CN106856458B

CN106856458B - Transmission layer source end rate control method

Info

Publication number: CN106856458B
Application number: CN201710048969.XA
Authority: CN
Inventors: 官权升; 马海城
Original assignee: South China University of Technology SCUT
Current assignee: South China University of Technology SCUT
Priority date: 2017-01-20
Filing date: 2017-01-20
Publication date: 2020-07-28
Anticipated expiration: 2037-01-20
Also published as: CN106856458A

Abstract

The invention discloses a transmission layer source end rate control method, which is characterized in that on the basis of a transmission layer protocol, a source end finds out a proper sending rate in a binary search mode according to RTS (request to send) times of each data packet of a link layer of a node, so that the RTS times is within a preset interval from 1 to the maximum RTS times. At this sending rate, severe channel competition and network congestion near the source end caused by data packet burst of the source end by the transport layer protocol can be effectively relieved. The method can also detect changes in network conditions based on relative changes in the number of adjacent RTS's, thereby restarting the process of finding the most appropriate sending rate. The method has four states, namely an aggressive sending state, a slow starting state, a binary search state and a stable state. The four states can mutually shift under certain conditions.

Description

Transmission layer source end rate control method

Technical Field

The invention relates to the technical field of wireless network congestion control, in particular to a transmission layer source end rate control method.

Background

Existing research has shown that the Transmission Control Protocol (TCP) exhibits severe performance degradation in multi-hop wireless networks. The main reason for this is that the transport layer often misjudges the congestion status of the network. The loss of a packet is considered a sign of network congestion. However, there are a number of other factors in a wireless network that can lead to packet loss. Such as high error rate of wireless channels, collisions of open shared channels, and routing failures caused by movement of terminals. Frequent packet loss suppresses the increase of the transmission window of the transmission layer, thereby reducing throughput.

The three reasons for packet loss occur in the physical layer, the link layer and the network layer, respectively. Research has indicated that packet loss caused by open shared channel collision occurring at the link layer is a major factor of wireless network packet loss, and channel error at the physical layer and route failure reconstruction at the network layer have relatively small influence. Therefore, in order to effectively improve the transmission performance of TCP in a wireless network, attention is focused on mitigating channel collision at a link layer.

TCP employs a send window for network congestion control. If the transmission window is increased, the data packets that can be transmitted within one RTT are increased, and the throughput of TCP is also increased, and vice versa. However, TCP does not indicate how the data packets that can be sent within the send window will be sent, so most TCP versions send down in a best effort manner at once, i.e., all data packets that can be sent are sent down regardless of the node cache cannot accommodate it. This behavior does not cause too much problem in the wired network, but in the wireless network, due to the problem of the 4-hop interference range specific to the wireless network, the channel competition near the source end is too severe, which finally wastes the channel resource and reduces the network transmission efficiency.

Similar to TCP, when the transmission rate of UDP is too high, a situation similar to TCP data packet burst occurs, and network resources are wasted due to severe contention of wireless channel, which eventually leads to network performance degradation.

In summary, the transport layer needs a source rate control mechanism to alleviate congestion in the wireless network. TCP has congestion control, flow control and reliability control functions, but lacks a source-side rate control function at the micro-level. The underlying links in wired networks indirectly implement the rate control function, but not in wireless networks. Therefore, a transport layer rate control function is necessary. The same is true for UDP. Rate control is seen as a very useful tool for congestion control.

Disclosure of Invention

The invention aims to solve the defects in the prior art, and provides a transmission layer source end rate control method, which determines a sending rate according to RTS times and MAC sending time parameters on a link layer of a source node to ensure that the RTS times are in a preset interval between the maximum value and the minimum value, so as to achieve the purposes of slowing down the intense channel competition and serious network congestion near the source node, effectively improve the throughput, reduce the time delay and improve the performance of a transmission layer.

