KR100672979B1 - Congestion control method - Google Patents

Abstract

A congestion control method is provided to improve an availability ratio of an entire bandwidth by reducing linearly a size of a congestion window. A slow start process is performed to increase exponentially a size of a congestion window when a transmission side does not generate packet loss(S101,S102). A congestion avoidance process is performed to perform repeatedly the slow start process when the congestion window is less than a slow start threshold or increase linearly the size of the congestion window when the congestion window is less than the slow start threshold(S103). A constant process is performed to maintain the congestion window at a maximum state when the size of the congestion window approaches the maximum congestion window(S104). A reverse congestion avoidance process is performed to reduce linearly the size of the congestion window when the packet loss is generated from each of the slow start process, the congestion avoidance process, and the constant process. A slow stop process is performed to reduce exponentially the size of the congestion window.

Description

Congestion control method

1 is a flow chart of a congestion control method according to an embodiment of the present invention.

2A is a flowchart illustrating a process of performing a slow start step of a congestion control method according to an exemplary embodiment of the present invention.

delete

2B is a flowchart illustrating a process of performing a congestion avoidance step of a congestion control method according to an embodiment of the present invention.

2C is a flowchart illustrating a process of performing a constant step of a congestion control method according to an embodiment of the present invention.

2d is a flowchart illustrating a process of performing a congestion avoidance step of a congestion control method according to an exemplary embodiment of the present invention.
Figure 2e is a flow chart illustrating a process of performing a slow stop step of the congestion control method according to an embodiment of the present invention.

3 is a diagram illustrating a packet loss degree setting method for a congestion control method according to an embodiment of the present invention.

4 is a network structural diagram for a congestion control method according to an embodiment of the present invention;

<Description of the symbols for the main parts of the drawings>

200: gateway

210, 220, 230: base station
240: wireless mobile terminal device

The present invention relates to a congestion control method based on a Datagram Congestion Control Protocol (hereinafter referred to as "DCCP") that supports mobility of an embedded terminal. In particular, the size of the congestion window is adjusted according to packet loss. To improve bandwidth and to manage sessions fairly through a gateway.

In a ubiquitous computing environment, various terminals having heterogeneous characteristics exist in wired and wireless networks, and terminals with various kinds of computing capabilities coexist. In particular, in the ubiquitous network, in order to maximize the mobility of various terminals, many wireless access networks such as a mobile network or a sense network will be included, and the terminals will be mainly made up of lightweight and miniaturized embedded terminals. Mobility causes a rapid change in network bandwidth and congestion control methods are essential to efficiently handle it.

Transmission Control Protocol (hereinafter referred to as "TCP") or User Datagram Protocol to reliably transmit large amounts of continuity data such as multimedia or logistics data over the Internet; Hereinafter referred to as "UDP"). However, BSD (Berkeley Software Distribution) socket-based TCP consists of a thick protocol stack due to many features such as congestion control, flow control, and retransmission techniques, and limited hardware resources such as lightweight, miniaturized embedded terminals. The case is a very big burden.

UDP, on the other hand, has a lightweight protocol stack that lacks features such as congestion control, but does not guarantee reliability.

Accordingly, congestion control methods of transport layers such as TCP-Friendly and TCP-Like based on DCCP, which have a lightweight protocol and some reliability, have been proposed as a standard.

DCCP-based TCP-Friendly congestion control method is based on delay time and controls congestion almost similar to TCP, but sends ACK message using ACK vector which sends several ACK messages at once to improve transmission rate. Reduce

DCCP-based TCP-Like congestion control method is based on the packet loss rate, TCP-Friendly Rate Control (TFRC) method for TCP-friendly flow, the congestion window gradually increases in accordance with the packet loss rate, If consecutive packet loss occurs, reduce the congestion window in half.

These are all unreliable transmissions, just like UDP, but with connection establishment and teardown to provide some reliability.

Since the conventional congestion control methods are implemented based on a wired environment, if they are introduced without modification into a wireless environment in which a high bit error rate occurs, bandwidth degradation occurs. If a packet is treated as a loss due to this bit error rate, no congestion has occurred, but conventional congestion control methods recognize it as congestion and reduce the amount of transmission by reducing the size of the congestion window. As such, the conventional congestion control methods cannot distinguish packet loss due to bit error and packet loss due to congestion.

