JP2000078195A - Retransmission control method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To efficiently transfer data even on the condition of frequently generating packet loss by immediately retransmitting a reported lost packet only when the packet loss is reported by selective dispatch confirmation information and a congestion window size is equal to or smaller than a threshold. SOLUTION: The flow control of a TCP (protocol layer 4) to be executed by a layer 4 processing part 6 is made into window system and as a parameter for congestion control, in addition to the reception window size declared as a packet reception enable amount from a communicating party, the congestion window size showing a packet transmission enable amount reflecting the transfer condition of a communication network 3 is provided. Based on both the parameters, a transmission node 1 transmits the prescribed number of data packets to a reception node 2 every time. When the packet loss is detected based on the selective confirmation information reported from the reception node 2, only when the congestion window size is equal to or smaller than the threshold, that lost packet is sent again.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a retransmission control method, and more particularly to a retransmission control method in TCP having a function of performing selective transmission confirmation.

[0002]

2. Description of the Related Art Generally, TCP (Trunk) widely used on the Internet is a typical example of a layer 4 protocol.
In the Reno algorithm, which is an Ansmission Control Protocol, a timeout occurs or three duplicate ACs
It detects packet loss in response to K packet reception and performs retransmission control. Since retransmission by timeout is determined by the RTT (Round Trip Time) and the accuracy of the timer on the transmitting side, it takes a relatively long time to start up.

[0003] In the retransmission by receiving the duplicate ACK packet, as shown in FIG.
(Reception acknowledgment) By receiving three ACK packets 42 to 44 storing the same reception sequence number (Acknowledgement Number) as the packet 41, it is determined that the packet is lost, and the retransmission operation is performed. This is in consideration of the fact that the delivery order is not guaranteed in the lower layer, and to cope with the case where the packet arrival order has been changed.

[0004] The retransmission control based on the duplicate ACK packet is performed at a higher speed than retransmission based on timeout, and is therefore called Fast Retransmit (high-speed retransmission control). In addition, SACK (Selective Delivery Confirmation: Selective Ackn
The retransmission scheme using the owledgement option is not specified, but only states that the conventional scheme should be preserved. Therefore, like the conventional Reno-TCP,
Generally, a retransmission operation is performed by receiving a plurality of duplicate SACKs.

[0005]

However, in such a conventional retransmission control method, the packet loss detection by the duplicate ACK packet requires a total of 4 packets for the first ACK packet and three subsequent duplicate ACK packets on the transmitting side. Two A
CK packet needs to be received, but Reno-TC
In P, every time a packet loss due to a duplicate ACK packet is detected, the network is determined to be congested, and the value of the congestion window size cwnd indicating the number of transmittable packets on the transmitting side is halved. In some situations, the value of cwnd may be four packets or less.

[0006] Here, as shown in FIG. 5, for example, if the cwnd on the transmitting side is four packets, and one of these packets is lost, on the receiving side, every time a packet is received, Aw
Since the CK packet is returned, AC that can be received by the transmitting side
K packet is duplicate ACK with first ACK packet 51
Two packets 52 and 53 make a total of three, three duplicate ACK packets are not received, and the Fast Retransmit operation is not performed. Therefore, there is a problem that transmission is interrupted until a timeout is detected.

[0007] The number of duplicate ACK packets received by the transmitting side is determined by the number of packets that have arrived at the receiving side since the packet was lost.
It may change depending on the status of ACK packet transmission from the receiving side to the transmitting side. Therefore, the sender's cwnd
Is 5 or more, the three overlapping ACs
There is a problem that the K packet is not received and the transmission may be interrupted.

[0008] The present invention has been made to solve such a problem, and is provided with a Ren supporting a SACK option.
An object of the present invention is to provide a retransmission control method capable of efficiently performing data transfer even in a situation where packet loss frequently occurs while suppressing the influence of erroneous retransmission due to a change in packet order in o-TCP. I have.

[0009]

In order to achieve the above object, a retransmission control method according to the present invention provides a retransmission control method in which a transmitting node is notified of a packet loss by selective acknowledgment information, The notified lost packet is immediately retransmitted only when the window size cwnd is equal to or smaller than a predetermined threshold.

[0010]

Next, the present invention will be described with reference to the drawings. FIG. 1 is a flowchart illustrating a retransmission control method according to an embodiment of the present invention, FIG. 2 is a sequence diagram illustrating an example of a retransmission control operation, and FIG. 3 is a block diagram illustrating a communication system to which the retransmission control method is applied. In FIG.
The transmitting node 1 and the receiving node 2 are connected to a communication network (network) 3, respectively, establish a predetermined connection in the communication network 3, and perform packet-based data transfer via the connection.

In the transmitting node 1 and the receiving node 2, the application processing unit 5 is a processing unit that executes a higher-level application. The layer 4 processing unit 6
Is a processing unit that performs data transmission / reception processing based on connection setting release, flow control, retransmission control, and the like. The lower layer processing unit 7 performs layer 2 processing such as frame assembly / disassembly.
And a layer 3 process such as routing.

The flow control of the TCP (protocol layer 4) executed by the layer 4 processing unit 6 is a window system, and a reception window size adw declared as a packet receivable amount by a communication partner as a congestion control parameter.
In addition to nd, the transmission node 1 has a congestion window size cwnd indicating the packet transmission amount reflecting the transfer status of the communication network 3. Send to

In the present invention, it is assumed that the TCP SACK option is used. In the TCP connection setting process executed by the layer 4 processing unit, the connection between the transmitting node 1 and the receiving node 2 is set. , SACK option is mutually confirmed. Note that the SACK option means that when a loss of a data packet is detected by the receiving node 2, SACK information (selective acknowledgment information) indicating data packets received after the lost packet is converted to an ACK (reception acknowledgment) packet. This is a function of storing and notifying the transmitting node 1.

