JP2002374302A - Rtt-measuring method and system - Google Patents

Abstract

PROBLEM TO BE SOLVED: To realize a RTT-measuring method, capable of efficiently measuring an RTT value, taking into consideration for the presence/absence of congestion. SOLUTION: The RTT value which the time for transmitting and receiving a packet is measured by using a retransmission request packet to be transmitted toward an upstream node in the case of a data packet loss in a network, containing a plurality of nodes and a downstream direction packet (a retransmission request suppressing packet or a retransmission data packet) to be transmitted to a downstream node in response to the packet. A receiver node, detecting the loss of a data packet transmits the retransmission request packet including a time stamp value and the address of its own node to the upstream node by a unicast. And the upstream node receiving the retransmission request packet transmits the downstream direction packet including the time stamp value and the address to the downstream node by a multicast. Based on the time stamp value included in the downstream direction packet, the receiver node measures the RTT value.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

The present invention relates to an RTT measurement method and an RTT measurement system, and more particularly to an RTT (Round Trip) in a network including a plurality of nodes.
Time) and a measurement system.

[0002]

2. Description of the Related Art In a best-effort network such as the existing Internet, a receiver needs to respond to reception in order to ensure reliability.
In the unicast communication, an acknowledgment (ACK) is generally transmitted. On the other hand, in multicast communication, there is a plurality of receivers for a packet transmitted by one sender, and when all the receivers transmit ACKs, a problem occurs in which ACKs are concentrated on the senders and scalability is reduced. . For this reason, in multicast communication, it is common to transmit a retransmission request (Negative ACK; hereinafter, abbreviated as NAK) only when a receiver cannot receive a packet.

[0003] A NAK packet is transmitted from a receiver to a sender. However, if there is congestion or the like in the route to the sender, the NAK packet may be lost and may not reach the sender. Therefore, after transmitting the NAK packet, the receiver needs to retransmit the NAK packet if there is no response for a certain period of time. The time interval before the retransmission is set based on the RTT, which is the round-trip time between the sender and the receiver, so that the receiver can use the RT from his or her own to the sender.
You need to know the T value.

If the time interval until retransmission is short, even if the sender transmits retransmission data, the NAK packet will be retransmitted before it reaches the receiver. Then
The sender performs retransmission again, and sends a useless packet to the network. Also, if the time interval before retransmission is long, the time interval until the next retransmission will be long if the NAK packet for the sender is lost or if the data retransmitted by the sender is lost. Therefore, it takes time to repair the lost packet. Therefore, if the RTT value can be accurately estimated, the efficiency of the reliable multicast communication can be improved.

[0005] Also, the NA for the same data packet is
Only one K packet needs to be sent to the sender, and when many receivers send NAK packets at the same time for the same data packet, useless NA to the sender is lost.
K concentration occurs. For this reason, the receiver waits for a random time before transmitting the NAK packet, shifts the transmission time of the NAK packet for each receiver, and
Send the packet to all recipients and senders other than yourself by multicast. This allows the recipient to
A technique has been proposed in which a NAK packet is not transmitted when a NAK packet from another receiver is received before transmitting the packet (NAK Suppression: NAK suppression). The random time to wait at this time is generally set based on the RTT value of the recipient having the largest RTT value in the group of recipients (the RTT value of the farthest recipient).

[0006] Hereinafter, multicast,
Reliable multicast, network node (router)
Reliable multicast with support, NAK aggregation, N
AK suppression, local recovery processing, NAK suppression and NAK
Simultaneous use with aggregation, RT in reliable multicast
The method of measuring the T value will be described. (Multicast) Multicast is a technology for performing 1: n or n: n communication. When the sender distributes the same data to n recipients, instead of delivering it to n recipients n times, once the transmission is performed, the multicast router in the network copies the data, Deliver to recipients.

(Reliability Multicast) Since the Internet is a best-effort type network (which makes an effort to deliver data to a destination, but does not guarantee reliability), even if data is transmitted, the sender cannot receive data. It is not always possible to receive data reliably. Reliable multicast is a technology that guarantees that data is reliably delivered to all recipients in multicast communication. This includes a method of returning an acknowledgment to the sender when the receiver has successfully received the data, and the continuity between the data and the previously received data when certain data is received. When there is no data, there is a method of detecting a loss of data and transmitting a NAK packet requesting retransmission of the data. When transmitting ACK, all receivers return ACK for all received data. For this reason,
Since a large amount of ACK is returned to the sender, the method using NAK packets generally has higher scalability.

(Reliable Multicast with Network Node (Router) Support) A multicast router duplicates data at the time of distribution from a sender to a receiver, and transmits one data transmitted by the sender to a plurality of receivers. The data is distributed, but only the NAK packet is relayed on the path from the receiver to the sender, and the data is not duplicated.
On the other hand, there has been proposed a method of efficiently performing the reliable multicast distribution by causing a router to perform processing even on a path from a receiver to a sender. Specifically, as the processing performed by the router, it is sufficient to send only one NAK packet to a sender for one data from a plurality of receivers. The router forwards the NAK packet, and NAK Aggregation (NAK aggregation) processing that the router discards for the duplicate NAK packet that comes later, or saves the data transmitted from the sender in the cache, and the NAK from the receiver is stored in the cache. There is a local recovery process for performing retransmission on behalf of the sender when it arrives. It is known that scalability is improved by performing these operations. In addition, there is a concept that a router has a function to perform these processes, and a concept that a server device is arranged adjacent to the router and the server device performs these processes. A description will be given as a node device (hereinafter, referred to as a node).

[0009] (NAK aggregation) AC to ensure reliability
Even when a NAK packet is used instead of K, network congestion generally occurs in a router queue. For this reason, if some data is discarded in the router queue, all the recipients delivered via the router will detect the same data loss, and all those recipients will send NAK packets, Many N for the same data
The AK will reach the sender. When the sender receives a retransmission request for data, the sender retransmits the data. However, when a plurality of NAKs for the same data arrive, the sender retransmits the first NAK packet. However, if retransmission is performed also on a NAK packet that arrives later, retransmission is duplicated and a useless packet is transmitted to the network. Therefore, retransmission of data is not performed for the duplicate NAK packet, and the duplicate NAK packet is discarded.

[0010] When congestion occurs in a router queue located close to the sender, many receivers detect the loss of the same data and transmit NAK packets, resulting in a large amount of useless NAKs. However, it is conceivable that the band of the link toward the sender is compressed and the scalability is reduced. To solve this,
In the node device (router) in the network, the NA
A method of integrating K, reducing the amount of NAKs reaching the sender and improving scalability has been proposed, which is called NAK aggregation.

When the node device receives the NAK packet for the data, if the NAK is the first NAK for the data, the node device forwards the NAK packet to the sender, otherwise discards the NAK packet. Will do. In order to realize this processing, the node device holds a packet sequence number and a counter value indicating how many times the NAK packet has been transferred upstream. Here, it is necessary to hold the counter value indicating the number of transfers for the following reason. That is,
Even if a NAK packet is transmitted once, the data may drop again. When a NAK packet is transmitted, a timer is set, and if no data arrives before the timer expires, the NAK packet is transmitted again. This is because, in order to perform the processing, in order to perform the NAK aggregation, it is necessary to count the number of times the node device has transmitted the NAK packet. The timer value applied at this time is generally based on the RTT value between the receiver and the sender.