In addition, in the congestion control method of the conventional TCP or a similar policy, when the loss occurs, the congestion window is reduced in half, and thereafter, the size of the congestion window is linearly increased. However, if the congestion is not severe, a large portion of the network bandwidth will not be used if the transmission rate is cut in half at the same time in a large congestion window. In addition, the amount of congestion window increase in another terminal with a relatively small congestion window is less than this, which causes a problem of wasting network bandwidth for a certain time.

The current network structure is divided into the access network and the Internet used by terminals to access the Internet based on the gateway. Among them, the access network is divided into a wireless access network and a wired access network according to the connection method of the terminals. In the case of a wired access network, since a fixed number of terminals are connected, traffic fluctuations are smaller than that of the Internet or wireless access network. However, in the case of a wireless access network, due to the mobility of the terminals, the number of terminals that can access the base station, which is a practical access point, is variable. The traffic is therefore very variable. That is, if many terminals suddenly move and a data transmission request comes in more than one base station's processing limit, the cell is difficult to transmit all data required by the terminals due to congestion. Conventional congestion control methods have not been designed based on such a congestion structure, and thus, congestion control is difficult.

Accordingly, the present invention has been proposed to solve the above problems, and an object of the present invention is to provide a congestion window in order to distinguish between loss due to congestion and loss due to bit error caused by characteristics of a wireless environment. By linearly reducing the size, the overall bandwidth utilization is improved by excluding the possibility of the congestion window being halved in non-congestion situations, and exponentially reducing the size of the congestion window to half when there is some congestion. In addition, by reducing the number of occurrences of an error exponentially when additional errors occur in the process of reducing, to provide a congestion control method to prevent the waste of bandwidth that can occur due to a significant reduction in bandwidth.

Another object of the present invention is to provide a congestion control method in which the gateway can manage the sessions of the entire terminals by managing the sessions of the entire terminals.

 In order to achieve the above object, a congestion control method according to the present invention provides a method of controlling congestion by increasing or decreasing the size of a congestion window according to whether a packet is lost or not. If the packet is not lost due to the packet transmission, the slow start step of exponentially increasing the size of the congestion window, and the slow start step is repeated if the congestion window is less than the slow start threshold in the slow start step A congestion avoidance step of linearly increasing the size of the congestion window if the congestion window is greater than or equal to the slow start threshold; and maximum size of the congestion window if the size of the congestion window reaches a maximum in the congestion avoidance step Constant step to maintain the song When a new side transmits a packet corresponding to the size of the congestion window, and a packet loss occurs, the size of the congestion window is linearly reduced to distinguish between bit error caused by a characteristic of a wireless environment and loss caused by congestion. A congestion avoidance step and a slow stop step of exponentially reducing the size of the congestion window until the size of the congestion window is half to minimize waste of bandwidth due to congestion. When a loss occurs, each of the above steps is characterized by performing any one of a slow start step, a reverse congestion avoidance step, and a slow stop step according to the degree of packet loss.

The packet loss degree is a value obtained by subtracting the packet loss number from the number of packets corresponding to the increased congestion window size.

The slow start step may include: checking the slow stop flag for determining whether to perform the slow stop step and performing the slow stop step when the slow stop flag is 1, and when the slow stop flag is 0, If the number of packet losses is less than or equal to the number of packets by the increased congestion window size half, a reverse congestion avoidance step is performed, and the number of packet losses is less than the number of packets by the increased congestion window size and the increased congestion window size half If the number of packets is greater than the number of packets, the slow stop step is performed. If the number of packet losses is greater than or equal to the number of packets corresponding to the increased congestion window size, the size of the congestion window is reduced by half, and then the slow start step is performed. It is done.

The congestion avoidance step may include: checking the slow stop flag and performing the slow stop step when the slow stop flag is 1; and when the slow stop flag is 0, the number of packet loss increases the congestion window. If the size of one packet equals the size, a congestion avoidance step is performed; if the number of packet losses is greater than the number of packets equal to the increased congestion window size and smaller than all the congestion window sizes increased in this step, a slow stop step is performed. If the number of packet losses is greater than or equal to the number of packets corresponding to all the congestion window sizes increased in this step, the current congestion window is reduced by half, and then a slow start step is performed.