The purpose of the invention can be achieved by adopting the following technical scheme:

a transport layer source end rate control method, the method comprising:

and step of aggressive sending: requesting a data packet from a transmission layer as long as a link layer queue is empty, then sending the requested data packet out, and simultaneously measuring RTS (request to send) times and MAC (media access control) sending time required for sending each data packet at the moment and recording the RTS times and the MAC sending time as maximum RTS times and maximum MAC sending time; after a certain number of data packets are sent, uploading the measured values of the maximum RTS times and the longest MAC sending time to a transmission layer, and entering a slow start step;

a slow start step: recording the reciprocal of the longest MAC sending time measured in the step of aggressive sending as an initial sending rate, and searching a lower bound rate, sending a data packet downwards at the initial sending rate in a transmission layer, measuring the current RTS (request to send) times after sending a certain number of data packets, if the current RTS times is less than half of the maximum RTS times, multiplying the sending rate, and then sending the data packet at a new sending rate; if the current RTS times is more than half of the maximum RTS times, recording the current sending rate as a search upper bound rate, and entering a binary search step;

a binary search step: recording a search interval from a search lower bound rate to a search upper bound rate as an initial rate search interval, and searching a rate in a binary search mode in the interval, so that when a data packet is sent at the sending rate, the measured RTS (request to send) times is within a preset interval between 1 and the maximum RTS time, if the search is successful, entering a stable step, and if the search is failed, entering an aggressive sending step again;

a stabilizing step: and sending a data packet at the sending rate searched in the two-step searching step, monitoring the relative change between the current RTS time and the last RTS time, if the current RTS time is greater than a certain multiple of the last RTS time or less than the certain multiple of the last RTS time, re-entering the step of aggressive sending, and otherwise, keeping the sending rate unchanged.

Further, the RTS number and the MAC sending time are both measured by a sliding window averaging method, specifically, a sliding window average value calculated after a certain number of data packets are sent.

Further, the step of aggressive sending continuously sends data packets, when a transmission layer has data packets that can be sent, a link layer queue pulls the data packets upwards, and then tries to send the data packets; the link layer queues in the slow start step, the binary search step and the stabilization step do not actively pull data packets.

Further, in the slow start step, the binary search step and the stabilization step, a sending timer is used for controlling the sending rate, and the timeout time of the sending timer is set to be the reciprocal of the sending rate; when the sending timer is overtime, the transmission layer sends a data packet downwards, then the timer is restarted, and new overtime time needs to be calculated according to the current rate when the sending timer is restarted each time.

Further, the binary search step specifically includes:

firstly, setting a sending rate as a middle value of a rate search interval, sending a data packet by the sending rate, and then adjusting the sending rate according to the current RTS times measured at the sending rate, wherein the method comprises the following three conditions:

(1) if the current RTS times are larger than the maximum value of the RTS time preset interval, setting the search rate upper bound as the current rate, and then setting the current rate as the middle value of the search rate interval;

(2) if the current RTS times are smaller than the minimum value of the RTS time preset interval, setting the lower bound of the search rate as the current rate, and then setting the current rate as the middle value of the search rate interval;

(3) and if the current RTS times is within the RTS time preset interval, stopping the search process of the rate and entering the stabilizing step.

Further, the RTS number preset interval is a range artificially set between the minimum value of the RTS number and the maximum value of the RTS number, and is set to be 0.3 to 0.5 times the interval between the minimum RTS number and the maximum RTS number.

Further, in the stabilizing step, if the RTS number changes greatly, which indicates that the network condition changes, a new round of rate control process needs to be performed again, where the switching condition is set to be that the RTS number is greater than 1.2 times or less than 0.8 times of the last RTS number.

Further, if the current RTS number is within the RTS number preset interval, in the search process of the stop rate, the upper and lower boundary distances of the stop search are 5% of the initial upper and lower boundary distances.

Compared with the prior art, the invention has the following advantages and effects:

1. the invention has the advantages of simplicity and easy implementation. The method is improved on the basis of a transport layer protocol, the sending rate is only adjusted at the source node according to the channel information of the node, the change of the transport layer of the source node is little, and the interactive process between adjacent nodes is not changed.