Each receiver also records how many NAK packets have been transmitted for the missing data, as in the node device. When transmitting a NAK packet, a count value indicating the number of the NAK packet is set to NA
It is transmitted in K packets. The node device compares the NAK count value of the NAK packet received from each receiver with the NAK count value held by the node device. If the NAK count value of the NAK packet from the receiver is larger, it is sent to the sender. It does the transfer. On the other hand, if the NAK count of the NAK packet is equal to or smaller than the NAK count held by itself, it is discarded because it has already been transferred. Figure 6 shows NA
The processing at the node at the time of K aggregation is shown. In the figure, processing is performed as follows at a node at the time of NAK aggregation. (1) The sender T transmits data one by one in order, and the node device (multicast router) duplicates the packet,
Delivery is made to the recipients R1 and R2, respectively. (2) Upon receiving the data, each receiver checks for any missing previous data, and if there is any missing data, requests the node device N to retransmit with NAK. It should be noted that, in the figure, the shaded portion is a portion where data is missing. (3) At the time of NAK transmission, the receiver R1 sets a timer,
If no data is sent before the timer expires,
Retransmit the NAK packet. (4) Upon receiving the NAK packet from the receiver, the node checks whether the NAK is the NAK of which number. If the NAK is the first NAK of that number, the node transmits the NAK to the sender, otherwise discards it. (5) In the above example, the receiver R1 transmits a NAK packet for data 3 and 5, and the receiver R2 transmits a NAK packet for data 2 and 5, so that the node N has a total of four NAKs. However, the node N forwards the NAK packet for the data 5 to the sender T by transmitting the NAK packet from the receiver R1 that arrives first, and discards the NAK packet from the receiver R2 that arrives later.
K integration is performed.

(NAK suppression) Only one NAK packet for the same data packet needs to be delivered to the sender, and when many receivers transmit NAK packets for the same data packet at the same time, Useless NAK concentration on the sender occurs. Therefore, the recipient is NAK
Wait a random time before sending a packet, N
The transmission time of the AK packet is shifted for each receiver, and the receiver transmits the NAK packet by multicast to all the receivers other than itself and the sender. When a NAK packet is received from the server, the NAK packet is not transmitted. This technique is NAK suppression. By using this, the sending of the NAK of the receiver can be suppressed,
Concentration of NAKs on the sender can be avoided. As the random time to wait at this time, the RTT value of the receiver having the largest RTT value in the receiver group (RTT value of the farthest receiver)
It is generally set based on (TT value). FIG. 7 shows the NAK suppression processing. In the figure,
The NAK suppression processing is performed as follows. (1) DATA (2) is lost between the sender T and the node N (loss in the sending queue of the sender), and the receivers R1 and R2
When both receive DATA (3) or later, DATA (2)
To detect the loss. (2) Both the receivers R1 and R2 individually and individually set a timer, and wait until the timer expires to transmit NAK (2). (3) The timer of the receiver R1 expires first, and the receiver R1 transmits NAK (2) by multicast. (4) Node N is NAK which is a multicast packet
(2) is duplicated and distributed to the sender T and the receiver R2. (5) Since the receiver R2 receives NAK (2) before the timer expires, it does not transmit NAK (2). (6) The sender T retransmits DATA (2). (7) Since the number of NAKs (2) reaching the sender T is reduced to one, concentration of NAKs can be avoided.

(Local Recovery Processing) In the node device, data distributed from the sender is stored in a cache, and when a NAK from the receiver arrives, the node performs distribution on behalf of the sender, The time required for recovery can be reduced. Also, send a NAK to the sender
It is recorded whether or not the packet has been transmitted, and since the data for which the NAK packet has not been transmitted does not need to be distributed to the subordinate receivers, the process of discarding the data is also performed. Can also be prevented. Since the amount of data that can be stored in the node device is limited, the cache method is FIFO (First I / O).
n First Out), there are a method of caching only retransmitted data and a method of caching all data transmitted by the sender. The process performed by the node is NAK from the subordinate
When a packet is received, it checks whether there is data in its own cache, if it exists, distributes it, otherwise records that the NAK packet has been transferred upstream, and records the NAK packet upstream Is transferred. When the retransmitted data arrives, it checks whether the NAK packet for the data has been transferred upstream. If the NAK packet has been transferred before, the data is distributed by multicast to the subordinate receivers. If not, the data is discarded. At this time, if necessary, a process of caching data is performed in order to respond to the NAK packet again. The local recovery processing will be described with reference to FIG. In the figure, the local recovery process is performed as follows. (1) The receiver R1 detects that DATA (2) is missing and transmits NAK (2). (2) The node N2 transfers NAK (2) upstream since DATA (2) is not stored in the cache. (3) The node N1 multicasts DATA (2) because DATA (2) is stored in the cache. (4) Since the node N2 has a record of transferring NAK (2) upstream, it multicasts DATA (2) under its control. (5) Since the node N3 has not forwarded NAK (2) upstream, it discards DATA (2) without transmitting it under it. (6) The receiver R1 receives DATA (2). (7) Since the receiver R2 already has DATA (2) and has not transmitted NAK (2), the receiver R2 only discards it even if it receives it. (8) Here, the reason why the node device multicasts DATA besides the receiver to which the NAK has arrived is that, due to NAK suppression, it is necessary to transmit DATA even to the receiver that did not transmit the NAK packet. That's why.

(Simultaneous Use of NAK Suppression and NAK Aggregation)
For simplicity, assume a tree-like multicast group with one sender and multiple recipients. When each receiver detects the loss of data, it transmits a NAK packet to the sender by unicast. When a node in the network receives a NAK packet, it determines by NAK aggregation whether it should be forwarded to the sender. When forwarding NAK packets to the sender,
In order to prevent other recipients from sending any more NAK packets, NAK packets are transmitted to subordinate recipients by multicast. Multicast NAK
The recipient of the packet receives a NAK for the data
The timer for stopping transmission of the packet and retransmitting the NAK packet is reset. By performing the same processing as that of the receiver for the node devices arranged in a tree shape, the number of NAKs transmitted to the network can be reduced, and scalability is improved. In the case where only NAK suppression is performed, the receiver always sends a NAK packet by multicast after waiting for a random time. It may happen that a multicast NAK packet is transmitted. This often occurs when the distance between the recipients is long, and the like, and although the NAK is transmitted in the network, the recipient has not yet reached.

This problem will be described with reference to FIG. FIG. 5 shows a problem when only NAK suppression is performed. (1) Assume that the receivers R1, R2, and R3 detect the same data loss. (2) Each receiver individually sets a timer at random time based on the RTT value up to the sender. (3) The timer of the receiver R1 expires and the receiver R1 transmits a NAK packet by multicast. (4) The receiver R2 receives the NAK packet before the timer expires, and stops transmitting the NAK packet. (5) The receiver R3 cannot receive the NAK packet before the timer expires, and transmits the NAK packet by multicast after the timer expires. (6) Since the receiver R3 receives the NAK packet transmitted by multicast after transmitting the NAK packet, the effect of NAK suppression does not work. (7) The receiver R2 and the sender T receive N for the same data.
As a result, two AK packets are received, and wasteful packets flow in the network.

However, in the method in which NAK suppression and NAK aggregation are simultaneously used, transmission is performed by unicast to a router in the network, and the router needs to transfer the NAK packet to the sender, or Multicast N for NAK packets that have never been forwarded
NAK only when AK packet is received from upstream
If a router on the way has previously received a NAK packet from another receiver in order to multicast-distribute the packet under it, the router can forward the NAK packet to the sender or multicast NAK under it. No packets will be sent. For this reason, even if a receiver sends a NAK packet while a NAK packet is flowing through the network, the router sends the NAK packet.
There is an advantage that packets are discarded and useless NAK packets are not spread. When NAK suppression and NAK aggregation are used at the same time, the value of the random timer set at the time of NAK suppression is set based on the RTT value of the farthest receiver under the control of the immediately upstream node device. The retransmission timer is based on the RTT value up to the sender. FIG.
0, simultaneous use of NAK suppression and NAK aggregation will be described. In the figure, in the case of simultaneous use of NAK suppression and NAK aggregation, processing is performed as follows. (1) In the process of FIG. 10, the same state as in FIG. 9 is assumed, but each node records whether a NAK packet has been received and whether or not a NAK packet has been transferred. (2) The receiver R1 transmits a NAK packet to the node N2, and the node N2 transmits the NAK packet to the receivers R1 and R2 by multicast, and transmits the NAK packet to the node N1 by unicast. (3) Since the receiver R2 receives the NAK packet before the timer expires, the receiver R2 stops transmitting the NAK. (4) Upon receiving the NAK packet from the node N2, the node N1 transmits a NAK packet to the sender by unicast and to the nodes N2 and N3 by multicast. (5) The receiver R3 cannot receive the NAK packet before the timer expires, and transmits the NAK packet to the node N3 by unicast after the timer expires. (6) The node N3 receives the NAK packet from the receiver R3, but does not transmit the NAK packet to the node N1 because the node N3 receives the NAK packet distributed by multicast from the node N1 and distributes it under the node. (7) The sender and receiver R2 receive the NAK packet only once, and NAK suppression works well.