The constant step may include the same congestion control as in the congestion avoidance step when the slow stop flag is inspected and the slow stop flag is 1, and the slow stop flag is 0 and the slow stop flag is 0. It characterized in that to perform.

The reverse congestion avoidance step may include: checking the slow stop flag to perform the slow stop step when the slow stop flag is 1; and when the slow stop flag is 0, when the packet loss count is 0, the reverse stop. Perform the previous step of the congestion avoidance step, perform the reverse congestion avoidance step if the packet loss number is 1, perform the slow stop step if the number of packet loss is less than or equal to the packet loss number of the step before the congestion avoidance step If the number of packet losses is greater than the number of packet losses before the step of avoiding congestion, the size of the congestion window is reduced by half, and then the slow start step is performed.

The slow stop step may include: checking the slow stop flag to perform the slow stop step when the slow stop flag is 1, and when the slow stop flag is 0, exponentially increasing the size of the congestion window. While reducing, by reducing the number of occurrences of an error exponentially weighted in the process of further error reduction, characterized in that the congestion avoidance step is performed.

The maximum congestion window threshold and the slow start threshold may be set to values calculated by the following equation for fair session management based on the gateway.

은 세션

의 최대 혼잡 윈도우 임계치를 나타내고, T

은 세션

의 슬로우 스타트 임계치를 나타내고, N은 게이트웨이를 통해서 서비스 중인 모든 세션 수를 나타내고, W

은 게이트웨이의 최대 가용 윈도우 크기를 나타내 고, W

은 세션

의 최대 수신 윈도우 크기를 나타낸다.Where T

Session

Represents the maximum congestion window threshold of

Session

Slow start threshold of N, N denotes the number of all sessions in service through the gateway, W

Represents the maximum available window size of the gateway, W

Session

Represents the maximum receive window size.

In addition, the gateway recognizes the congestion window size of each terminal connected to the gateway and the available window size of the gateway through the maximum congestion window threshold and the slow start threshold.

In addition, if the packet loss does not occur, the present invention performs the slow start step if the congestion window size is less than the slow start threshold, the congestion window size is greater than the slow start threshold or In the same case, if the congestion window size is maximum, the constant step is performed; otherwise, the congestion avoidance step is performed.

Hereinafter, embodiments of the present invention as described above will be described in detail with reference to the accompanying drawings.

1 is a flowchart of a congestion control method according to an embodiment of the present invention.

As shown in FIG. 1, in a congestion control method according to an embodiment of the present invention, if a transmitting side transmits a packet corresponding to the congestion window size, there is no packet loss, and a slow start step is performed if the congestion window is less than the slow start threshold. A congestion avoidance step (S108) of linearly increasing the size of the congestion window if the congestion window is greater than or equal to the slow start threshold, and the congestion avoidance step (S108) A constant step (S109) of maintaining the maximum size of the congestion window when the maximum is reached, and a packet loss caused by the transmitting side transmitting a packet as large as the congestion window size, Inversely reducing the size of the congestion window to distinguish between losses caused by bit errors and congestion Job avoidance step (S103), and when the transmitting side transmits a packet corresponding to the size of the congestion window and receives a negative response, until the size of the congestion window is half to minimize waste of bandwidth due to congestion; And a slow stop step S104 of reducing the size of the congestion window.

Here, if packet loss occurs in the slow start step, the mixed avoidance step, and the constant step, the slow start step (S102), the reverse congestion avoidance step (S103), and the slow stop step (depending on the packet loss amount according to the number of packet losses) Perform any one of S104). On the other hand, if no packet loss occurs, the next state is determined by comparing the congestion window and the slow start threshold (S105). That is, if no packet loss occurs (S101) and the size of the congestion window is smaller than the slow start threshold (S105), the slow start step is performed (S106), and no packet loss occurs (S101). Is not smaller than the slow start threshold (S105) and the size of the congestion window is smaller than the maximum congestion window (S107), the congestion avoidance step is performed (S110), and no packet loss occurs (S101). If is not smaller than the slow start threshold (S105) and the size of the congestion window is not smaller than the size of the maximum congestion window (S107) a constant step is performed (S109). For reference, the maximum congestion window size here means a window state of a size that can transmit the most packets.

Meanwhile, the packet loss degree means a value obtained by subtracting the number of lost packets from the increased congestion window size.