The operation of the present invention will be described below with reference to FIGS. In FIG. 2, first, data packets (DT1 to DT1)
T4) is transmitted. In the receiving node 2, the transmitting node 1
Each time a data packet is received from, the transmission sequence number stored in the data packet is detected,
This is stored in an ACK packet as a reception sequence number and returned. Therefore, when DT1 is received, AC
K packet (ACK1 indicating an ACK packet of DT1)
21 is returned.

If DT2 is lost, the receiving node 2 receives DT3 next to DT1. And SA
As the CK information, the transmission sequence number indicating the data packet DT3 received after the packet loss is stored in the ACK packet 22 and returned. At this time, as the reception sequence number of the ACK packet 22, a sequence number indicating the delivery confirmation of the last data packet DT1 that has been continuously and normally received is set.

On the other hand, the transmitting node 1 confirms the reception of the data packet DT1 by the ACK packet 21.
Then, in the next ACK packet 22, based on the SACK information stored in the ACK packet 22, a packet loss, that is, a discontinuous portion (hole) of the received data packet is detected, and the retransmission control process shown in FIG. I do.

First, it is determined whether or not cwnd upon detection of a hole is equal to or smaller than a predetermined threshold th (step 11). If cwnd is equal to or smaller than the threshold th (step 11: YES), retransmission processing is immediately executed (step 11). 12). In this retransmission process, a data packet corresponding to the detected hole, here DT2, is retransmitted to the receiving node 2.

If cwnd at the time of detecting a hole is larger than a predetermined threshold value th (step 11: NO),
It is determined whether the value of the overlap hall counter is equal to or less than a specified value (step 13). If the value is equal to or smaller than the specified value (step 13: YES), the overlapping hole counter is incremented (added) by the amount of the loss notified by the SACK information (step 14).

On the other hand, if the duplicated hole counter is a specified value (for example, based on the idea of Fast Retransmit, the specified value =
If 3) is exceeded (step 13: NO), retransmission processing is performed in the same manner as step 12 described above (step 1).
5). By this means, when the cwnd value is larger than the threshold value, the same retransmission control as that of packet loss detection using duplicate ACK packets can be realized even when SACK information is used. In the retransmission processing (steps 12 and 15), transmission of the corresponding packet, cancellation of the timeout monitoring timer, clearing of the duplicate hole counter, and the like are performed.

As described above, when packet loss is detected based on the SACK information notified from the receiving node 2, the lost packet is retransmitted only when the cwnd value is equal to or less than the threshold value. Therefore, if the cwnd value is small and it is not possible to transmit a data packet that can obtain the number of ACK packets required for retransmission necessity determination by the duplicate ACK packet of Fast Retransmit, or if the ACK packet itself is lost and the retransmission necessity is determined by the duplicate ACK packet Even if the required number of ACK packets is not obtained, lost packets can be immediately retransmitted without waiting for activation of retransmission control due to timeout.

In TCP that does not support the SACK option, only the last part of continuous data is known, and it is not possible to determine which part is lost after that. However, it is not efficient to retransmit all subsequent parts. Absent. However, since the number of lost packets is limited by the SACK information, even if retransmission is performed erroneously, the retransmission efficiency does not significantly decrease.

[0022]

As described above, the present invention relates to the case where packet loss is notified by the selective acknowledgment information (SACK information) and the congestion window size cw
Only when nd is equal to or less than a predetermined threshold, the notified lost packet is immediately retransmitted. Therefore, in Reno-TCP supporting the SACK option, while suppressing the influence of erroneous retransmission due to the change of the packet order, Even in a situation where packet loss frequently occurs, data transfer can be performed efficiently.

[Brief description of the drawings]

FIG. 1 is a flowchart illustrating a retransmission control method according to an embodiment of the present invention.

FIG. 2 is a sequence diagram showing an example of a retransmission control operation according to the present invention.

FIG. 3 is a block diagram showing a communication system to which the retransmission control method of the present invention is applied.

FIG. 4 is a sequence diagram showing an example of a conventional retransmission control operation.

FIG. 5 is a sequence diagram showing another conventional retransmission control operation example.

[Explanation of symbols]

DESCRIPTION OF SYMBOLS 1 ... Transmission node, 2 ... Receiving node, 3 ... Communication network, 5 ... Application processing part, 6 ... Layer 4 processing part, 7 ... Lower layer processing part.

Continuation of the front page (72) Inventor Yukio Atsumi 5-2-2 Toranomon, Minato-ku, Tokyo F-term in the ultra-high-speed network computer technology laboratory (reference) 5K014 AA01 EA08 FA03 5K030 GA03 HA08 LA02 LC03 MB13 5K034 AA05 EE11 HH01 HH02 HH06 HH11 KK28 MM03 MM16 NN26

Claims (1)

[Claims] 1. A congestion window size cwn obtained by estimating a transfer capability of a network between a transmitting and receiving node connected via a predetermined connection by a layer 4 protocol.
transmitting a data packet from the transmitting node based on d.
In response to receiving the data packet, the receiving node returns an ACK packet indicating the reception acknowledgment,
In a communication system in which selective acknowledgment information (SACK information) indicating a data packet loss situation is notified by an ACK packet, the transmission node may be configured to notify the packet loss by the selective acknowledgment information and to determine whether the packet loss has occurred. A retransmission control method characterized by immediately retransmitting a reported lost packet only when the size cwnd is equal to or smaller than a predetermined threshold.