(Method of Measuring RTT Value in Reliable Multicast) As described above, since it is the receiver that needs to know the RTT value, the RTT value is measured mainly by the receiver. The receiver periodically transmits a packet for RTT measurement to the sender. First, it is assumed that there is only a sender and a receiver. In this case, the receiver transmits a packet to which the transmission time is added as a time stamp to the sender by unicast. Upon receiving the packet, the sender adds the time stamp value of the received packet to the reply packet and transmits the packet. The receiver calculates the difference between the time stamp value added to the returned packet and the reception time,
The RTT value can be estimated.

For NAK suppression, all receivers need to know the RTT value of the receiver having the largest RTT value in the receiver group. For this purpose, the receiver adds the RTT value obtained in the previous RTT measurement to the RTT measurement packet and transmits it to the sender. The sender holds the RTT value of the receiver having the largest RTT value, and transmits the RTT value transmitted from each receiver.
When receiving the TT packet, the RTT value added to the packet is compared with the held RTT value, and if the RTT value added to the packet is larger, the held information is updated. Then, the sender adds the held RTT value to the returned RTT measurement packet and transmits the packet.

Thus, upon receiving the RTT measurement packet, the receiver can know both its own RTT value and the largest RTT value among all the receivers. In this method, since the RTT value of the information held by the sender is not updated with a small value, the sender always holds the RTT value in a bad state. Since this has a problem, it is necessary to prevent the held RTT value from remaining in a bad state. For this, the address of the recipient having the largest RTT value is recorded, and the RT from the recipient is recorded.
For the T value report, a method of updating even a small value or a method of updating after a certain period of time even if the reported RTT value is a small value is used.

The method of measuring the RTT will be described with reference to FIG. FIG. 3 shows a process in the case where the RTT measurement is performed only between the sender and the receiver. (1) Each of the receivers R1, R2, and R3 transmits an RTT measurement packet to the sender T at random. To the RTT measurement packet, the time stamp of the transmission time and the RTT value measured up to the sender last time are added as data. (2) Upon receiving the RTT measurement packet, the sender T receives the RTT value added to the packet and the held Dow.
Compare with n Worst RTT value. If the RTT value added to the packet is worse, Dow
Update n WorstRTT value. In addition, Down
Worst RTT value is the worst RTT value going downstream. (3) The sender T returns a response packet to the receivers R1, R2, R3. The reply packet contains the received R
The time stamp of the TT measurement packet and DownWor
The st RTT value is entered as data. (4) Upon receiving a response packet to the RTT measurement packet, the receivers R1, R2, and R3 determine their own R based on the difference between the time stamp value added to the packet and the reception time.
The TT value is estimated and Down added to the packet
UP Wors held by Worst RTT value
Update the t RTT value. In addition, UP Worst R
The TT value is the worst upstream RTT value.

In the above case, the RTT value is measured between the sender and the receiver. When performing NAK suppression or NAK aggregation, the node devices in the network are also NAK
Since the packet is transmitted, it is necessary to know the RTT value up to the sender, and the receiver and the node device are the receivers under the upstream node device and receive the largest RTT value from the node device for NAK suppression. Since it is necessary to know the RTT value of the receiver or the node device, the RTT measurement exchanged between the receiver and the sender is performed in a tree shape.
As a result, each node device can know the RTT parameters necessary for performing NAK suppression and NAK aggregation, and also reduce the number of RTT measurement packets flowing through the network.

In particular, when measuring the RTT value only between the receiver and the sender, the sender must process the RTT measurement packets from all the receivers. When performing the RTT measurement, only the RTT measurement packet from the node device immediately below or the receiver needs to be processed, and the load on the sender is reduced. In order to perform this method, each node device needs to hold the RTT value of the receiver having the largest RTT value among the subordinate receivers.

Also, each node adds the RTT value from itself to the sender to the response packet to the RTT measurement packet, so that the receiver can estimate the RTT value from the sender. Here, Table 1 and FIG.
2, a method of performing the RTT measurement in a tree shape will be described. Table 1 is a table showing the contents of the RTT measurement packet and the response packet when the RTT measurement is performed between the sender and the receiver and when the RTT measurement is performed using the tree structure.

[0025]

[Table 1]

As shown in Table 1, when the method of measuring the RTT value is only between the transmitter and the receiver, the RTT measurement packet includes the time stamp of the transmission time and the RTT value from itself to the sender as data. The response packet includes the time stamp of the RTT measurement packet and D
own Worst RTT value is included as data. When the method of measuring the RTT value uses a tree structure, the RTT measurement packet includes the time stamp of the transmission time and the RTT value from itself to the upper node (UP
RTT) is included as data, and the response packet includes the time stamp of the RTT measurement packet and Down.
Worst RTT value and R from itself to sender
The TT value is included as data.

FIG. 12 is a diagram for explaining RTT measurement using a tree structure. (1) Recipients R1, R2, R3 and nodes N1, N
2 and N3 randomly transmit the RTT measurement packet to the upper node. The RTT measurement packet includes the time stamp of the transmission time and the UP
Enter the RTT value. (2) The sender T and the nodes N1, N2, N3
When a measurement packet is received, the RTT value added to the packet and the Down Worst R stored therein are stored.
Compare the TT value, and if the RTT value of the packet is worse, update the Down Worst RTT value. (3) The sender T and the nodes N1, N2, N3
A response packet is returned to the sender of the measurement packet. To the reply packet, the time stamp of the received RTT measurement packet, the Down Worst RTT value, and its own RTT value (0 in the case of the sender T) are added as data. (4) Receivers R1, R2, R3 and nodes N1, N2,
Upon receiving the response packet to the RTT measurement packet, N3 estimates the UP RTT value between itself and the upper node from the difference between the time stamp value added to the packet and the reception time, and adds the packet to the packet. The RTT value from itself to the sender T is added to the RTT value from the upper node to the sender T. Also, the UP Worst RTT value held by the Down Worst RTT value added to the packet is updated.

The conventional RTT measurement described above shows that the receiver
In order to measure the TT value, a packet including the transmission time of the packet as a time stamp is transmitted to the sender, and the sender receiving the reply adds the time stamp to the response packet and returns the response packet. The RTT value is estimated from the difference between the reception time and the time stamp value. Usually, this procedure is performed at regular intervals. The receiver sets the RTT value estimated in the previous RTT measurement to the RT value.
By transmitting the packet by adding it to the packet for T measurement, the sender estimates the RTT value of the farthest receiver in the receiver group, and adds that value to the response packet.
The receiver receiving the response packet sets its own RTT value,
The RTT value of the farthest recipient can be known.

[0029]

Generally, when congestion occurs in a certain part of the network, the RTT is temporarily set.
The value becomes very large. In this case, in the above-described conventional method, the receiver can only measure the RTT value at regular time intervals using the RTT measurement packet.
Accurate RTT of own network following dynamic changes
It is difficult to estimate the value. Similarly, even when the congestion is resolved and the RTT value is reduced, there is a time interval before the RTT value is measured, and it is difficult to follow an accurate RTT value.