More specifically, as shown in FIG. 3, since the number of packets of the congestion window in the interval i is m and no packets are lost, all m bits of the ACK vector are filled with one. As a result, the congestion window increases by the number of n packets in the i + 1 interval, and k packets are lost after transmitting a total of m + n packets. This loss is caused by a new increase in n packets instead of m packets which have already confirmed lossless transmission in interval i. Therefore, in the embodiment of the present invention, the degree of packet loss is defined as the value obtained by subtracting the number of k lost packets lost from n increased packets.

On the other hand, the two thresholds (maximum congestion window threshold, slow start threshold) is set to the following values for fair session management based on the gateway. Here, the slow start threshold means a limit of a slow start step in which a congestion window of the terminal transitions to a congestion avoidance step of performing a linear increase. These threshold values are obtained by considering the entire sessions at the gateway.

은 세션

의 최대 혼잡 윈도우 임계치를 나타내고, N은 게이트웨이를 통해서 서비스 중인 모든 세션 수를 나타내고,

는 게이트웨이의 최대 가용 윈도우 크기를 나타내고, W

은 세션

의 최대 수신 윈도우 크기를 나타낸다.Where T

Session

Denotes the maximum congestion window threshold of N, N denotes the number of all sessions in service through the gateway,

Represents the maximum available window size of the gateway, W

Session

Represents the maximum receive window size.

Equation 1 may be used to set the maximum congestion window threshold for fair session management based on the gateway.

Referring to Equation 1, the maximum congestion window threshold is set to the minimum of the maximum available window size of the gateway divided by the number of all sessions being served through the gateway and the maximum received window size of the session. In other words, a minimum value is set between the maximum congestion window size that the gateway can allocate for the session and the reception window size of the session. This maximum congestion window threshold resets its value each time a new session is added or terminated. This makes it possible to manage the entire session fairly and to use the bandwidth efficiently.

은 세션

의 슬로우 스타트 임계치이고, T

은 세션

의 최대 혼잡 윈도우 임계치이고, N은 게이트웨이를 통해서 서비스 중인 모든 세션 수이고, W

은 게이트웨이의 최대 가용 윈도우 크기이다.Where T

Session

Is the slow start threshold of T

Session

Is the maximum congestion window threshold of N, N is the number of all sessions in service through the gateway, W

Is the maximum available window size of the gateway.

Using Equation 2 above, a slow start threshold for fair session management based on a gateway may be set.

)× T

로 설정된다. 즉, 게이트웨이의 가용 윈도우에 여유가 있을 때는 T

(최대 혼잡 윈도우 임계치)가 설정되고, 게이트웨이의 가용 윈도우에 여유가 없을 때는 (1-

)× T

가 설정된다. 이는 새로운 세션이 들어올 여유 폭을 미리 남겨두어서 혼잡을 사전에 방지하기 위한 것이다. 이러한 슬로우 스타트 임계치는 새로운 세션이 추가되면 수학식 2를 재계산하여 새로운 슬로우 스타트 임계치를 구한다. 이때, 혼잡에 의해서 패킷 손실이 발생하면, 슬로우 스타트 임계치를 최대 혼잡 윈도우 임계치의 절반으로 설정한다. 이로써, 트래픽이 매우 가변적인 무선 접속망에서 갑자기 많은 단말들이 접속함으로 인해 발생하는 혼잡을 효율적으로 제어할 수 있게 된다.Referring to Equation 2, the slow start threshold is set to the maximum congestion window threshold if the sum of the maximum congestion window thresholds of all sessions serving through the gateway is less than the maximum available window size of the gateway, otherwise (1-

) × T

Is set to. That is, if there is room in the available window of the gateway, T

When (Maximum Congestion Window Threshold) is set, and there is no room in the available window of the gateway,

) × T

Is set. This is to prevent congestion in advance by leaving room for new sessions in advance. This slow start threshold is recalculated by Equation 2 when a new session is added to obtain a new slow start threshold. At this time, if packet loss occurs due to congestion, the slow start threshold is set to half of the maximum congestion window threshold. As a result, in a wireless access network where traffic is very variable, congestion caused by sudden access of many terminals can be efficiently controlled.

As such, the gateway sets the maximum congestion window size and slow start threshold for gateway-based fair session management. Therefore, all terminals are under the control of the gateway, thereby enabling efficient congestion control.