2. The invention has the advantage of wide applicability. The RTS number is a very accurate indicator of the network congestion condition, and is true for different channel bandwidths. The RTS times are controlled between the maximum value and the minimum value, so that network congestion can be obviously reduced, and meanwhile, the utilization rate of network resources cannot be lost.

3. The invention has the advantage of rapid reaction. The RTS number used for adjusting the rate is only measured in the source node and transmitted in the node, so that the information is accurate and timely. The search rate uses a binary search algorithm, which can meet the requirements with the fastest convergence rate.

Drawings

Fig. 1 is a transport layer state transition diagram of a transport layer source end rate control method disclosed in the present invention;

fig. 2 is a flowchart of RTS number measurement of a transmission layer source end rate control method disclosed in the present invention;

fig. 3 is a flow chart of binary search of a method for controlling a rate at a source end of a transport layer according to the present invention.

Detailed Description

In order to make the objects, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are some, but not all, embodiments of the present invention. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.

Examples

The invention aims to avoid serious channel collision and network congestion caused by data burst at the source end of a transmission layer, and the source end performs sending rate control according to RTS times. And finding a sending rate by a binary search mode, so that the RTS number is in a preset interval between the maximum RTS number and the minimum RTS number at the sending rate. When there is a significant change in the network conditions, the method can detect this change and resume a new transmission rate control procedure.

Based on the principle, the specific scheme of the invention is as follows:

and step of aggressive sending: requesting a data packet from a transmission layer as long as a link layer queue is empty, then sending the requested data packet out, and simultaneously measuring RTS (request to send) times and MAC (media access control) sending time required for sending each data packet at the moment, and recording the RTS times and the MAC sending time as maximum RTS times and maximum MAC sending time. And after a certain number of data packets are sent, uploading the measured values of the maximum RTS times and the longest MAC sending time to a transmission layer, and entering a slow start step.

A slow start step: and recording the reciprocal of the longest MAC transmission time measured in the step of aggressive transmission as an initial transmission rate and searching a lower bound rate. The data packets are sent down at the transport layer at the initial sending rate. And after a certain number of data packets are sent, the current RTS times are measured. If the current RTS times is less than half of the maximum RTS times, multiplying the sending rate, and then sending a data packet at a new sending rate; if the current RTS times is more than half of the maximum RTS times, the current sending rate is recorded as the upper limit searching rate, and a binary searching step is carried out.

A binary search step: note the initial rate search interval from the lower bound rate of the search to the upper bound rate of the search. And searching a rate in a binary search mode in the interval, so that when the data packet is sent at the sending rate, the measured RTS number is within a preset interval between 1 and the maximum RTS number. And if the search is successful, entering a stabilizing step, and if the search is failed, re-entering an aggressive sending step. The search failure means that when the rate search interval is smaller than a certain proportion of the initial rate search interval, the current RTS number does not enter a preset interval of the RTS number.

A stabilizing step: and sending the data packet at the sending rate searched in the two-step searching step, and simultaneously monitoring the relative change between the current RTS times and the last RTS times. And if the current RTS times is larger than a certain multiple of the last RTS times or smaller than a certain multiple of the last RTS times, re-entering the step of aggressive sending. Otherwise, the sending rate is kept unchanged.

The four steps respectively correspond to four states in a transmission layer, namely an aggressive sending state, a slow starting state, a binary search state and a stable state. The four states can be mutually converted under specific conditions, and the specific conversion process is shown in fig. 1.

The method for controlling the rate of the source end of the transmission layer based on the scheme has the following specific implementation process:

1. the link layer transitions between two phases. The difference between the step of aggressive sending and the steps of slow start, binary search and stabilization is that the step of aggressive sending continuously sends data packets, and when the data packets can be sent in the transmission layer, the link layer queue pulls the data packets upwards and then tries to send. In the slow start step, the binary search step and the stabilization step, the link layer queue does not actively pull the data packet. Therefore, there are two different stages in the link layer, which are set as an aggressive sending stage and a rate control stage, respectively, corresponding to an aggressive sending step, a slow start step, a binary search step, and a stabilization step. When switching between different steps, it is necessary to synchronize the phases of changing the link layer.