In order to solve this problem, it is considered that the estimated interval time of the RTT value should be shortened. However,
Reducing the measurement time interval means that more packets are sent out over the network, which increases the overhead. Further, even in a state where there is no congestion and there is not much change in the RTT value, the RTT measurement packet is transmitted, and there is a problem that useless packets flow in the network and network resources cannot be effectively used.

Further, the RTT measurement only between the transmitter and the receiver has a problem in that the processing load of the RTT packet is imposed on the transmitter. When the RTT measurement is performed hierarchically using a tree structure, the load on the sender is reduced, but the RTT measurement is performed randomly by each node and receiver, so that congestion may occur on a certain link. When the RTT value greatly changes, it may take time for the information to be hierarchically transmitted to the receiver.

This problem will be described with reference to FIG. The figure shows a measurement delay in RTT measurement using a tree structure. (1) As shown in FIG. 13, the RTT measurement interval is 2
Seconds, each link delay without congestion is 10 ms,
The initial measurement time randomly set by each node and the receiver is 0.5 seconds for the receiver, 1.5 seconds for the node N1, and 1.0 seconds for the node N2. (2) Congestion occurs at one second between the sender T and the node N1, and the link delay changes from 10 ms to 50 ms. (3) Since the time required for the RTT measurement is smaller than the measurement interval and is ignored, when there is no congestion, the node N
The RTT values of 1, N2 and R are 10 ms and 2 respectively.
0 ms and 30 ms. (4) After 1 second, congestion occurs between the sender T and the node N1, but the node N1 can know it.
Since it is every RTT measurement interval, it is 1.5 seconds later. (5) The node N2 measures the RTT value at the time of 1.0 second, but at that time, the node N1 has not yet detected the congestion. 3.0 seconds after the second measurement. (6) The receiver R performs the second measurement after 2.5 seconds,
Since the node N2 has not detected congestion at that time, the receiver R can know that the RTT value has changed due to the congestion after 4.5 seconds, which is the third measurement. (7) It turns out that a time lag of 3.5 seconds has occurred before the congestion information that has occurred at the time of 1 second reaches the receiver, and a measurement delay larger than the measurement interval has occurred. (8) This delay is caused by RT in order from the node close to the sender T.
The problem can be solved by performing the T measurement. However, it is difficult for the downstream node to know when the RTT value has been measured at the upstream node, and an appropriate measurement time is set for each node N
1, N2 and receiver R are difficult to set, and even if set properly, a delay greater than the RTT measurement interval can occur.

The present invention has been made to solve the above-mentioned drawbacks of the prior art, and an object of the present invention is to provide an RTT measurement method capable of more efficiently measuring an RTT value in consideration of the presence or absence of congestion. And to provide an RTT measurement system.

[0034]

An RTT measurement method according to claim 1 of the present invention provides a retransmission request packet transmitted to an upstream node when a data packet is lost and a response to the packet in a network including a plurality of nodes. RTT measurement method for measuring the RTT value, which is the round trip time of packet transmission and reception using the downstream packet transmitted toward the downstream node, the receiver node that has detected the loss of the data packet, The retransmission request packet including the time stamp value and the address of the own node is transmitted to the upstream node by unicast, and the upstream node that has received the retransmission request packet downstreamly transmits the downstream packet including the time stamp value and the address by multicast. To the node and, at the receiver node, included in the downstream packet And measuring the RTT value based on the time stamp values are.

[0035] The RTT measuring method according to claim 2 of the present invention is characterized in that, in claim 1, the downstream packet is a retransmission request suppression packet transmitted from an upstream node for retransmission request suppression. Claim 3 of the present invention
2. The RTT measurement method according to claim 1, wherein the downstream packet is a retransmission data packet transmitted from a sender node that is a source of the data packet in response to the retransmission request packet. .

The RTT measuring method according to claim 4 of the present invention is the method according to any one of claims 1 to 3, wherein the upstream node transmits the latest packet from its own node to the sender node when transmitting the downstream packet. The node that has added the RTT value and received the packet sets the RTT value from its own node to the sender node with reference to the RTT value and the RTT value from its own node to the upstream node. It is characterized by having done.

The RTT measuring method according to claim 5 of the present invention is the method according to any one of claims 1 to 4, wherein the upstream node transmits the downstream packet from its own node to the farthest node immediately below it. The node that has added the RTT value and received the downstream packet, refers to the RTT value, and checks the RT from the upstream node to the farthest node immediately below it.
It is characterized in that a T value is set.

[0038] In the RTT measuring method according to claim 6 of the present invention, in any one of claims 1 to 5, the upstream node transmits only the earliest retransmission request packet upstream, and transmits the other retransmission request packets upstream. It is characterized by being discarded. The method for measuring RTT according to claim 7 of the present invention comprises:
The upstream node according to any one of claims 1 to 6,
It is characterized in that data to be distributed is held, and when the retransmission request packet is received, local recovery for distributing a data packet is performed instead of the sender node.

An RTT measurement system according to claim 8 of the present invention provides a retransmission request packet transmitted to an upstream node when a data packet is lost, and a retransmission request packet transmitted to a downstream node in response to the packet in a network including a plurality of nodes. An RTT measurement system for measuring an RTT value, which is a round trip time of packet transmission and reception, using a downstream packet transmitted by a wireless communication apparatus, wherein when a loss of a data packet is detected, a time stamp value and an address of the own node are used. And a receiver node for transmitting a retransmission request packet including the timestamp value and the address to the downstream node by multicast when receiving the retransmission request packet. A downstream direction at the recipient node. And measuring the RTT value based on the time stamp value contained in packets.

[0040] The RTT measuring system according to claim 9 of the present invention is characterized in that, in claim 8, the downstream packet is a retransmission request suppression packet transmitted from an upstream node for retransmission request suppression. The RTT measurement system according to claim 10 of the present invention is characterized in that in claim 8,
The downstream packet is a retransmission data packet transmitted from a sender node that is a transmission source of the data packet in response to the retransmission request packet.

The RTT measurement system according to claim 11 of the present invention is the RTT measurement system according to any one of claims 8 to 10, wherein the upstream node transmits the latest packet from its own node to the sender node when transmitting the downstream packet. The node that has added the RTT value and received the packet receives the RTT value,
The RTT value from the own node to the sender node is set with reference to the RTT value from the own node to the upstream node.

The RTT measurement system according to claim 12 of the present invention is the RTT measurement system according to any one of claims 8 to 11, wherein the upstream node transmits the downstream packet from its own node to the farthest node immediately below it. Add the RTT value,
The node that has received the downstream packet is the RTT
The RTT value from the upstream node to the farthest node immediately below it is set with reference to the value.

In the RTT measurement system according to claim 13 of the present invention, in any one of claims 8 to 12, the upstream node transmits only the earliest retransmission request packet upstream, and transmits the other retransmission request packets upstream. It is characterized by being discarded. The RTT measurement system according to claim 14 of the present invention is the RTT measurement system according to any one of claims 8 to 13, wherein the upstream node holds data to be delivered and, when receiving the retransmission request packet, the sender node Instead, local recovery for distributing data packets is performed.

In short, since the NAK packet is transmitted when the receiver detects a packet loss, there is a high possibility that network congestion has occurred at the time of transmitting the NAK packet. Therefore, in the present invention, in addition to the packet for estimating the RTT value periodically, the RTT is also used for the NAK packet.
Estimate the value. Accordingly, when congestion does not occur in the network, the RTT value is estimated periodically as in the related art, and when congestion occurs, the RTT value can be measured at shorter intervals.
Therefore, it is possible to send an unnecessary packet to the network and accurately estimate the RTT value without imposing a load.
It is possible to accurately estimate the AK packet retransmission timer interval, the NAK suppression timer interval, and the like.