Hereinafter, a process of performing each step of the congestion control method according to an embodiment of the present invention will be described.

For reference, in the process of performing each step of the congestion control method according to an embodiment of the present invention, the slow stop flag is first examined. The slow stop flag is a flag indicating whether the current slow stop step is being performed. If the value is 1, the slow stop flag indicates that the slow stop step is being performed. The reason for checking the slow stop flag first is that if the slow stop flag has a value of 1 and performs the slow stop step, it continues until the congestion window is reduced to half the size of the congestion window from the step prior to the slow stop. This is to maintain the slow stop phase.

2A is a flowchart illustrating a slow start step of a congestion control method according to an exemplary embodiment of the present invention. In the slow start step, if packet loss does not occur, a slow start step → congestion avoidance depending on the size of a congestion window. Increase the number of packets that can be sent by following the steps-> constant steps. On the other hand, if a packet loss occurs, the reverse congestion avoidance step, the slow stop step, or the slow start step is performed according to the packet loss level.

As shown in FIG. 2A, the slow start step of the congestion control method according to the embodiment of the present invention first checks the slow stop flag when a packet loss occurs, and if the value is greater than zero (S111), a slow stop step. To perform (S112). On the other hand, if the slow stop flag is not greater than zero (S111), the packet loss degree (p_loss) is greater than or equal to half (w_inc / 2) of the number of packets by the increased congestion window size and greater than the increased congestion window size. If it is small (w_inc / 2 <= p_loss <w_inc), the congestion avoidance step is performed (S114), and if the packet loss degree (p_loss) is smaller than half of the increased mixed window size, and greater than 0 (0 <p_loss <w_inc / 2) (S115), the slow stop step is performed (S112), if all of the above conditions are not satisfied, the size of the congestion window is reduced to half (S116) and then the slow start step is performed (S117).

2B is a flowchart illustrating a congestion avoidance step of the congestion control method according to an embodiment of the present invention. In the congestion avoidance step, if a packet loss does not occur, the size of the congestion window is linearly increased. If no loss occurs and the size of the congestion window reaches its maximum, a constant step is performed. On the other hand, if a packet is lost, a slow start step, a slow stop step, or a reverse congestion avoidance step is performed according to the packet loss level.

As shown in FIG. 2B, the congestion avoidance step of the congestion control method according to the embodiment of the present invention first checks the slow stop flag when a packet loss occurs, and if the value is greater than zero (S121), the slow stop step. To perform (S122). On the other hand, if the slow stop flag is not greater than 0 (S121), if the packet loss degree (p_loss) is smaller than the number of packets (w_inc) by the increased congestion window size and is greater than or equal to 0 (0 <= p_loss <w_inc) In step S123, the congestion avoidance step is performed (S124), and the packet loss degree (p_loss) is less than 0, and the packet count (cws) corresponding to the current congestion window size is subtracted from the packet count (ssthresh) by the slow start threshold. If greater than the value (ssthresh-cwnd <p_loss <0) (S125), the slow stop step is performed (S122). If all of the above conditions are not satisfied, the size of the congestion window is reduced to half (S126) and then the slow start is performed. Step S127 is performed.

2C is a flowchart illustrating a process of performing a constant step of a congestion control method according to an embodiment of the present invention. The constant step of the congestion control method according to an embodiment of the present invention is an extension step of the congestion avoidance step. In other words, the packet loss does not occur, and the congestion avoidance step is performed when the congestion window reaches the maximum congestion window.

delete

As shown in FIG. 2C, when the packet loss occurs, the constant step of the congestion control method according to the embodiment of the present invention first checks the slow stop flag, and if the value is greater than 0 (S131), the slow stop step. To perform (S132). On the other hand, when the slow stop flag is not greater than 0 (S131), if the packet loss degree (p_loss) is greater than or equal to -1 and less than 0 (-1 <= p_loss <0) (S133), the reverse congestion avoiding step (S134), if the packet loss degree (p_loss) is less than -1 and is greater than the packet count (ssthresh) by the slow start threshold minus the packet count (maxthresh) by the maximum congestion window threshold (ssthresh-maxthresh < p_loss <-1) (S135), the slow stop step is performed (S132), if all of the above conditions are not satisfied, the size of the congestion window is reduced to half (S136) and then the slow start step is performed (S137). .

