span\; class="notranslate">\; size</span>}}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; *</span>\mathrm{<span\; class="notranslate">\; BaseRTT</span>}<span\; class="notranslate">\; ,</span>$ in, $\mathrm{Expected}=\frac{\mathrm{cwnd}}{\mathrm{BaseRtt}}$ is the expected rate at the current moment, cwnd indicates the size of the congestion window, BaseRtt is the minimum round-trip time of messages between switching nodes, seg_size is the size of the message segment as mentioned above, and BWE is the current network bandwidth obtained in the above step A1 Estimated value BWE[k]. 7. The method for adjusting the congestion window in a communication network according to claim 1, characterized in that said step A comprises: Calculate the buffer queue length of the switching node according to the expected rate of the current network, the actual rate and the minimum round-trip time of packets between the switching nodes. 8. The method for adjusting the congestion window in a communication network according to claim 7, characterized in that said step A further comprises: Calculate the expected rate Expected and the actual rate Actual: $\mathrm{<span\; class="notranslate">\; Expected</span>}<span\; class="notranslate">\; =</span>\frac{\mathrm{<span\; class="notranslate">\; cwnd</span>}}{\mathrm{<span\; class="notranslate">\; BaseRtt</span>}}<span\; class="notranslate">\; ,</span>$ $\mathrm{<span\; class="notranslate">\; Actual</span>}<span\; class="notranslate">\; =</span>\frac{\mathrm{<span\; class="notranslate">\; cwnd</span>}}{\mathrm{<span\; class="notranslate">\; Rtt</span>}}<span\; class="notranslate">\; ;</span>$ Among them, cwnd represents the congestion window size, BaseRtt represents the minimum round-trip time of messages between switching nodes, and RTT represents the actual round-trip time of messages between switching nodes; Calculate and determine the buffer queue length N of the switching node as: N=Actual*(RTT-BaseRTT)=Diff*BaseRTT, Wherein, Diff=Expected-Actual, which is the difference between the expected rate Expected and the actual rate Actual. 9. The method for adjusting the congestion window in a communication network according to claim 1, 2, 3, 7 or 8, characterized in that said step B comprises: B1. If the length of the buffer queue is greater than the set queue length threshold, it is determined that the network bandwidth is saturated, and the packet loss generated in the network is caused by congestion, and the switching node is performed according to the corresponding congestion control strategy Adjustment of window and activation threshold; B2. If the length of the buffer queue is less than the set queue length threshold, it is determined that the network bandwidth is not saturated, and the packet loss generated in the network is caused by a change in link characteristics, and according to the change based on link characteristics The control strategy adjusts the switching node's window and startup threshold. 10. The method for adjusting the congestion window in a communication network according to claim 9, characterized in that the adjustment of the switching node's window and start-up threshold according to the control strategy based on link characteristic changes described in step B2 is : If three repeated ACK messages are received, adjust the slow start threshold equal to the ratio of the product of the estimated bandwidth and the minimum round-trip time to the size of the TCP segment, and if the current congestion window is greater than the slow start threshold, set the congestion window Equal to the slow start threshold; If the retransmission timer expires, then make the congestion window equal to 1, make the slow start threshold equal to the ratio of the product of the estimated bandwidth and the minimum round-trip time to the size of the TCP segment, and if the slow start threshold is less than 2, then make the slow start threshold equal to 2.

Images