[0045]

Next, an embodiment of the present invention will be described with reference to the drawings. In the drawings referred to in the following description, the same parts as those in the other drawings are indicated by the same reference numerals. As described above, when congestion occurs in the network and a packet loss occurs, the receiver detects the packet loss and transmits a NAK packet. That is, N
When the AK is sent, there is some change in the network,
It is highly possible that the RTT value has changed. Therefore, in the present invention, the RTT value is measured using the NAK packet.

When NAK aggregation is not performed, NAKs are exchanged only between the sender and the receiver. When the receiver detects a packet loss, the receiver transmits a NAK packet by multicast. At this time, a time stamp and a previously measured RTT value are added to the NAK packet to be transmitted. When each receiver receives a NAK packet transmitted by multicast, it does not change the processing, except that it stops transmitting the NAK packet. On the other hand, the sender is NAK
When the packet arrives, the RT added to the NAK packet
The T value is compared with the RTT value of the subordinate farthest receiver, and if necessary, the RTT value of the subordinate farthest receiver is updated. NAK
Since the data requested by the packet is retransmitted, the data packet to be retransmitted includes the time stamp added to the NAK packet, the sender address of the NAK packet, and the RTT value of the subordinate farthest receiver. The packet is added as data, and the data packet is transmitted to the multicast address.

Among the receivers that have received the retransmission data, the receiver that matches the NAK sender address added to the data packet estimates the RTT value and updates the worst RTT value upstream (UP Worst RTT). .
For other recipients, data recovery and worst-case upstream RTT
It only updates the value, but does not estimate the RTT value.

Thus, the receiver that has transmitted the NAK packet receives the retransmission data packet,
T value estimates can be updated, and other recipients
The RTT value serving as the reference for performing K suppression can be updated. The sender appends the address of the NAK sender to the data packet because the data packet
This is to indicate whether the packet corresponds to an AK packet. Also, unlike the RTT measurement packet transmitted by unicast and the response packet to the RTT measurement packet, the NAK packet and the retransmission data packet are transmitted by multicast.
This is because it is necessary to clearly indicate whether the response is to a packet. Note that the updating of the UP Worst RTT value can be performed for each receiver by comparing it at the time of receiving the NAK.

Unlike the RTT measurement packet, only the receiver that detects the packet loss due to congestion, etc., can perform the RTT measurement non-periodically during the period of congestion, rather than periodically. Follow R
The TT value can be measured. In addition, since the RTT value serving as a reference for NAK suppression can be uniformly and accurately known for all receivers, NAK suppression works efficiently. If the reference RTT value differs for each receiver, and NAK suppression is performed based on a small RTT value, the number of recipients who transmit a NAK packet before the NAK transmitted by multicast arrives increases, resulting in NAK suppression. Does not work efficiently, but this is not the case with this method.

Due to the effect of NAK suppression, a receiver whose transmission of NAK packets has been suppressed cannot measure the RTT value using NAK packets. However, receivers located under a congested link frequently experience packet loss due to congestion, transmit NAK packets more easily than other receivers, and have an opportunity to estimate the RTT value using NAK packets. Therefore, suppression of NAK transmission by NAK suppression is not so problematic.

The advantage of the above method is that a receiver who is considered to be in a poor state, that is, a fluctuation of the RTT value is large, is frequently used in the R system.
The TT value can be measured. On the other hand, a receiver in good condition will only measure the regular RTT value on a regular basis, and the time interval for dynamically estimating the RTT value according to the network condition will be set. The point is that it can be changed.
Further, since the estimation of the RTT value and the notification of the worst RTT value under the estimation are performed using the NAK packet and the retransmission data packet, the amount of packets transmitted to the network does not change compared to the conventional method. Therefore, the influence on the network is reduced by the overhead of about several bytes due to the addition of the data in the packet, and the RTT value can be measured efficiently.

Here, referring to FIG. 1, a method of estimating the RTT value using the NAK packet and the retransmission data packet when estimating the RTT value only between the sender and the receiver will be described. Table 2 is a table showing parameters added to the NAK packet and the retransmission data packet used when estimating the RTT value only between the transmitting and receiving parties.

[0053]

[Table 2]

Referring to Table 2, a NAK packet includes a request data number, a time stamp of a transmission time, and an RTT value from itself to a sender as data.
The retransmission data packet includes a data number and data,
Time stamp added to NAK packet, NA
K sender address and Down Worst RT
The T value is included as data.

FIG. 1 shows an embodiment of the RTT measuring system according to the present invention. (1) At a certain time, congestion occurs in the link between the node N1 and the node N2, and the RTT value between the links is 1
It is assumed that the time has changed from 0 ms to 25 ms and a packet loss has occurred. (2) The receivers R1 and R2 detect packet loss and
After waiting for a random time due to K suppression, the receiver R2 first transmits a NAK packet by multicast. At this time, the receiver R2 adds the previously measured RTT value (50 ms) and the time stamp to the NAK packet. (3) The receiver R1 receives the multicast NAK packet and suppresses transmission of the NAK packet.
Further, the sender receives the NAK packet, and adds the time stamp, the address of the receiver R2, and Down Worst RTT (60 ms) to the retransmission data packet, and transmits the packet. (4) The receiver R1 receives the data and repairs the packet loss. Upon receiving the data, the receiver R2 receives its own R2.
It updates the TT value and knows that the RTT value has reached 65 ms. The receiver R3 receives the data but discards it because it already has the data. At this point, the UP Worst RTT value has not been updated for all recipients. Recipient R
2 may update the UP Worst RTT value at this time, but since the same value cannot be unified for all recipients,
It is more desirable to perform this at the time of retransmission data reception. (5) When the congestion continues, packet loss occurs again, and the receivers R1 and R2 detect the packet loss several times. Due to the effect of NAK suppression, either one will transmit a NAK packet. (6) Assuming that the receiver R2 transmits the NAK packet again, the receiver R2 transmits the time stamp and the RT measured last time.
The T value (65 ms) is added to the NAK packet and transmitted. (7) The receivers R1 and R3 receive the NAK packet transmitted by multicast. Recipient R3 is a useless NAK
Therefore, the receiver R1 discards the N.
Suppress AK transmission. The receivers R1 and R3 may update the UP WorstRTT value at this point, but since the same value cannot be unified for all the receivers, it is preferable to update the value at the time of receiving retransmission data. (8) When the sender receives the NAK packet, the RTT value (65 ms) added to the packet is larger than the held Down Worst RTT value (60 ms). Update the Down Worst RTT value. (9) The sender sends a time stamp to the retransmission data,
2 and the Down Worst RTT value. (10) The receivers R1 and R2 receive the retransmitted data and repair the data. The receiver R2 estimates the RTT value, and all the receivers receive the Down W added to the packet.
The UP Worst RTT value is updated with the orst RTT value. (11) At this time, the receiver R1 has not been able to measure an accurate RTT value, but if the receiver R1 transmits a NAK packet, the receiver R1 can also measure an accurate RTT value. (12) Since the receiver R3 in a good state does not transmit a NAK packet, the measurement of the RTT value is regular. On the other hand, the receivers R1 and R2 have an opportunity to transmit a NAK, and thus can measure their own RTT values. Also, all the receivers including the receiver R3 that does not transmit the NAK packet can update the UP Worst RTT value by the retransmission data.