The slow start step, congestion avoidance step and constant step perform a slow start step, a reverse congestion avoidance step or a slow stop step depending on the degree of congestion at the time of packet loss. That is, the slow start step, congestion avoidance step, and constant step are each a slow start step and a reverse congestion avoidance step according to the packet loss degree, which is a value obtained by subtracting the packet loss count from the packet count corresponding to the increased congestion window size when packet loss occurs. To perform any one of the steps, slow stop.

Hereinafter, the reverse congestion avoidance step and the slow stop step of reducing the size of the congestion window in order to efficiently handle the congestion, rather than increasing or maintaining the size of the congestion window, will be described.

Here, the congestion avoidance step or the slow stop step is not performed when packet loss does not occur, and is performed only when packet loss occurs.

2d is a flowchart illustrating a process of performing a congestion avoidance step of a congestion control method according to an embodiment of the present invention. In the reverse congestion avoidance step, a current loss state is a characteristic of a wireless network while decreasing the size of a congestion window by one. It distinguishes between bit error caused by packet loss and packet loss caused by congestion. Therefore, the congestion avoidance step controls congestion due to packet loss occurring in the step before the congestion avoidance step.

As shown in FIG. 2D, the congestion avoidance step of the congestion control method according to an embodiment of the present invention first checks the slow stop flag when a packet loss occurs, and if the value is greater than zero (S141), a slow stop is performed. To perform the step (S142). On the other hand, if the slow stop flag is not greater than zero (S141), if the number of packets lost is zero (nack = 0) (S143), the previous step of the decongestion avoidance step is performed (S144), and the number of lost packets is If 1 (nack = 1) (S145), the reverse congestion avoidance step is performed (S146), and if the number of lost packets is less than or equal to the number of lost packets from the step before the decongestion avoidance step (nack <= pre_rca_nack) (S147) In step S142, if the above conditions are not satisfied, the size of the congestion window is reduced to half (S148), and then the slow start step is performed (S149).

Meanwhile, FIG. 2E is a flowchart illustrating a slow stop step of the congestion control method according to an embodiment of the present invention. In the slow stop step, when the packet loss is small, that is, when the degree of congestion is low, bandwidth is increased due to congestion. Minimize waste.

As shown in FIG. 2E, the slow stop step of the congestion control method according to the embodiment of the present invention first checks the slow stop flag, and if the value is greater than zero (S151), performs a slow stop step (S152). If the slow stop flag is not greater than zero (S151), a congestion avoidance step is performed (S153). In other words, the slow stop phase maintains the slow stop phase by exponentially reducing the size of the congestion window by setting the slow stop flag to 1 until the size of the congestion window is half the size of the congestion window. In case of occurrence, the decrease is increased by exponentially weighting the number of occurrences of the error. When the congestion window is half the size and the slow stop step is completed, the congestion avoidance step is performed after setting the slow stop flag to zero.

In the congestion control method according to the embodiment of the present invention as shown in FIG. 4, the gateway 200 sets a threshold value so that the base stations 210, 220, and 230 manage sessions fairly. Congestion does not occur in the base station 210, 220, 230 even when the terminal 240 connected to the station 210, 220, 230 moves while transmitting data. In this way, the gateway 200 may know the packet loss and delay, as well as the size of the congestion window that all the terminals 240 obtain in the transmission process, and all terminals currently serviced by the gateway 200. The total transmission window size of 240 may also be known. Therefore, it is possible to perform a fair and efficient congestion control by considering not only packet loss and delay but also the window size of each terminal 240 observed by the gateway 200 and the available window size of the gateway 200.

For reference, after performing each of the above steps (slow start step, congestion avoidance step, constant step, reverse congestion avoidance step, slow stop step), the sender changes the congestion window size changed in the data packet to a change option. In addition, the receiving side changes the ack ratio to the value of the change option transmitted from the gateway, waits until receiving the number of packets, and then changes the packet to the changed ack ratio. After receiving, the ACK vector is transmitted to the transmitting side again.