In order to reduce errors caused by fluctuation of network conditions, RTS times and MAC sending time values measured in all steps are average values of a sliding window calculated after a certain number of data packets are sent.

In specific application, RTS times and MAC sending time are measured in a sliding window average mode. Each time the current RTS number and MAC transmission time are measured, the current value is averaged with a previous number of measurements. The RTS times and the MAC sending time are uploaded to the transmission layer after the link layer sends a certain number of data packets, so that the influence of the previous network condition on the current measured value can be relieved.

The different states of the link layer only differ in whether a packet is requested upwards when the queue has space, and have no effect on the measurement of the number of RTS and the MAC transmission time. In this example, the number of packets transmitted is set to 10. The specific process is shown in fig. 2.

2. After the maximum RTS times and the maximum MAC sending time are measured in the aggressive sending step, the link layer is converted into a rate control stage, and the maximum RTS times and the maximum MAC sending time are uploaded to the transmission layer. And the transmission layer sets a sending timer according to the reciprocal of the maximum MAC sending time and enters a slow starting step. And after the link layer sends 10 data packets, uploading the newly counted RTS times and informing the transmission layer of updating the rate. If the RTS number is more than half of the maximum RTS number, which indicates that the current sending rate is high enough, the slow start process stops with the rate as the upper search limit, and the binary search step is entered. If the RTS number is less than half of the maximum RTS number, the current sending rate is multiplied, and the data is continuously sent.

3. In order to autonomously control the transmission rate, a transmission timer is provided. The transmission timer has two states, a stop state and an active state. The timer is in a stop state during the step of aggressive sending, and is in a running state during other steps. And after the sending timer is overtime, if the transmission layer caches a data packet, immediately sending a data packet.

And controlling the sending rate in the slow starting step, the binary searching step and the stabilizing step by using a sending timer. The timeout time of the transmission timer is set to the inverse of the transmission rate. When the sending timer is overtime, the transmission layer sends a data packet downwards, and then the timer is restarted. Each time the timer is restarted, a new timeout time needs to be calculated based on the current rate.

4. After the aggressive sending step and the slow start step, a search interval for the rate is determined. Other methods may be used to find a suitable rate. Such as a linear search. However, there is a linear relationship between the sending rate and the number of RTS, so that a binary search can be used for finding, which is the fastest method. Similarly, a rate that can make the RTS number be at a fixed point can be searched, but in order to increase the search convergence rate, the search process can be stopped when the RTS number is within a preset RTS number interval. The RTS number preset interval is empirical data, and the value is different under different networks.

The RTS number preset interval refers to a range artificially set from the minimum RTS number to the maximum RTS number. The minimum number of RTS times is 1. The maximum value of the RTS times is measured when the data packet is sent according to the aggressive sending step. When the RTS times are at a certain position between the two extreme values, the utilization rate of the network resources reaches the maximum. To speed up the convergence rate, the point is set to an interval. This interval is an empirical value and varies with different network conditions.

In this example, the interval is set to 0.3 to 0.5 times the minimum RTS number to maximum RTS number interval. The change in transmission rate is relatively independent of the actual transmission of the data packets. Changing the transmission rate does not affect the transmission timer that is currently timing. After the sending timer sends data, the sending rate is adjusted until the link layer uploads the latest RTS times after 10 data packets are actually sent.

Referring to fig. 3, a specific flow analysis for binary search is as follows:

(1) if the RTS times are above a preset interval, which indicates that the current rate is too large, setting the current rate as an upper search bound, and setting the current rate as a middle value between the upper search bound and a lower search bound;

(2) if the RTS times are below a preset interval, which indicates that the current rate is too small, setting the current rate as a lower search bound, and setting the current rate as a middle value between the lower search bound and an upper search bound;

(3) and if the RTS times are within a preset interval, indicating that the rate is a proper rate, stopping the searching process and setting the state of the transmission layer to be a stable state.