(RTT Measurement Using Tree Structure) Next, a case where the method of the present invention is used for RTT measurement using a tree structure will be considered. In this case, since NAK aggregation and NAK suppression are simultaneously used, useless packets flowing in the network can be reduced, and RTT measurement can be performed irregularly in a portion where congestion occurs in the network. And more efficiently,
In addition, the RTT value can be measured tightly. In this case, upon detecting a packet loss, the receiver transmits a NAK packet by unicast to the upstream node. The receiver adds the time stamp of the transmission time and the RTT value measured last time by itself to the NAK packet, and transmits the NAK packet in the same manner as in the method of measuring the RTT using the NAK packet performed only between the sender and the receiver. When the upstream node receives the NAK packet, it compares the RTT value added to the NAK packet with the Down Worst RTT value held by itself and updates the Down Worst RTT value if necessary. Next, N added to the NAK packet
The AK count is compared with the NAK count held by itself. NAK added to NAK packet
If the count is larger, N
Since the AK packet needs to be transmitted, a time stamp of the transmission time and the RTT value between itself and the upstream node are added to the NAK packet and transmitted to the upstream node. At the same time, it is necessary to multicast the NAK packet to the subordinate receivers for NAK suppression.
Address of the recipient who sent the K packet, Down
WorstRTT value and RT from itself to sender
A T value is added and the data is transmitted to a subordinate receiver by multicast. Among the recipients who have received the multicast NAK packet, those who have not yet transmitted the NAK packet suppress NAK transmission by NAK suppression. N
The receiver that has transmitted the AK packet updates the RTT value from itself to the upstream node. Also, all receivers update the RTT values from themselves to the sender by summing the RTT values from the upstream node to the sender in the multicast NAK with the sender.

The above is repeated hierarchically up to the sender.
Then, when the NAK reaches the sender, the sender is N
RTT value added to AK packet and Down W
Similarly, the value of Down Worst RTT is updated if necessary. Next, the sender checks whether it is necessary to retransmit the data requested by the NAK by comparing the number of retransmissions with the number of NAK transmissions added to the NAK packet, and retransmits if necessary. When performing this retransmission, the data packet is transmitted with the time stamp, the address of the NAK transmitting node or receiver, and the Down Worst RTT value added. The node receiving the retransmission data transmits the Down Worst added to the retransmission data packet.
The UP Worst RTT value is updated by RTT, and whether the retransmitted data packet is a response to the one requested by the NAK is checked by address comparison. Estimate.

Next, the node checks whether it is necessary to transfer the retransmitted data packet to the subordinate.
The TT value and the RTT value from itself to the sender are added to the data packet, and the data packet is distributed by multicast. At this time, the time stamp value and the address of the NAK sender are invalid. The node or receiver that has received the data updates the Down Worst RTT value, and updates the RTT value from the upper node to the sender added to the data packet and the RTT value from itself to the upper node held by itself. And the RTT value from itself to the sender is estimated. When hierarchical retransmission data distribution is performed in this manner, Down Worst R
The TT and RTT values are updated.

When the RTT value is estimated using the tree structure, unlike the case where the RTT measurement is performed only between the transmitting and receiving sides, the receiver that first transmits the unicast NAK packet is multicast from the upstream node to suppress the NAK. By receiving the delivered NAK packet, the RTT value with the upstream node can be estimated. Next, after the NAK reaches the sender, the retransmission data is delivered.
The upstream node can know the RTT value and Down Worst RTT value sequentially obtained by NAK transmission at the time of retransmission data reception. Therefore, in one NAK transmission, the RTT value from itself to the sender and Down Wo
Both rst RTT values can be updated. For this reason, when the RTT measurement is performed using the NAK packet, the delay time until an accurate RTT value is measured, which has been a problem of the related art, is eliminated, and the NAK packet is eliminated.
The RTT value can be estimated by receiving the multicast NAK and the retransmission data after the transmission.

Referring to FIG. 2 and Table 3, a method of estimating the RTT value using the NAK packet and the retransmission data packet when estimating the RTT value using the tree structure will be described. Table 3 is a table showing parameters added to the unicast NAK packet, the multicast NAK packet, and the retransmission data packet used in that case.

[0061]

[Table 3]

Referring to Table 3, using a tree structure, R
When measuring the TT value, the unicast NAK packet includes, as data, the requested data number, the time stamp of the transmission time, and the own RTT value obtained last time. Also, the multicast NAK packet includes the request data number, the time stamp in the unicast NAK, the address of the node or receiver that transmitted the unicast NAK packet, the Down Worst RTT value, and the RTT value from itself to the sender. Included as However, if there is no record transmitted upstream for the time stamp in the unicast NAK, specify an invalid value. Value.

Further, the retransmission data packet includes a data number and data, a time stamp in the unicast NAK, an address of a node or a receiver that transmitted the unicast NAK packet, Down Worst RTT.
The value and the RTT value from itself to the sender are included as data. However, the time stamp in the unicast NAK is set to NULL or -1 unless the sender transmits retransmission data, and the node that transmitted the unicast NAK packet or the address of the receiver is also a node other than the sender. Specifies an invalid value.
For the RTT value from itself to the sender, specify 0 for the sender.