The present invention improves overall bandwidth utilization by linearly reducing the size of the congestion window in order to distinguish between loss due to bit error and loss due to congestion caused by the characteristics of the wireless environment when loss occurs. Bandwidth generated by drastically reducing the size of the congestion window by exponentially reducing the size of the congestion window by half, while exponentially weighting the number of occurrences of the error when additional errors occur in the decreasing process. It is possible to efficiently manage the bandwidth by judging the situation caused by congestion and the loss due to bit error while preventing waste of data.

In addition, according to the present invention, the gateway can manage the sessions of all terminals in an integrated manner.

In addition, in the present invention, since the gateway performs all congestion control functions, the terminal has a transport layer protocol stack having only a simple function of receiving data and transmitting a response message including an Ack vector, thereby reducing the overall burden on the terminal. The performance of the terminal can be improved.

Claims (9)

In the method of controlling the congestion by increasing or decreasing the size of the congestion window according to the packet loss, A slow start step of exponentially increasing the size of the congestion window if a transmitting side transmits the packet as large as the congestion window and no packet loss occurs; A congestion avoidance step of repeating the slow start step if the congestion window is less than the slow start threshold in the slow start step and linearly increasing the size of the congestion window if the congestion window is above the slow start threshold; A constant step of maintaining the maximum size of the congestion window when the size of the congestion window reaches a maximum congestion window in the congestion avoidance step; When the transmitting side transmits a packet corresponding to the size of the congestion window in each step in the slow start, congestion avoidance, and constant phase, if packet loss occurs, the loss caused by the bit error and congestion caused by the characteristics of the wireless environment is eliminated. In order to distinguish the congestion window, the size of the congestion window is linearly reduced, and if it is found to be a bit error, the process returns to the previous stage of error occurrence. A reverse congestion avoiding step transitioned to the step; If the cause of the error is found to be congestion in the congestion avoidance step, the size of the congestion window is exponentially reduced to half of the current congestion window size and the congestion window size is reduced in order to control the congestion while minimizing wasting bandwidth. If an additional error occurs during the process, a slow stop step is performed to control congestion quickly by reducing the size of the congestion window by multiplying the existing reduction by 2 ^{(number of errors)} . When the packet loss occurs in the slow start, congestion avoidance, and the constant step, the congestion control method is performed by transferring the congestion control step as shown in Equation 1), Equation 2), Equation 3) in each step. Equation 1) Congestion Control Phase Transition According to Packet Loss in Slow Start Phase

Equation 2) Congestion Control Stage Transition Formula According to Packet Loss in Congestion Avoidance Stage

Equation 3) Congestion Control Phase Transition According to Packet Loss in Constant Phase

Packet loss

: Current Congestion Window Size in Session i

: Increased Congestion Window Size in Session i

: Maximum Congestion Window Threshold for Session i

: Slow Start Threshold in Session i The method of claim 1, The packet loss degree is a congestion control method, characterized in that the number of packets as much as the increased congestion window size minus the number of packet loss. delete delete delete The method of claim 1, wherein the congestion avoidance step, If packet loss occurs in the slow start phase, congestion avoidance phase, and constant phase, and the packet loss degree is not large by using Equation 1) to Equation 3), whether the packet is actually congested in each step, A congestion control method characterized by controlling the congestion window size and the congestion control step using Equation 4) whether or not it is a bit error according to the characteristics of the environment. Equation 4

: Number of packets lost in the congestion avoidance phase

: Number of packets lost in the stage before transitioning to the congestion avoidance phase The method of claim 1, wherein the slow stop step, If packet loss occurs in the slow start phase, congestion avoidance phase, and constant phase, and the packet loss degree is large using Equation 1) to Equation 3), it is proved to be a real congestion situation and the congestion window size is expressed by Equation 5 Congestion control method characterized in that the control using. Equation 5)

: Current Congestion Window Size

: Congestion window size before transition to slow stop

: Congestion Window Size After Slow Stop Phase Ends

: Number of times of slow stop step progress (default: 0)

: Number of error occurrences during the slow stop phase (initial value: 0) The method of claim 1, The maximum congestion window threshold indicating the maximum value of the congestion window size and the slow start threshold indicating the entry value from the slow start step to the congestion avoidance step are set to values calculated by Equation 6) for fair session management based on the gateway. Congestion control method characterized in that. (6)

: Maximum congestion window setting formula

: Slow start threshold setting formula

: All sessions in service through the gateway

: Maximum Available Window Size of Gateway

: Maximum receive window size for session i delete

Images