Considering that the sending rate meeting the condition is in a range, if the distance between the upper and lower boundaries in the binary search process is very short, the condition of stopping the search is still not achieved, which indicates that the search process may have misjudgment or the network condition has sudden change, so the network condition needs to be detected again, the transmission layer enters the step of aggressive sending, and simultaneously the state of the link layer also needs to be converted from the rate control stage to the stage of aggressive sending. The upper and lower bounds distance at which the search was stopped in this example is 5% of the initial upper and lower bounds distance.

5. The rate needs to adapt to changing network conditions, which can result in a change in the number of RTS. When the network is in a stable state, if the RTS frequency changes greatly, which indicates that the network state changes, a new rate control process needs to be performed again. The changeover condition is set in this example to be 1.2 times or less than 0.8 times the number of RTS times greater than the last RTS time.

6. The transport layer needs to be changed accordingly:

(1) if the transport layer protocol is UDP, a queue is arranged on the transport layer for buffering the data packet sent by UDP. The queue processes overflow data in a tail discard mode. The queue length is not set too small, which may better smooth the impact of UDP packet bursts. In this example set to a length of 50 packets.

(2) If the transmission layer protocol is TCP, the limitation of the sending window on the sending of the data packet is removed, and the function of the TCP for actively sending the data packet is cancelled. Whether TCP is able to send data packets depends on whether the send timer is expired and whether the link layer queue is empty in the aggressive sending step. The other functions of TCP are left unchanged.

The above embodiments are preferred embodiments of the present invention, but the present invention is not limited to the above embodiments, and any other changes, modifications, substitutions, combinations, and simplifications which do not depart from the spirit and principle of the present invention should be construed as equivalents thereof, and all such changes, modifications, substitutions, combinations, and simplifications are intended to be included in the scope of the present invention.

Claims

1. A method for controlling a rate at a source end of a transport layer, the method comprising:

and step of aggressive sending: requesting a data packet from a transmission layer as long as a link layer queue is empty, then sending the requested data packet out, and simultaneously measuring RTS (request to send) times required for sending each data packet and MAC (media access control) sending time required for each data packet from entering an MAC (media access control) layer to receiving a confirmation frame at the moment, and recording the RTS times and the MAC sending time as the maximum RTS times and the longest MAC sending time; after a certain number of data packets are sent, uploading the measured values of the maximum RTS times and the longest MAC sending time to a transmission layer, and entering a slow start step;

2. The method according to claim 1, wherein the RTS number and the MAC sending time are both measured by a sliding window averaging method, specifically, a sliding window average value calculated after a certain number of data packets are sent.

3. The method of claim 1, wherein the aggressive sending step continues sending data packets, and when there is data packets that can be sent in the transport layer, the link layer queue pulls up the data packets and then tries to send; the link layer queues in the slow start step, the binary search step and the stabilization step do not actively pull data packets.

4. The method of claim 1, wherein a sending timer is used in the slow start step, the binary search step and the stabilization step to control the sending rate, and the timeout time of the sending timer is set to the reciprocal of the sending rate; when the sending timer is overtime, the transmission layer sends a data packet downwards, then the timer is restarted, and each time the sending timer is restarted, new overtime time needs to be calculated according to the current rate.

5. The method for rate control at the source end of the transport layer according to claim 1, wherein the binary search step specifically comprises:

6. The method as claimed in claim 5, wherein the RTS number preset interval is a range artificially set between a minimum RTS number and a maximum RTS number, and the RTS number preset interval is set between 0.3 and 0.5 times the minimum RTS number and the maximum RTS number.

7. The method as claimed in claim 1, wherein in the stabilizing step, if the RTS times change significantly, which indicates that the network condition changes, a new round of rate control process needs to be performed again, and the transition condition is set such that the RTS times are greater than 1.2 times or less than 0.8 times of the last RTS times.

8. The method of claim 5, wherein if the current RTS number is within a preset RTS number interval, the upper and lower bounds of the distance to stop the search in the search process of the stop rate are 5% of the initial upper and lower bounds of the distance.

9. The method of claim 1, wherein the failure of search means that the current RTS number does not yet enter a preset RTS number interval when the rate search interval is smaller than a certain proportion of an initial rate search interval.