FIG. 2 is a diagram showing an RTT measurement method using a NAK packet using a tree structure according to the present invention. (1) As in the case of FIG. 1, congestion occurs between the nodes N1 and N2 at a certain time, the RTT value changes from 10 ms to 25 ms, and packet loss occurs. (2) The receivers R1 and R2 detect packet loss and
After waiting a random time due to K suppression, the receiver R2 transmits a NAK packet first. (3) The receiver R2 NAKs the time stamp of the transmission time and the UP RTT value (30 ms) previously measured by the receiver R2.
The NAK packet is added to the packet and transmitted to the node N2 by unicast. (4) When receiving the NAK packet, the node N2 receives the Down Worst RTT (30 m
s) and the RTT value (30) added to the NAK packet.
ms) and, if necessary, Down Worst
Update the RTT value. (5) The node N2 is the N added to the NAK packet.
By comparing the AK count number with the NAK count number of the held NAK information, it is determined whether or not the NAK packet should be transmitted upstream. (6) If the NAK packet needs to be transmitted upstream, the time stamp of the time when the node N2 transmits the NAK packet and the UP RTT value (10
ms) is added to the NAK packet, and this packet is transmitted. In order to suppress NAK, the receiver under the
Since it is necessary to multicast the K packet, the time stamp added to the NAK packet from the receiver R2, the address of the receiver R2, Down Worst
RTT value (30 ms) and R from itself to sender
The TT value (20 ms) is added to the NAK packet, and this packet is multicast to the subordinate receiver. (7) When the receivers R1 and R2 receive the multicast NAK packet, Dow added to the NAK packet
UP Wors with n Worst RTT value (30ms)
Update the t RTT value (30 ms). Also, the receiver R
2 is the unicast NA added to the NAK packet
Since the sender's address matches his own,
The UP RTT value (30 ms) is estimated from the difference between the reception time of the NAK packet and the time stamp value added to the NAK packet. Finally, the receivers R1 and R2 estimate the RTT value from themselves to the sender by taking the sum of the RTT value from the node added to the NAK packet to the sender and the UP RTT value of the receiver itself. (8) The node N1 is a unicast NA from the node N2.
Upon receiving the K packet, the node N2 performs the same process as when receiving the unicast NAK packet from the receiver R2. (9) The node N2 receives the multicast N from the node N1.
When the AK packet is received, the own UP Worst RTT value (15 ms) is updated with the Down Worst RTT value (15 ms) added to the NAK packet, and the unicast NA added to the NAK packet is updated.
In order for the address of the K sender to match its own address, it estimates its UP RTT value (25 ms). Finally, it is checked whether or not the NAK packet has been transferred upstream, and since the NAK packet has been transferred, the NAK packet is discarded. (10) Upon receiving the multicast NAK packet from the node N1, the node N3 receives the Down Worst RTT value (15 ms) added to the NAK packet.
And updates its own UP Worst RTT value (15 ms), and calculates the sender's UP RTT value (15 ms) by adding the RTT value to the sender (10 ms) added to the multicast NAK packet from node N1 to the sender. Estimate the RTT value up to. Finally, it is checked whether or not the NAK packet has been transferred upstream. (11) Since the node N3 has not forwarded the NAK packet to the upstream, and has not multicast-distributed the NAK packet under the node N3, the Down Worst RTT value (35 ms) and the RTT value from itself to the sender (25
ms) is added to the NAK packet, and multicast transmission is performed under the packet. (12) When receiving the multicast NAK packet, the receiver R3 receives the Down added to the NAK packet.
Worst RTT value (35ms) and own UPWor
Update the st RTT value (35 ms). (13) The sender is a unicast NAK from node N1
When a packet is received, the UP RTT value (10 ms) added to the NAK packet and the D
Compare with the Own Worst RTT value (10 ms) and, if necessary, update the Down Worst RTT value. (14) Does the sender need to resend data? NAK
The NAK count added to the packet is compared with the number of data retransmissions, and if necessary, the data is retransmitted. (15) When retransmitting data, the sender transmits the time stamp added to the NAK packet, the sender address of the NAK, and the Down Worst RTT value (10
ms) is added to the retransmission data packet and transmitted by multicast. (16) Upon receiving the retransmission data packet, the node N1 has Down Wor added to the retransmission data packet.
Up Worst of own with st RTT value (10 ms)
The RTT value (10 ms) is updated, and since the NAK sender address added to the NAK packet matches its own address, its own UP RTT value (10 ms) is estimated. Finally, the sum of the RTT value up to the sender added to the retransmission data packet (in this case, 0 ms because it is direct retransmission data from the sender) and its own UPRTT value gives the RTT from itself to the sender. Value (10m
s) is estimated. (17) Since the node N1 is transmitting the NAK packet upstream, the node N1 multicasts retransmission data under the node N1. In this case, Down Worst held by itself is
RTT value (15 ms) and RTT value up to the sender (1
0 ms) is added to the retransmission data packet and transmitted by multicast. (18) When receiving the retransmission data packet, the node N3 receives the Down Wor added to the retransmission data packet.
own UP Worst with st RTT value (15ms)
The RTT value (15 ms) is updated, the sum of the RTT value of the sender (10 ms) added to the data packet to the sender (10 ms) and the own UP RTT value (20 ms) is calculated, and the RTT value of the sender to the sender (25 ms) is added. ). Finally, investigate whether the data should be distributed to the subordinates.
In this case, since the NAK packet has not been transmitted upstream, the data is discarded without being distributed to the subordinate. (19) Upon receiving the retransmitted data packet, the node N2 adds Down Wor added to the retransmitted data packet.
own UP Worst with st RTT value (15ms)
The RTT value (15 ms) is updated, and the RTT value (35 ms) from itself to the sender is calculated by adding the RTT value (10 ms) to the sender added to the data packet (10 ms) and the own UP RTT value (25 ms). estimate. Finally, investigate whether the data should be distributed to the subordinates. In this case, since the NAK packet is transmitted upstream, it is necessary to distribute the data under the NAK packet.
The TT value (30 ms) and the RTT value from itself to the sender (35 ms) are added to the data packet, and the data packet is multicast-distributed under the control. (20) Upon receiving the retransmitted data packet, the receivers R1 and R2 receive the Down added to the retransmitted data packet.
Worst RTT value (30ms) and own UPWor
The st RTT value is updated, the RTT value (35 ms) to the sender added to the data packet and the own UP
The RTT value from itself to the sender is estimated by the sum with the RTT value. (21) By processing these NAK and retransmission data,
The receivers R1 and R2 can estimate an accurate RTT value up to the sender, and the RTT value can be estimated by one retransmission process. (22) Down Wors held by the node N1
Although the t RTT value is not an accurate value, the next time the NAK is transmitted via the node N2, this value will also be an accurate value. (23) In Table 3, it is described that an invalid value is specified for data, but there is no particular need to change it. If not changed, the address will not be processed because the address does not match under the subordinate.

(RTT Measurement for Performing Local Recovery) Finally, a method of measuring RTT using NAK and retransmission data for performing local recovery will be described. This means that the node performing the local recovery plays the role of the sender of the RTT measurement method in the above-mentioned tree structure, and the other parts are almost the same. The difference is that when the sender performs retransmission, the RTT value from the sender to the sender is set to 0 and added to the retransmitted data packet and transmitted. , The RTT value up to the sender.

Similarly, when a node device in the network monitors data reception, and if a loss of data can be found in the node device, the node device performs the same processing as that of the above-mentioned receiver, thereby An accurate measurement of the RTT value becomes possible. Note that the original data (not the retransmission data according to the NAK packet but the data transmitted first from the sender) is used for the notification of the RTT value (Down Worst RTT value and the RTT value of the worst recipient under the subordinate) from the upstream node. This is not used because the RTT value added to the data is rewritten in each node in the network in the present method, and the recalculation of the checksum and the like is required each time, so that the processing overhead in the network node is considered to be large. It is.

When the RTT measurement is performed only between the sender and the receiver, the RTT of the farthest receiver under the original data in the original data is used.
The T value may be notified. This is because there is no processing in the network node. If the processing at the network node is greatly accelerated in the future and the processing delay is almost eliminated, the RTT value of the farthest receiver and the RTT value of the farthest receiver can be obtained even if the RTT measurement is performed using the tree structure.
The RTT value from the node device to the sender may be notified.

FIGS. 3, 4 and 5 show processing at the time of receiving a unicast NAK, receiving a multicast NAK, and receiving retransmission data at a node (sender, receiver, router) when this method is applied. Are respectively shown. FIG. 3 shows a unicast N at each node.
The processing at the time of receiving AK is shown. In the figure, when a unicast NAK packet is received (step S3)
01), first, the RTT value is updated (step S)
302). Next, if necessary, NAK suppression processing is performed (step S303). In this case, a multicast NAK packet is transmitted.

Further, it is determined whether retransmission data is stored in the node (step S304). When retransmission data is stored in the node, retransmission data distribution is performed if necessary (step S304 → S305). In this case, the retransmission data is transmitted by multicast. On the other hand, if the retransmission data is not stored in the node, NAK aggregation processing is performed if necessary (step S304 → S30).
6). In this case, a NAK packet is transmitted by unicast.

FIG. 4 shows a process when each node receives a multicast NAK. In the figure,
Upon receiving a multicast NAK packet (step S401), first, an RTT value update process is performed (step S402). Next, if necessary, NAK suppression processing is performed (step S403). In this case, N
Send an AK packet.

FIG. 5 shows processing at the time of receiving retransmission data in each node. In the figure, when retransmission data is received (step S501), an RTT value update process is first performed (step S502). Next, if necessary, retransmission data distribution is performed (step S503). In this case, the retransmission data is transmitted by multicast. As described above, in the processing of FIGS. 3 to 5, the RTT updating processing is first performed when a packet is received. Therefore, NA
Even if a packet is discarded due to K aggregation or the like, R
TT update processing can be performed.

As described above, in the RTT measurement method according to the present invention, in a network including a plurality of nodes, a retransmission request packet transmitted to an upstream node when a data packet is lost and a downstream node responds to the packet to the downstream node. This is an RTT measurement method for measuring an RTT value, which is a round trip time of packet transmission and reception, using a downstream packet transmitted toward the network. When a loss of a data packet is detected,
A receiver node for transmitting, by unicast, a retransmission request packet including a time stamp value and an address of the own node to an upstream node, and a downstream packet including the time stamp value and the address when the retransmission request packet is received. And an upstream node transmitting the downstream packet to the downstream node in a multicast manner, wherein the receiver node measures an RTT value based on a time stamp value included in the downstream packet. . The downstream packet is transmitted from a retransmission request suppression packet transmitted from an upstream node for suppressing a retransmission request, or transmitted from a sender node which is a transmission source of the data packet in response to the retransmission request packet. This is a retransmission data packet.

When transmitting the downstream packet, the upstream node adds the latest RTT value from its own node to the sender node, and the node receiving the packet adds the latest RTT value to the An RTT value from the own node to the sender node is set with reference to an RTT value from a node to the upstream node.

Further, when transmitting the downstream packet, the upstream node adds an RTT value from its own node to the farthest node immediately below it, and the node receiving the downstream packet adds the RTT value to the node. The RTT value from the upstream node to the farthest node immediately below it is set by referring to the RTT value. Further, the upstream node transmits only the earliest retransmission request packet upstream, and discards other retransmission request packets.

The upstream node may be a node that holds data to be delivered and performs local recovery for delivering a data packet instead of the sender node when receiving the retransmission request packet. is there.

[0076]

As described above, according to the present invention, when congestion occurs in a network and a receiver detects a packet loss, it is necessary to perform a normal RTT measurement in addition to a normal periodic RTT measurement.
The RTT value can be measured using the K packet (unicast, multicast) and the retransmission data packet, so that the RTT value can be measured more accurately at the time of congestion where the RTT value changes drastically. That is, a node or a receiver that is considered to be congested in the transmission path from the sender and whose RTT value is likely to change is R
Many opportunities for measuring the TT value can be obtained, and conversely, a node or a receiver existing on a path without congestion has a normal R
Since only the TT measurement is performed, the density of the RTT measurement can be changed only in a necessary part according to the dynamic change of the network, and there is an effect that the RTT measurement can be performed efficiently.

Further, according to the present invention, each receiver can efficiently and accurately measure the RTT value at the time of congestion, and the sender or the node device can set the RT of the farthest receiver in the receiver group.
Estimate the T value and efficiently notify the subordinate node or receiver of the T value. Based on these values, a timer for NAK suppression and a timer for retransmitting NAK packets can be set. So useless N
The transmission of AK and the delivery of useless retransmission data due to the useless NAK reaching the sender and the node performing local recovery are reduced. Therefore, there is an effect that reliable multicast distribution can be efficiently performed.

Further, when performing congestion control in the reliable multicast, the receiver calculates its own throughput based on its own loss rate and RTT value, and transmits the value to the sender to transmit. In general, the receiver controls transmission according to the throughput of the worst receiver among all the receivers. On the other hand, in the present invention, since the estimation of the RTT value can be accurately performed, the selection of the worst receiver can be performed accurately. It is possible to prevent a phenomenon in which a receiver in poor condition frequently causes packet loss by transmitting to a receiver with a high throughput, and the data can be efficiently delivered. is there.

Even when reliable multicast communication is performed only between the sender and the receiver, when the node in the network supports it, and when NAK suppression or NAK aggregation is performed, the node performing local recovery is also used in the network. , The RTT value can be measured using NAKs (unicast, multicast) and retransmission data transmitted from the receiver upon a packet loss at the time of congestion.

Further, according to the present invention, there is an effect that the amount of packets transmitted to the network hardly changes, and the overhead of the data added to the packets is only slightly increased.

[Brief description of the drawings]

FIG. 1 is a diagram for explaining an RTT measurement method according to the present invention.

FIG. 2 is a diagram illustrating a case where an RTT value is estimated using a tree structure in the RTT measurement method according to the present invention.

FIG. 3 is a flowchart showing a unicast NAK reception process in a node.

FIG. 4 is a flowchart showing a multicast NAK reception process in a node.

FIG. 5 is a flowchart showing retransmission data reception processing in a node.

FIG. 6 is a diagram illustrating processing in a node at the time of NAK aggregation.

FIG. 7 is a diagram illustrating NAK suppression processing.

FIG. 8 is a diagram showing a local recovery process.

FIG. 9 is a diagram for explaining a problem when only NAK suppression is performed.

FIG. 10 is a diagram for explaining simultaneous use of NAK suppression and NAK aggregation.

FIG. 11 is a diagram illustrating a case where RTT measurement is performed only between a sender and a receiver.

FIG. 12 is a diagram illustrating a case where RTT measurement is performed using a tree structure.

FIG. 13 is a diagram showing a measurement delay in RTT measurement using a tree structure.

[Explanation of symbols]

 N, N1, N2, N3 Node R, R1, R2, R3 Receiver T Sender

 ──────────────────────────────────────────────────続 き Continued on the front page F term (reference) 5K014 AA01 DA02 FA05 FA11 5K030 GA14 HA08 HB15 LA02 LD06 MB06

Claims (14)

[Claims] In a network including a plurality of nodes, a retransmission request packet transmitted to an upstream node when a data packet is lost and a downstream packet transmitted to a downstream node in response to the packet are used. A RTT measurement method for measuring a round trip time of packet transmission and reception, wherein a receiver node that detects a loss of a data packet transmits a retransmission request packet including a time stamp value and an address of its own node. The upstream node that has received the retransmission request packet by multicasting transmits the downstream packet including the time stamp value and the address to the downstream node by multicast, and at the receiver node, the downstream packet Measure the RTT value based on the timestamp value contained in the RTT measurement method characterized by the above-mentioned. 2. The RTT measurement method according to claim 1, wherein the downstream packet is a retransmission request suppression packet transmitted from an upstream node for retransmission request suppression. 3. The RTT according to claim 1, wherein the downstream packet is a retransmission data packet transmitted from a sender node that is a source of the data packet in response to the retransmission request packet. Measuring method. 4. The upstream node adds the latest RTT value from its own node to the sender node when transmitting the downstream packet, and the node receiving the packet adds the latest RTT value to the The RTT measurement method according to any one of claims 1 to 3, wherein the RTT value from the own node to the sender node is set with reference to the RTT value from a node to the upstream node. . 5. The upstream node, when transmitting the downstream packet, adds an RTT value from its own node to a farthest node immediately below the node, and the node receiving the downstream packet adds the RTT value to the downstream node. The RTT according to any one of claims 1 to 4, wherein an RTT value from an upstream node to a farthest node immediately below the node is set with reference to the RTT value.
TT measurement method. 6. The RTT measurement according to claim 1, wherein the upstream node transmits only the earliest retransmission request packet upstream and discards other retransmission request packets. Method. 7. The upstream node holds data to be distributed and, when receiving the retransmission request packet,
7. The method according to claim 1, wherein a local recovery for delivering a data packet is performed on behalf of the sender node.
RTT measurement method according to any one of the above. 8. In a network including a plurality of nodes, a retransmission request packet transmitted to an upstream node when a data packet is lost and a downstream packet transmitted to a downstream node in response to the packet are used. An RTT measurement system for measuring an RTT value, which is a round trip time of packet transmission and reception, when detecting a loss of a data packet, upstream of a retransmission request packet including a time stamp value and an address of its own node by unicast. A receiver node for transmitting to the node, and an upstream node for transmitting a downstream packet including the time stamp value and the address to the downstream node by multicast when receiving the retransmission request packet, the receiver node At the time based on the time stamp value included in the downstream packet. RTT
An RTT measurement system for measuring a value. 9. The RTT measurement system according to claim 8, wherein the downstream packet is a retransmission request suppression packet transmitted from an upstream node for retransmission request suppression. 10. The RTT according to claim 8, wherein the downstream packet is a retransmission data packet transmitted from a sender node that is a source of the data packet in response to the retransmission request packet. Measurement system. 11. The upstream node adds the latest RTT value from its own node to the sender node when transmitting the downstream packet, and the node receiving the packet adds the latest RTT value to the The RTT measurement system according to any one of claims 8 to 10, wherein an RTT value from the own node to the sender node is set with reference to an RTT value from a node to the upstream node. . 12. The upstream node, when transmitting the downstream packet, adds an RTT value from its own node to the farthest node immediately below the node, and the node that has received the downstream packet updates the RTT value. The RTT measurement system according to any one of claims 8 to 11, wherein an RTT value from the upstream node to a farthest node immediately below the upstream node is set with reference to the RTT value. 13. The apparatus according to claim 8, wherein the upstream node transmits only the earliest retransmission request packet upstream and discards other retransmission request packets.
RTT measurement system according to any one of the above. 14. The method according to claim 11, wherein the upstream node holds data to be delivered, and when receiving the retransmission request packet, performs local recovery for delivering a data packet instead of the sender node. Claim 8
The RTT measurement system according to any one of claims 13 to 13.